diff options
author | Christophe Lyon <christophe.lyon@linaro.org> | 2014-06-13 11:09:24 +0200 |
---|---|---|
committer | Christophe Lyon <christophe.lyon@linaro.org> | 2014-06-13 11:09:24 +0200 |
commit | 66f9ff92503cbf8efe9670ff13cd2ef8df8c320d (patch) | |
tree | 117c107ebda267cf27d1d996b10ac73c2e61f372 | |
parent | 8e606b4b671544f55b807eea103a6083dea33ec0 (diff) | |
download | platform_external_arm-neon-tests-66f9ff92503cbf8efe9670ff13cd2ef8df8c320d.tar.gz platform_external_arm-neon-tests-66f9ff92503cbf8efe9670ff13cd2ef8df8c320d.tar.bz2 platform_external_arm-neon-tests-66f9ff92503cbf8efe9670ff13cd2ef8df8c320d.zip |
Add a new expected results file, useful for the conversion of the
tests into GCC's dejagnu testsuite.
-rw-r--r-- | expected_input4gcc.txt | 7063 |
1 files changed, 7063 insertions, 0 deletions
diff --git a/expected_input4gcc.txt b/expected_input4gcc.txt new file mode 100644 index 0000000..485ba33 --- /dev/null +++ b/expected_input4gcc.txt @@ -0,0 +1,7063 @@ + +VLD1/VLD1Q output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, 0xc1600000, 0xc1500000 }; + +VADD/VADDQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xffec, 0xffed, 0xffee, 0xffef }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff3, 0xfffffff4 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x54 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x4, 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xe, 0xf, 0x10, 0x11 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x18, 0x19 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xe6, 0xe7, 0xe8, 0xe9, 0xea, 0xeb, 0xec, 0xed, 0xee, 0xef, 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffdc, 0xffdd, 0xffde, 0xffdf, 0xffe0, 0xffe1, 0xffe2, 0xffe3 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffd2, 0xffffffd3, 0xffffffd4, 0xffffffd5 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x8, 0x9 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xfc, 0xfd, 0xfe, 0xff, 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7, 0xfff8, 0xfff9, 0xfffa }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x27, 0x28, 0x29, 0x2a }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff3, 0xfffffffffffffff4 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x40d9999a, 0x40d9999a }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x41100000, 0x41100000, 0x41100000, 0x41100000 }; + +VLD1_LANE/VLD1_LANEQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xf0, 0xaa }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xfff0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xaaaaaaaa, 0xfffffff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xf0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xfff0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xaaaaaaaa, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xf0 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xfff0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xaaaaaaaa, 0xc1800000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xf0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xfff0, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xfffffff0, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xaaaaaaaaaaaaaaaa, 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xf0, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xfff0, 0xaaaa }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xfffffff0, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0xaaaaaaaaaaaaaaaa }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xf0, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xfff0, 0xaaaa }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xc1800000, 0xaaaaaaaa }; + +VLD1_DUP/VLD1_DUPQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1800000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff0, 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff0, 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1800000, 0xc1800000, 0xc1800000 }; + +VLD1_DUP/VLD1_DUPQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1700000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff1, 0xfffffff1, 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff1, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff1, 0xfffffff1, 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff1, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1700000, 0xc1700000, 0xc1700000, 0xc1700000 }; + +VLD1_DUP/VLD1_DUPQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1600000, 0xc1600000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff2, 0xfffffff2, 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff2, 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff2, 0xfffffff2, 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff2, 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1600000, 0xc1600000, 0xc1600000, 0xc1600000 }; + +VDUP/VDUPQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1800000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff0, 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff0, 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1800000, 0xc1800000, 0xc1800000 }; + +VDUP/VDUPQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1700000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff1, 0xfffffff1, 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff1, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff1, 0xfffffff1, 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff1, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1700000, 0xc1700000, 0xc1700000, 0xc1700000 }; + +VDUP/VDUPQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1600000, 0xc1600000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff2, 0xfffffff2, 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff2, 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff2, 0xfffffff2, 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff2, 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1600000, 0xc1600000, 0xc1600000, 0xc1600000 }; + +VMOV/VMOVQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1800000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff0, 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff0, 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1800000, 0xc1800000, 0xc1800000 }; + +VMOV/VMOVQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1700000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff1, 0xfffffff1, 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff1, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff1, 0xfffffff1, 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff1, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1700000, 0xc1700000, 0xc1700000, 0xc1700000 }; + +VMOV/VMOVQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1600000, 0xc1600000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff2, 0xfffffff2, 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff2, 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff2, 0xfffffff2, 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff2, 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1600000, 0xc1600000, 0xc1600000, 0xc1600000 }; + +VGET_HIGH output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1600000, 0xc1500000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VGET_LOW output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; + +VQDMLAL_LANE output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7c1e, 0x7c1f, 0x7c20, 0x7c21 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x7c1e, 0x7c1f }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; + +VQDMLAL_LANE (mul with input=0) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; + +VQDMLAL_LANE (check mul cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffef, 0x7ffffff0, 0x7ffffff1, 0x7ffffff2 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x7fffffffffffffef, 0x7ffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; + +VQDMLSL_LANE output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffff83c2, 0xffff83c3, 0xffff83c4, 0xffff83c5 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffff83c2, 0xffffffffffff83c3 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; + +VQDMLSL_LANE (mul with input=0) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; + +VQDMLSL_LANE (check mul cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x8000000000000000, 0x8000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; + +VQDMLAL_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x1684, 0x1685, 0x1686, 0x1687 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x21ce, 0x21cf }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; + +VQDMLAL_N (check mul cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffef, 0x7ffffff0, 0x7ffffff1, 0x7ffffff2 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x7fffffffffffffef, 0x7ffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; + +VQDMLSL_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffe95c, 0xffffe95d, 0xffffe95e, 0xffffe95f }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffde12, 0xffffffffffffde13 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; + +VQDMLSL_N (check mul cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x8000000000000000, 0x8000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VEXT/VEXTQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf7, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff3, 0x22, 0x22, 0x22 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff1, 0x33 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf6, 0xf7, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff2, 0xfff3, 0x66, 0x66 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff1, 0x77 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf6, 0xf7, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff2, 0xfff3, 0x66, 0x66 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1700000, 0x42066666 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xfe, 0xff, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff7, 0x22, 0x22, 0x22, 0x22, 0x22, 0x22, 0x22 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff3, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff1, 0x44 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xfc, 0xfd, 0xfe, 0xff, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff6, 0xfff7, 0x66, 0x66, 0x66, 0x66, 0x66, 0x66 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff3, 0x77, 0x77, 0x77 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff1, 0x88 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xfc, 0xfd, 0xfe, 0xff, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff6, 0xfff7, 0x66, 0x66, 0x66, 0x66, 0x66, 0x66 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1500000, 0x4204cccd, 0x4204cccd, 0x4204cccd }; + +VSHR_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf8, 0xf9, 0xf9, 0xfa, 0xfa, 0xfb, 0xfb }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffffc, 0xfffffffc }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x3c, 0x3c, 0x3c, 0x3c, 0x3d, 0x3d, 0x3d, 0x3d }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x1ffe, 0x1ffe, 0x1ffe, 0x1ffe }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x7ffffff, 0x7ffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x7fffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf8, 0xf8, 0xf9, 0xf9, 0xfa, 0xfa, 0xfb, 0xfb, 0xfc, 0xfc, 0xfd, 0xfd, 0xfe, 0xfe, 0xff, 0xff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffffc, 0xfffffffc, 0xfffffffc, 0xfffffffc }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x3c, 0x3c, 0x3c, 0x3c, 0x3d, 0x3d, 0x3d, 0x3d, 0x3e, 0x3e, 0x3e, 0x3e, 0x3f, 0x3f, 0x3f, 0x3f }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x1ffe, 0x1ffe, 0x1ffe, 0x1ffe, 0x1ffe, 0x1ffe, 0x1ffe, 0x1ffe }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x7ffffff, 0x7ffffff, 0x7ffffff, 0x7ffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VSHRN_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf8, 0xf9, 0xf9, 0xfa, 0xfa, 0xfb, 0xfb }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff8, 0xfff8, 0xfff9, 0xfff9 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffffc, 0xfffffffc }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xfc, 0xfc, 0xfc, 0xfc, 0xfd, 0xfd, 0xfd, 0xfd }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfffe, 0xfffe, 0xfffe, 0xfffe }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffffe, 0xfffffffe }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHRN_N (with input = 0) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHRN_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf9, 0xf9, 0xfa, 0xfa, 0xfb, 0xfb, 0xfc }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff8, 0xfff9, 0xfff9, 0xfffa }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffffc, 0xfffffffc }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xfc, 0xfc, 0xfd, 0xfd, 0xfd, 0xfd, 0xfe, 0xfe }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfffe, 0xfffe, 0xfffe, 0xfffe }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffffe, 0xfffffffe }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHRN_N (with large shift amount) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; + +VQRSHRN_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf9, 0xf9, 0xfa, 0xfa, 0xfb, 0xfb, 0xfc }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff8, 0xfff9, 0xfff9, 0xfffa }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffffc, 0xfffffffc }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; + +VQRSHRN_N (check saturation: shift by 3) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; + +VQRSHRN_N (check saturation: shift by max) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VSET_LANE/VSET_LANEQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0x11 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0x22 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0x33 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x44 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0x55, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0x66, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0x77 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x88 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0x55, 0xf7 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0x66, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0x4204cccd }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0x99 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xaa, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xbb }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0xcc }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xdd, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xee, 0xfff7 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xff, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0x11 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xdd, 0xff }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xee, 0xfff7 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, 0xc1600000, 0x41333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; + +VQSUB/VQSUBQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xdf, 0xe0, 0xe1, 0xe2, 0xe3, 0xe4, 0xe5, 0xe6 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xffce, 0xffcf, 0xffd0, 0xffd1 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xffffffbd, 0xffffffbe }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffffac }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x9b, 0x9c, 0x9d, 0x9e, 0x9f, 0xa0, 0xa1, 0xa2 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xff8a, 0xff8b, 0xff8c, 0xff8d }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffff79, 0xffffff7a }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffff68 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xdf, 0xe0, 0xe1, 0xe2, 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea, 0xeb, 0xec, 0xed, 0xee }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffce, 0xffcf, 0xffd0, 0xffd1, 0xffd2, 0xffd3, 0xffd4, 0xffd5 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffbd, 0xffffffbe, 0xffffffbf, 0xffffffc0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffffac, 0xffffffffffffffad }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x9b, 0x9c, 0x9d, 0x9e, 0x9f, 0xa0, 0xa1, 0xa2, 0xa3, 0xa4, 0xa5, 0xa6, 0xa7, 0xa8, 0xa9, 0xaa }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xff8a, 0xff8b, 0xff8c, 0xff8d, 0xff8e, 0xff8f, 0xff90, 0xff91 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffff79, 0xffffff7a, 0xffffff7b, 0xffffff7c }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffff68, 0xffffffffffffff69 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffffac }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffff68 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffffac, 0xffffffffffffffad }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffff68, 0xffffffffffffff69 }; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x8000000000000000 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x7fffffffffffffff, 0x7fffffffffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x0 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +VECT_VAR_DECL(expected,int,8,8) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; + +VQDMULH output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; + +VQDMULH output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; + +VQDMULH_LANE output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; + +VQDMULH_LANE (check mul cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x19, 0x19, 0x19, 0x19 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x4, 0x4 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xa, 0xa, 0xa, 0xa }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; + +VQDMULH_N (check mul cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; + +VQDMULL output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x200, 0x1c2, 0x188, 0x152 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x200, 0x1c2 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; + +VQDMULL (check mul cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x7fffffffffffffff, 0x7fffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; + +VQDMLAL output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7c1e, 0x7c1f, 0x7c20, 0x7c21 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x7c1e, 0x7c1f }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; + +VQDMLAL (check mul cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffef, 0x7ffffff0, 0x7ffffff1, 0x7ffffff2 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x7fffffffffffffef, 0x7ffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; + +VQDMLSL output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffff83c2, 0xffff83c3, 0xffff83c4, 0xffff83c5 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffff83c2, 0xffffffffffff83c3 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; + +VQDMLSL (check mul cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x8000000000000000, 0x8000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xff, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0xffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0xff, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0xffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xff, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0xff, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0xff, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0xff, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; + +VBSL/VBSLQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf6, 0xf6, 0xf6, 0xf6 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff0, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffffd }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf3, 0xf3, 0xf3, 0xf3, 0xf7, 0xf7, 0xf7, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff0, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffff1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf3, 0xf3, 0xf3, 0xf3, 0xf7, 0xf7, 0xf7, 0xf7 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff0, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800004, 0xc1700004 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf6, 0xf6, 0xf6, 0xf6, 0xf2, 0xf2, 0xf2, 0xf2, 0xf6, 0xf6, 0xf6, 0xf6 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff0, 0xfff2, 0xfff2, 0xfff4, 0xfff4, 0xfff6, 0xfff6 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff0, 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffffd, 0xfffffffffffffffd }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf3, 0xf3, 0xf3, 0xf3, 0xf7, 0xf7, 0xf7, 0xf7, 0xf3, 0xf3, 0xf3, 0xf3, 0xf7, 0xf7, 0xf7, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff0, 0xfff2, 0xfff2, 0xfff4, 0xfff4, 0xfff6, 0xfff6 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff0, 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf3, 0xf3, 0xf3, 0xf3, 0xf7, 0xf7, 0xf7, 0xf7, 0xf3, 0xf3, 0xf3, 0xf3, 0xf7, 0xf7, 0xf7, 0xf7 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff0, 0xfff2, 0xfff2, 0xfff4, 0xfff4, 0xfff6, 0xfff6 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800001, 0xc1700001, 0xc1600001, 0xc1500001 }; + +VSHL/VSHLQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xe0, 0xe2, 0xe4, 0xe6, 0xe8, 0xea, 0xec, 0xee }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xff80, 0xff88, 0xff90, 0xff98 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffff000, 0xfffff100 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffff80 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xe0, 0xe2, 0xe4, 0xe6, 0xe8, 0xea, 0xec, 0xee }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xff80, 0xff88, 0xff90, 0xff98 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffff000, 0xfffff100 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffff80 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x20, 0x40, 0x60, 0x80, 0xa0, 0xc0, 0xe0, 0x0, 0x20, 0x40, 0x60, 0x80, 0xa0, 0xc0, 0xe0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x1000, 0x2000, 0x3000, 0x4000, 0x5000, 0x6000, 0x7000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x40000000, 0x80000000, 0xc0000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x8000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x20, 0x40, 0x60, 0x80, 0xa0, 0xc0, 0xe0, 0x0, 0x20, 0x40, 0x60, 0x80, 0xa0, 0xc0, 0xe0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x1000, 0x2000, 0x3000, 0x4000, 0x5000, 0x6000, 0x7000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x40000000, 0x80000000, 0xc0000000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x8000000000000000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VSHL/VSHLQ (large shift amount) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VSHL/VSHLQ (negative shift amount) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf8, 0xf9, 0xf9, 0xfa, 0xfa, 0xfb, 0xfb }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff8, 0xfff8, 0xfff9, 0xfff9 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffffc, 0xfffffffc }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x78, 0x78, 0x79, 0x79, 0x7a, 0x7a, 0x7b, 0x7b }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x7ff8, 0x7ff8, 0x7ff9, 0x7ff9 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x3ffffffc, 0x3ffffffc }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xfc, 0xfc, 0xfc, 0xfc, 0xfd, 0xfd, 0xfd, 0xfd, 0xfe, 0xfe, 0xfe, 0xfe, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffffe, 0xfffffffe, 0xfffffffe, 0xfffffffe }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x3c, 0x3c, 0x3c, 0x3c, 0x3d, 0x3d, 0x3d, 0x3d, 0x3e, 0x3e, 0x3e, 0x3e, 0x3f, 0x3f, 0x3f, 0x3f }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x7ff, 0x7ff, 0x7ff, 0x7ff, 0x7ff, 0x7ff, 0x7ff, 0x7ff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x1ffffffe, 0x1ffffffe, 0x1ffffffe, 0x1ffffffe }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x7ffffffffffffff, 0x7ffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VSHL_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xe0, 0xe2, 0xe4, 0xe6, 0xe8, 0xea, 0xec, 0xee }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xffe0, 0xffe2, 0xffe4, 0xffe6 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xffffff80, 0xffffff88 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffffc0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xc0, 0xc4, 0xc8, 0xcc, 0xd0, 0xd4, 0xd8, 0xdc }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xff00, 0xff10, 0xff20, 0xff30 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffff80, 0xffffff88 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffffe0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x20, 0x40, 0x60, 0x80, 0xa0, 0xc0, 0xe0, 0x0, 0x20, 0x40, 0x60, 0x80, 0xa0, 0xc0, 0xe0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffe0, 0xffe2, 0xffe4, 0xffe6, 0xffe8, 0xffea, 0xffec, 0xffee }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffc0, 0xffffffc4, 0xffffffc8, 0xffffffcc }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffffc0, 0xffffffffffffffc4 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xc0, 0xc4, 0xc8, 0xcc, 0xd0, 0xd4, 0xd8, 0xdc, 0xe0, 0xe4, 0xe8, 0xec, 0xf0, 0xf4, 0xf8, 0xfc }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xff80, 0xff88, 0xff90, 0xff98, 0xffa0, 0xffa8, 0xffb0, 0xffb8 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffc0, 0xffffffc4, 0xffffffc8, 0xffffffcc }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffe0, 0xffffffffffffffe2 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; + +VQSHL/VQSHLQ (with input = 0) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; + +VQSHL/VQSHLQ (input 0 and negative shift amount) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; + +VQSHL/VQSHLQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xe0, 0xe2, 0xe4, 0xe6, 0xe8, 0xea, 0xec, 0xee }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xff80, 0xff88, 0xff90, 0xff98 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffff000, 0xfffff100 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffffe }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x1ffffffffffffffe }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x8000000000000000, 0x8000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; + +VQSHL/VQSHLQ (negative shift amount) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf8, 0xf9, 0xf9, 0xfa, 0xfa, 0xfb, 0xfb }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfffc, 0xfffc, 0xfffc, 0xfffc }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffffe, 0xfffffffe }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x78, 0x78, 0x79, 0x79, 0x7a, 0x7a, 0x7b, 0x7b }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3ffc, 0x3ffc, 0x3ffc, 0x3ffc }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x1ffffffe, 0x1ffffffe }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x1f, 0x1f, 0x1f, 0x1f, 0x1f, 0x1f, 0x1f, 0x1f }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x7ffff, 0x7ffff, 0x7ffff, 0x7ffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffff, 0xfffffffffff }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; + +VQSHL/VQSHLQ (large shift amount, negative input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x8000000000000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x8000000000000000, 0x8000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; + +VQSHL/VQSHLQ (check cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x3f, 0x3f, 0x3f, 0x3f, 0x3f, 0x3f, 0x3f, 0x3f }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3fff, 0x3fff, 0x3fff, 0x3fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x3fffffff, 0x3fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3fffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x7fffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x3f, 0x3f, 0x3f, 0x3f, 0x3f, 0x3f, 0x3f, 0x3f, 0x3f, 0x3f, 0x3f, 0x3f, 0x3f, 0x3f, 0x3f, 0x3f }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3fff, 0x3fff, 0x3fff, 0x3fff, 0x3fff, 0x3fff, 0x3fff, 0x3fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x3fffffff, 0x3fffffff, 0x3fffffff, 0x3fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3fffffffffffffff, 0x3fffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x7fffffffffffffff, 0x7fffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; + +VQSHL/VQSHLQ (large shift amount, positive input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x7fffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x7fffffffffffffff, 0x7fffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; + +VQSHL/VQSHLQ (check saturation on 64 bits) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x8000000000000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x7fffffffffffffff, 0x7fffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; + +VQSHL_N/VQSHLQ_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xc0, 0xc4, 0xc8, 0xcc, 0xd0, 0xd4, 0xd8, 0xdc }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xffe0, 0xffe2, 0xffe4, 0xffe6 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xffffffe0, 0xffffffe2 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffffc0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xc0, 0xc4, 0xc8, 0xcc, 0xd0, 0xd4, 0xd8, 0xdc, 0xe0, 0xe4, 0xe8, 0xec, 0xf0, 0xf4, 0xf8, 0xfc }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffe0, 0xffe2, 0xffe4, 0xffe6, 0xffe8, 0xffea, 0xffec, 0xffee }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffe0, 0xffffffe2, 0xffffffe4, 0xffffffe6 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffffc0, 0xffffffffffffffc4 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; + +VQSHL_N/VQSHLQ_N (check saturation with large positive input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x7fffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x7fffffffffffffff, 0x7fffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHL/VRSHLQ (with input = 0) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHL/VRSHLQ (input 0 and negative shift amount) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHL/VRSHLQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xe0, 0xe2, 0xe4, 0xe6, 0xe8, 0xea, 0xec, 0xee }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xff80, 0xff88, 0xff90, 0xff98 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffff000, 0xfffff100 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffffe }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xe0, 0xe2, 0xe4, 0xe6, 0xe8, 0xea, 0xec, 0xee }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xff80, 0xff88, 0xff90, 0xff98 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffff000, 0xfffff100 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x1ffffffffffffffe }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x1000, 0x2000, 0x3000, 0x4000, 0x5000, 0x6000, 0x7000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x8000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x1000, 0x2000, 0x3000, 0x4000, 0x5000, 0x6000, 0x7000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x8000000000000000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHL/VRSHLQ (negative shift amount) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf9, 0xf9, 0xfa, 0xfa, 0xfb, 0xfb, 0xfc }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfffc, 0xfffc, 0xfffd, 0xfffd }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffffe, 0xfffffffe }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x78, 0x79, 0x79, 0x7a, 0x7a, 0x7b, 0x7b, 0x7c }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3ffc, 0x3ffc, 0x3ffd, 0x3ffd }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x1ffffffe, 0x1ffffffe }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x80000, 0x80000, 0x80000, 0x80000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x100000000000, 0x100000000000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHL/VRSHLQ (checking round_const overflow: shift by -1) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x4000, 0x4000, 0x4000, 0x4000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x40000000, 0x40000000 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x4000000000000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x8000000000000000 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x4000, 0x4000, 0x4000, 0x4000, 0x4000, 0x4000, 0x4000, 0x4000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x40000000, 0x40000000, 0x40000000, 0x40000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x4000000000000000, 0x4000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x8000000000000000, 0x8000000000000000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHL/VRSHLQ (checking round_const overflow: shift by -3) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x1000, 0x1000, 0x1000, 0x1000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x10000000, 0x10000000 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x1000000000000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x2000, 0x2000, 0x2000, 0x2000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x20000000, 0x20000000 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x2000000000000000 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x1000, 0x1000, 0x1000, 0x1000, 0x1000, 0x1000, 0x1000, 0x1000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x10000000, 0x10000000, 0x10000000, 0x10000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x1000000000000000, 0x1000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x2000, 0x2000, 0x2000, 0x2000, 0x2000, 0x2000, 0x2000, 0x2000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x20000000, 0x20000000, 0x20000000, 0x20000000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x2000000000000000, 0x2000000000000000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHL/VRSHLQ (checking negative shift amount as large as input vector width) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x1, 0x1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHL/VRSHLQ (large shift amount) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHL/VRSHLQ (large negative shift amount) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x1, 0x1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VLD2/VLD2Q chunk 0 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, 0xc1600000, 0xc1500000 }; + +VLD2/VLD2Q chunk 1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1600000, 0xc1500000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff8, 0xfff9, 0xfffa, 0xfffb, 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff4, 0xfffffff5, 0xfffffff6, 0xfffffff7 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff8, 0xfff9, 0xfffa, 0xfffb, 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff4, 0xfffffff5, 0xfffffff6, 0xfffffff7 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff8, 0xfff9, 0xfffa, 0xfffb, 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1400000, 0xc1300000, 0xc1200000, 0xc1100000 }; + +VLD3/VLD3Q chunk 0 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, 0xc1600000, 0xc1500000 }; + +VLD3/VLD3Q chunk 1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1600000, 0xc1500000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff8, 0xfff9, 0xfffa, 0xfffb, 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff4, 0xfffffff5, 0xfffffff6, 0xfffffff7 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff8, 0xfff9, 0xfffa, 0xfffb, 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff4, 0xfffffff5, 0xfffffff6, 0xfffffff7 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff8, 0xfff9, 0xfffa, 0xfffb, 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1400000, 0xc1300000, 0xc1200000, 0xc1100000 }; + +VLD3/VLD3Q chunk 2 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff8, 0xfff9, 0xfffa, 0xfffb }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff4, 0xfffffff5 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff8, 0xfff9, 0xfffa, 0xfffb }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff4, 0xfffffff5 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff8, 0xfff9, 0xfffa, 0xfffb }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1400000, 0xc1300000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff8, 0xfffffff9, 0xfffffffa, 0xfffffffb }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff8, 0xfffffff9, 0xfffffffa, 0xfffffffb }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1000000, 0xc0e00000, 0xc0c00000, 0xc0a00000 }; + +VLD4/VLD4Q chunk 0 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, 0xc1600000, 0xc1500000 }; + +VLD4/VLD4Q chunk 1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1600000, 0xc1500000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff8, 0xfff9, 0xfffa, 0xfffb, 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff4, 0xfffffff5, 0xfffffff6, 0xfffffff7 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff8, 0xfff9, 0xfffa, 0xfffb, 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff4, 0xfffffff5, 0xfffffff6, 0xfffffff7 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff8, 0xfff9, 0xfffa, 0xfffb, 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1400000, 0xc1300000, 0xc1200000, 0xc1100000 }; + +VLD4/VLD4Q chunk 2 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff8, 0xfff9, 0xfffa, 0xfffb }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff4, 0xfffffff5 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff8, 0xfff9, 0xfffa, 0xfffb }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff4, 0xfffffff5 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff8, 0xfff9, 0xfffa, 0xfffb }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1400000, 0xc1300000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff8, 0xfffffff9, 0xfffffffa, 0xfffffffb }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff8, 0xfffffff9, 0xfffffffa, 0xfffffffb }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1000000, 0xc0e00000, 0xc0c00000, 0xc0a00000 }; + +VLD4/VLD4Q chunk 3 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff6, 0xfffffff7 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff3 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff6, 0xfffffff7 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff3 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1200000, 0xc1100000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffffc, 0xfffffffd, 0xfffffffe, 0xffffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffffc, 0xfffffffd, 0xfffffffe, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc0800000, 0xc0400000, 0xc0000000, 0xbf800000 }; + +VDUP_LANE/VDUP_LANEQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1, 0xf1 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf7, 0xf7, 0xf7, 0xf7, 0xf7, 0xf7, 0xf7, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff3, 0xfff3, 0xfff3, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf7, 0xf7, 0xf7, 0xf7, 0xf7, 0xf7, 0xf7, 0xf7 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff3, 0xfff3, 0xfff3, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1700000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff3, 0xfff3, 0xfff3, 0xfff3, 0xfff3, 0xfff3, 0xfff3, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff1, 0xfffffff1, 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff0, 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5, 0xf5 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1700000, 0xc1700000, 0xc1700000, 0xc1700000 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x4000, 0x4000 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x7fffffffffffffff, 0x7fffffffffffffff }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x44000, 0x44000, 0x44000, 0x44000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xaa000, 0xaa000 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x7fffffffffffffff, 0x7fffffffffffffff }; + +VST1_LANE/VST1_LANEQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf7, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff3, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff1, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf6, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff2, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf6, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff2, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1700000, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xff, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff5, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff1, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff1, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xfa, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff4, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff3, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xfa, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff4, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1700000, 0x33333333, 0x33333333, 0x33333333 }; + +VSUB/VSUBQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xee, 0xef, 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xffffffed, 0xffffffee }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffff8c }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xdc, 0xdd, 0xde, 0xdf, 0xe0, 0xe1, 0xe2, 0xe3 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffd2, 0xffd3, 0xffd4, 0xffd5 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffc8, 0xffffffc9 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffffee }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff, 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8, 0x9 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x4, 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xe, 0xf, 0x10, 0x11 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffffd8, 0xffffffffffffffd9 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea, 0xeb, 0xec, 0xed, 0xee, 0xef, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffed, 0xffee, 0xffef, 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffb9, 0xffffffba, 0xffffffbb, 0xffffffbc }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffed, 0xffffffffffffffee }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc00ccccd, 0xc00ccccd }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc00ccccc, 0xc00ccccc, 0xc00ccccc, 0xc00ccccc }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; + +VQADD/VQADDQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x12, 0x13, 0x14, 0x15 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x23, 0x24 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x34 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf, 0x10 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x23, 0x24, 0x25, 0x26 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x34, 0x35 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x34 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x34, 0x35 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x8000000000000000 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x7fffffffffffffff, 0x7fffffffffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +VECT_VAR_DECL(expected,int,8,8) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; + +VABS/VABSQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x10, 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x10, 0xf, 0xe, 0xd }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x10, 0xf }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x10, 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x10, 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x10, 0xf, 0xe, 0xd }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x40133333, 0x40133333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x4059999a, 0x4059999a, 0x4059999a, 0x4059999a }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; + +VQABS/VQABSQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x10, 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x10, 0xf, 0xe, 0xd }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x10, 0xf }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x10, 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x10, 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x10, 0xf, 0xe, 0xd }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; + +VQABS/VQABSQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VCOMBINE output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0x22, 0x22, 0x22, 0x22 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0x44 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0x66, 0x66, 0x66, 0x66 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0x77, 0x77 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0x88 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0x66, 0x66, 0x66, 0x66 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, 0x40533333, 0x40533333 }; + +VMAX/VMAXQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf3, 0xf3, 0xf3, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf3, 0xf3, 0xf3, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff1, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1780000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf4, 0xf4, 0xf4, 0xf4, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff3, 0xfff3, 0xfff3, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff1, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf9, 0xf9, 0xf9, 0xf9, 0xf9, 0xf9, 0xf9, 0xf9, 0xf9, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff2, 0xfff2, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff1, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1680000, 0xc1680000, 0xc1600000, 0xc1500000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x7fc00000, 0x7fc00000, 0x7fc00000, 0x7fc00000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x7fc00000, 0x7fc00000, 0x7fc00000, 0x7fc00000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x7f800000, 0x7f800000, 0x7f800000, 0x7f800000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; + +VMIN/VMINQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf3, 0xf3, 0xf3, 0xf3 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf3, 0xf3, 0xf3, 0xf3 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff1, 0xfff1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1780000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf4, 0xf4, 0xf4, 0xf4, 0xf4, 0xf4, 0xf4, 0xf4, 0xf4, 0xf4, 0xf4 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff3, 0xfff3, 0xfff3, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xf9, 0xf9, 0xf9, 0xf9, 0xf9, 0xf9 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, 0xc1680000, 0xc1680000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x7fc00000, 0x7fc00000, 0x7fc00000, 0x7fc00000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x7fc00000, 0x7fc00000, 0x7fc00000, 0x7fc00000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xff800000, 0xff800000, 0xff800000, 0xff800000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; + +VNEG/VNEGQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x10, 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x10, 0xf, 0xe, 0xd }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x10, 0xf }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x10, 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x10, 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x10, 0xf, 0xe, 0xd }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc0133333, 0xc0133333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc059999a, 0xc059999a, 0xc059999a, 0xc059999a }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; + +VQNEG/VQNEGQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x10, 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x10, 0xf, 0xe, 0xd }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x10, 0xf }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x10, 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x10, 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x10, 0xf, 0xe, 0xd }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; + +VQNEG/VQNEGQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VMLAL output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xe907, 0xe908, 0xe909, 0xe90a, 0xe90b, 0xe90c, 0xe90d, 0xe90e }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x3e07, 0x3e08, 0x3e09, 0x3e0a }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3e07, 0x3e08 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3e07, 0x3e08, 0x3e09, 0x3e0a, 0x3e0b, 0x3e0c, 0x3e0d, 0x3e0e }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x3e07, 0x3e08, 0x3e09, 0x3e0a }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3e07, 0x3e08 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VMLSL output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x16d9, 0x16da, 0x16db, 0x16dc, 0x16dd, 0x16de, 0x16df, 0x16e0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffc1d9, 0xffffc1da, 0xffffc1db, 0xffffc1dc }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffc1d9, 0xffffffffffffc1da }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xc1d9, 0xc1da, 0xc1db, 0xc1dc, 0xc1dd, 0xc1de, 0xc1df, 0xc1e0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffc1d9, 0xffffc1da, 0xffffc1db, 0xffffc1dc }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffc1d9, 0xffffffffffffc1da }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VMLAL_LANE output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x3e07, 0x3e08, 0x3e09, 0x3e0a }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3e07, 0x3e08 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x3e07, 0x3e08, 0x3e09, 0x3e0a }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3e07, 0x3e08 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VMLSL_LANE output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffc1d9, 0xffffc1da, 0xffffc1db, 0xffffc1dc }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffc1d9, 0xffffffffffffc1da }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffc1d9, 0xffffc1da, 0xffffc1db, 0xffffc1dc }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffc1d9, 0xffffffffffffc1da }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VMLAL_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x595, 0x596, 0x597, 0x598 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xb3a, 0xb3b }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x10df, 0x10e0, 0x10e1, 0x10e2 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x10df, 0x10e0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VMLSL_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffa4b, 0xfffffa4c, 0xfffffa4d, 0xfffffa4e }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffff4a6, 0xfffffffffffff4a7 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffef01, 0xffffef02, 0xffffef03, 0xffffef04 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffef01, 0xffffffffffffef02 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VMOVL output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VMOVN output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VMULL output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x100, 0xe1, 0xc4, 0xa9, 0x90, 0x79, 0x64, 0x51 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x100, 0xe1, 0xc4, 0xa9 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x100, 0xe1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xe100, 0xe2e1, 0xe4c4, 0xe6a9, 0xe890, 0xea79, 0xec64, 0xee51 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffe00100, 0xffe200e1, 0xffe400c4, 0xffe600a9 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffe000000100, 0xffffffe2000000e1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x5500, 0x5501, 0x5504, 0x5505, 0x5510, 0x5511, 0x5514, 0x5515 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VMULL_LANE output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x4000, 0x4000, 0x4000, 0x4000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x2000, 0x2000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x4000, 0x4000, 0x4000, 0x4000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x2000, 0x2000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VREV16 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf1, 0xf0, 0xf3, 0xf2, 0xf5, 0xf4, 0xf7, 0xf6 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf1, 0xf0, 0xf3, 0xf2, 0xf5, 0xf4, 0xf7, 0xf6 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf1, 0xf0, 0xf3, 0xf2, 0xf5, 0xf4, 0xf7, 0xf6 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf1, 0xf0, 0xf3, 0xf2, 0xf5, 0xf4, 0xf7, 0xf6, 0xf9, 0xf8, 0xfb, 0xfa, 0xfd, 0xfc, 0xff, 0xfe }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf1, 0xf0, 0xf3, 0xf2, 0xf5, 0xf4, 0xf7, 0xf6, 0xf9, 0xf8, 0xfb, 0xfa, 0xfd, 0xfc, 0xff, 0xfe }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf1, 0xf0, 0xf3, 0xf2, 0xf5, 0xf4, 0xf7, 0xf6, 0xf9, 0xf8, 0xfb, 0xfa, 0xfd, 0xfc, 0xff, 0xfe }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VREV32 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf3, 0xf2, 0xf1, 0xf0, 0xf7, 0xf6, 0xf5, 0xf4 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff1, 0xfff0, 0xfff3, 0xfff2 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf3, 0xf2, 0xf1, 0xf0, 0xf7, 0xf6, 0xf5, 0xf4 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff1, 0xfff0, 0xfff3, 0xfff2 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf3, 0xf2, 0xf1, 0xf0, 0xf7, 0xf6, 0xf5, 0xf4 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff1, 0xfff0, 0xfff3, 0xfff2 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf3, 0xf2, 0xf1, 0xf0, 0xf7, 0xf6, 0xf5, 0xf4, 0xfb, 0xfa, 0xf9, 0xf8, 0xff, 0xfe, 0xfd, 0xfc }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff1, 0xfff0, 0xfff3, 0xfff2, 0xfff5, 0xfff4, 0xfff7, 0xfff6 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf3, 0xf2, 0xf1, 0xf0, 0xf7, 0xf6, 0xf5, 0xf4, 0xfb, 0xfa, 0xf9, 0xf8, 0xff, 0xfe, 0xfd, 0xfc }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff1, 0xfff0, 0xfff3, 0xfff2, 0xfff5, 0xfff4, 0xfff7, 0xfff6 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf3, 0xf2, 0xf1, 0xf0, 0xf7, 0xf6, 0xf5, 0xf4, 0xfb, 0xfa, 0xf9, 0xf8, 0xff, 0xfe, 0xfd, 0xfc }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff1, 0xfff0, 0xfff3, 0xfff2, 0xfff5, 0xfff4, 0xfff7, 0xfff6 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VREV64 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf7, 0xf6, 0xf5, 0xf4, 0xf3, 0xf2, 0xf1, 0xf0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff3, 0xfff2, 0xfff1, 0xfff0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff1, 0xfffffff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf7, 0xf6, 0xf5, 0xf4, 0xf3, 0xf2, 0xf1, 0xf0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff3, 0xfff2, 0xfff1, 0xfff0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff1, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf7, 0xf6, 0xf5, 0xf4, 0xf3, 0xf2, 0xf1, 0xf0 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff3, 0xfff2, 0xfff1, 0xfff0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1700000, 0xc1800000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf7, 0xf6, 0xf5, 0xf4, 0xf3, 0xf2, 0xf1, 0xf0, 0xff, 0xfe, 0xfd, 0xfc, 0xfb, 0xfa, 0xf9, 0xf8 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff3, 0xfff2, 0xfff1, 0xfff0, 0xfff7, 0xfff6, 0xfff5, 0xfff4 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff1, 0xfffffff0, 0xfffffff3, 0xfffffff2 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf7, 0xf6, 0xf5, 0xf4, 0xf3, 0xf2, 0xf1, 0xf0, 0xff, 0xfe, 0xfd, 0xfc, 0xfb, 0xfa, 0xf9, 0xf8 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff3, 0xfff2, 0xfff1, 0xfff0, 0xfff7, 0xfff6, 0xfff5, 0xfff4 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff1, 0xfffffff0, 0xfffffff3, 0xfffffff2 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf7, 0xf6, 0xf5, 0xf4, 0xf3, 0xf2, 0xf1, 0xf0, 0xff, 0xfe, 0xfd, 0xfc, 0xfb, 0xfa, 0xf9, 0xf8 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff3, 0xfff2, 0xfff1, 0xfff0, 0xfff7, 0xfff6, 0xfff5, 0xfff4 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1700000, 0xc1800000, 0xc1500000, 0xc1600000 }; + +VSRA_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffffc, 0xfffffffd }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb, 0xc }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff3, 0xfffffff4 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff, 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffffc, 0xfffffffd, 0xfffffffe, 0xffffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf, 0x10, 0x11, 0x12, 0x13, 0x14 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfffc, 0xfffd, 0xfffe, 0xffff, 0x0, 0x1, 0x2, 0x3 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff3, 0xfffffff4, 0xfffffff5, 0xfffffff6 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VTRN/VTRNQ chunk 0 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0x11, 0x11, 0xf2, 0xf3, 0x11, 0x11 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0x22, 0x22 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0x55, 0x55, 0xf2, 0xf3, 0x55, 0x55 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0x66, 0x66 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0x55, 0x55, 0xf2, 0xf3, 0x55, 0x55 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0x66, 0x66 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf1, 0x11, 0x11, 0xf2, 0xf3, 0x11, 0x11, 0xf4, 0xf5, 0x11, 0x11, 0xf6, 0xf7, 0x11, 0x11 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0x22, 0x22, 0xfff2, 0xfff3, 0x22, 0x22 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf1, 0x55, 0x55, 0xf2, 0xf3, 0x55, 0x55, 0xf4, 0xf5, 0x55, 0x55, 0xf6, 0xf7, 0x55, 0x55 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0x66, 0x66, 0xfff2, 0xfff3, 0x66, 0x66 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0x77, 0x77 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf0, 0xf1, 0x55, 0x55, 0xf2, 0xf3, 0x55, 0x55, 0xf4, 0xf5, 0x55, 0x55, 0xf6, 0xf7, 0x55, 0x55 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff1, 0x66, 0x66, 0xfff2, 0xfff3, 0x66, 0x66 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, 0x42073333, 0x42073333 }; + +VTRN/VTRNQ chunk 1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf4, 0xf5, 0x11, 0x11, 0xf6, 0xf7, 0x11, 0x11 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff2, 0xfff3, 0x22, 0x22 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf4, 0xf5, 0x55, 0x55, 0xf6, 0xf7, 0x55, 0x55 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff2, 0xfff3, 0x66, 0x66 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x77, 0x77 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf4, 0xf5, 0x55, 0x55, 0xf6, 0xf7, 0x55, 0x55 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff2, 0xfff3, 0x66, 0x66 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x42066666, 0x42066666 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf8, 0xf9, 0x11, 0x11, 0xfa, 0xfb, 0x11, 0x11, 0xfc, 0xfd, 0x11, 0x11, 0xfe, 0xff, 0x11, 0x11 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff4, 0xfff5, 0x22, 0x22, 0xfff6, 0xfff7, 0x22, 0x22 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff2, 0xfffffff3, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf8, 0xf9, 0x55, 0x55, 0xfa, 0xfb, 0x55, 0x55, 0xfc, 0xfd, 0x55, 0x55, 0xfe, 0xff, 0x55, 0x55 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff4, 0xfff5, 0x66, 0x66, 0xfff6, 0xfff7, 0x66, 0x66 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff2, 0xfffffff3, 0x77, 0x77 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf8, 0xf9, 0x55, 0x55, 0xfa, 0xfb, 0x55, 0x55, 0xfc, 0xfd, 0x55, 0x55, 0xfe, 0xff, 0x55, 0x55 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff4, 0xfff5, 0x66, 0x66, 0xfff6, 0xfff7, 0x66, 0x66 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1600000, 0xc1500000, 0x42073333, 0x42073333 }; + +VUZP/VUZPQ chunk 0 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, 0xc1600000, 0xc1500000 }; + +VUZP/VUZPQ chunk 1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x22, 0x22, 0x22, 0x22 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x66, 0x66, 0x66, 0x66 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x77, 0x77 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x66, 0x66, 0x66, 0x66 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x42066666, 0x42066666 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x22, 0x22, 0x22, 0x22, 0x22, 0x22, 0x22, 0x22 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x66, 0x66, 0x66, 0x66, 0x66, 0x66, 0x66, 0x66 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x77, 0x77, 0x77, 0x77 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55, 0x55 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x66, 0x66, 0x66, 0x66, 0x66, 0x66, 0x66, 0x66 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x42073333, 0x42073333, 0x42073333, 0x42073333 }; + +VZIP/VZIPQ chunk 0 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf4, 0x11, 0x11, 0xf1, 0xf5, 0x11, 0x11 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff2, 0x22, 0x22 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf4, 0x55, 0x55, 0xf1, 0xf5, 0x55, 0x55 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff2, 0x66, 0x66 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf4, 0x55, 0x55, 0xf1, 0xf5, 0x55, 0x55 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff2, 0x66, 0x66 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf8, 0x11, 0x11, 0xf1, 0xf9, 0x11, 0x11, 0xf2, 0xfa, 0x11, 0x11, 0xf3, 0xfb, 0x11, 0x11 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff4, 0x22, 0x22, 0xfff1, 0xfff5, 0x22, 0x22 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff2, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf8, 0x55, 0x55, 0xf1, 0xf9, 0x55, 0x55, 0xf2, 0xfa, 0x55, 0x55, 0xf3, 0xfb, 0x55, 0x55 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff4, 0x66, 0x66, 0xfff1, 0xfff5, 0x66, 0x66 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff2, 0x77, 0x77 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf0, 0xf8, 0x55, 0x55, 0xf1, 0xf9, 0x55, 0x55, 0xf2, 0xfa, 0x55, 0x55, 0xf3, 0xfb, 0x55, 0x55 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff4, 0x66, 0x66, 0xfff1, 0xfff5, 0x66, 0x66 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1600000, 0x42073333, 0x42073333 }; + +VZIP/VZIPQ chunk 1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf2, 0xf6, 0x11, 0x11, 0xf3, 0xf7, 0x11, 0x11 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff1, 0xfff3, 0x22, 0x22 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf2, 0xf6, 0x55, 0x55, 0xf3, 0xf7, 0x55, 0x55 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff1, 0xfff3, 0x66, 0x66 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x77, 0x77 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf2, 0xf6, 0x55, 0x55, 0xf3, 0xf7, 0x55, 0x55 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff1, 0xfff3, 0x66, 0x66 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x42066666, 0x42066666 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf4, 0xfc, 0x11, 0x11, 0xf5, 0xfd, 0x11, 0x11, 0xf6, 0xfe, 0x11, 0x11, 0xf7, 0xff, 0x11, 0x11 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff2, 0xfff6, 0x22, 0x22, 0xfff3, 0xfff7, 0x22, 0x22 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff1, 0xfffffff3, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf4, 0xfc, 0x55, 0x55, 0xf5, 0xfd, 0x55, 0x55, 0xf6, 0xfe, 0x55, 0x55, 0xf7, 0xff, 0x55, 0x55 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff2, 0xfff6, 0x66, 0x66, 0xfff3, 0xfff7, 0x66, 0x66 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff1, 0xfffffff3, 0x77, 0x77 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf4, 0xfc, 0x55, 0x55, 0xf5, 0xfd, 0x55, 0x55, 0xf6, 0xfe, 0x55, 0x55, 0xf7, 0xff, 0x55, 0x55 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff2, 0xfff6, 0x66, 0x66, 0xfff3, 0xfff7, 0x66, 0x66 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1700000, 0xc1500000, 0x42073333, 0x42073333 }; +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xff, 0xf1, 0xff, 0xf2, 0xff, 0xf3, 0xff }; +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xff, 0xff, 0xff, 0xf1, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xff, 0xf1, 0xff, 0xf2, 0xff, 0xf3, 0xff }; +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xff, 0xff, 0xff, 0xf1, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xff, 0xf1, 0xff, 0xf2, 0xff, 0xf3, 0xff }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xf1f0, 0xf3f2, 0xf5f4, 0xf7f6 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xffff, 0xfff1, 0xffff }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xf1f0, 0xf3f2, 0xf5f4, 0xf7f6 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xffff, 0xfff1, 0xffff }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xf1f0, 0xf3f2, 0xf5f4, 0xf7f6 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xf3f2f1f0, 0xf7f6f5f4 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfff1fff0, 0xfff3fff2 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xffffffff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xf3f2f1f0, 0xf7f6f5f4 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfff1fff0, 0xfff3fff2 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xffffffff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xf3f2f1f0, 0xf7f6f5f4 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfff1fff0, 0xfff3fff2 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xf7f6f5f4f3f2f1f0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfff3fff2fff1fff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffff1fffffff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xf7f6f5f4f3f2f1f0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfff3fff2fff1fff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffff1fffffff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xf7f6f5f4f3f2f1f0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfff3fff2fff1fff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xff, 0xf1, 0xff, 0xf2, 0xff, 0xf3, 0xff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xff, 0xff, 0xff, 0xf1, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xff, 0xf1, 0xff, 0xf2, 0xff, 0xf3, 0xff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xff, 0xff, 0xff, 0xf1, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xff, 0xf1, 0xff, 0xf2, 0xff, 0xf3, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xf1f0, 0xf3f2, 0xf5f4, 0xf7f6 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xffff, 0xfff1, 0xffff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xf1f0, 0xf3f2, 0xf5f4, 0xf7f6 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xffff, 0xfff1, 0xffff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xf1f0, 0xf3f2, 0xf5f4, 0xf7f6 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xf3f2f1f0, 0xf7f6f5f4 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfff1fff0, 0xfff3fff2 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xf3f2f1f0, 0xf7f6f5f4 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfff1fff0, 0xfff3fff2 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xf3f2f1f0, 0xf7f6f5f4 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfff1fff0, 0xfff3fff2 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xf7f6f5f4f3f2f1f0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfff3fff2fff1fff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffff1fffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xf7f6f5f4f3f2f1f0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfff3fff2fff1fff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffff1fffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xf7f6f5f4f3f2f1f0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfff3fff2fff1fff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xff, 0xf1, 0xff, 0xf2, 0xff, 0xf3, 0xff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xff, 0xff, 0xff, 0xf1, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xff, 0xf1, 0xff, 0xf2, 0xff, 0xf3, 0xff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xff, 0xff, 0xff, 0xf1, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xff, 0xf1, 0xff, 0xf2, 0xff, 0xf3, 0xff }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xf1f0, 0xf3f2, 0xf5f4, 0xf7f6 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xffff, 0xfff1, 0xffff }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xf1f0, 0xf3f2, 0xf5f4, 0xf7f6 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xffff, 0xfff1, 0xffff }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xf1f0, 0xf3f2, 0xf5f4, 0xf7f6 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xff, 0xf1, 0xff, 0xf2, 0xff, 0xf3, 0xff, 0xf4, 0xff, 0xf5, 0xff, 0xf6, 0xff, 0xf7, 0xff }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xff, 0xff, 0xff, 0xf1, 0xff, 0xff, 0xff, 0xf2, 0xff, 0xff, 0xff, 0xf3, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xf1, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xff, 0xf1, 0xff, 0xf2, 0xff, 0xf3, 0xff, 0xf4, 0xff, 0xf5, 0xff, 0xf6, 0xff, 0xf7, 0xff }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xff, 0xff, 0xff, 0xf1, 0xff, 0xff, 0xff, 0xf2, 0xff, 0xff, 0xff, 0xf3, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xf1, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xff, 0xf1, 0xff, 0xf2, 0xff, 0xf3, 0xff, 0xf4, 0xff, 0xf5, 0xff, 0xf6, 0xff, 0xf7, 0xff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xf1f0, 0xf3f2, 0xf5f4, 0xf7f6, 0xf9f8, 0xfbfa, 0xfdfc, 0xfffe }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xffff, 0xfff1, 0xffff, 0xfff2, 0xffff, 0xfff3, 0xffff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xffff, 0xffff, 0xffff, 0xfff1, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xf1f0, 0xf3f2, 0xf5f4, 0xf7f6, 0xf9f8, 0xfbfa, 0xfdfc, 0xfffe }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xffff, 0xfff1, 0xffff, 0xfff2, 0xffff, 0xfff3, 0xffff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xffff, 0xffff, 0xffff, 0xfff1, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xf1f0, 0xf3f2, 0xf5f4, 0xf7f6, 0xf9f8, 0xfbfa, 0xfdfc, 0xfffe }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xf3f2f1f0, 0xf7f6f5f4, 0xfbfaf9f8, 0xfffefdfc }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfff1fff0, 0xfff3fff2, 0xfff5fff4, 0xfff7fff6 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xffffffff, 0xfffffff1, 0xffffffff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xf3f2f1f0, 0xf7f6f5f4, 0xfbfaf9f8, 0xfffefdfc }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfff1fff0, 0xfff3fff2, 0xfff5fff4, 0xfff7fff6 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xffffffff, 0xfffffff1, 0xffffffff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xf3f2f1f0, 0xf7f6f5f4, 0xfbfaf9f8, 0xfffefdfc }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfff1fff0, 0xfff3fff2, 0xfff5fff4, 0xfff7fff6 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xf7f6f5f4f3f2f1f0, 0xfffefdfcfbfaf9f8 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfff3fff2fff1fff0, 0xfff7fff6fff5fff4 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffff1fffffff0, 0xfffffff3fffffff2 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xf7f6f5f4f3f2f1f0, 0xfffefdfcfbfaf9f8 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfff3fff2fff1fff0, 0xfff7fff6fff5fff4 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffff1fffffff0, 0xfffffff3fffffff2 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xf7f6f5f4f3f2f1f0, 0xfffefdfcfbfaf9f8 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfff3fff2fff1fff0, 0xfff7fff6fff5fff4 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xf1f0, 0xf3f2, 0xf5f4, 0xf7f6, 0xf9f8, 0xfbfa, 0xfdfc, 0xfffe }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xffff, 0xfff1, 0xffff, 0xfff2, 0xffff, 0xfff3, 0xffff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xffff, 0xffff, 0xffff, 0xfff1, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xf1f0, 0xf3f2, 0xf5f4, 0xf7f6, 0xf9f8, 0xfbfa, 0xfdfc, 0xfffe }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xffff, 0xfff1, 0xffff, 0xfff2, 0xffff, 0xfff3, 0xffff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xffff, 0xffff, 0xffff, 0xfff1, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xf1f0, 0xf3f2, 0xf5f4, 0xf7f6, 0xf9f8, 0xfbfa, 0xfdfc, 0xfffe }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xf3f2f1f0, 0xf7f6f5f4, 0xfbfaf9f8, 0xfffefdfc }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfff1fff0, 0xfff3fff2, 0xfff5fff4, 0xfff7fff6 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xffffffff, 0xfffffff1, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xf3f2f1f0, 0xf7f6f5f4, 0xfbfaf9f8, 0xfffefdfc }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfff1fff0, 0xfff3fff2, 0xfff5fff4, 0xfff7fff6 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xffffffff, 0xfffffff1, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xf3f2f1f0, 0xf7f6f5f4, 0xfbfaf9f8, 0xfffefdfc }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfff1fff0, 0xfff3fff2, 0xfff5fff4, 0xfff7fff6 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xf7f6f5f4f3f2f1f0, 0xfffefdfcfbfaf9f8 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfff3fff2fff1fff0, 0xfff7fff6fff5fff4 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffff1fffffff0, 0xfffffff3fffffff2 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xf7f6f5f4f3f2f1f0, 0xfffefdfcfbfaf9f8 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfff3fff2fff1fff0, 0xfff7fff6fff5fff4 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffff1fffffff0, 0xfffffff3fffffff2 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xf7f6f5f4f3f2f1f0, 0xfffefdfcfbfaf9f8 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfff3fff2fff1fff0, 0xfff7fff6fff5fff4 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xff, 0xf1, 0xff, 0xf2, 0xff, 0xf3, 0xff, 0xf4, 0xff, 0xf5, 0xff, 0xf6, 0xff, 0xf7, 0xff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xff, 0xff, 0xff, 0xf1, 0xff, 0xff, 0xff, 0xf2, 0xff, 0xff, 0xff, 0xf3, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xf1, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xff, 0xf1, 0xff, 0xf2, 0xff, 0xf3, 0xff, 0xf4, 0xff, 0xf5, 0xff, 0xf6, 0xff, 0xf7, 0xff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xff, 0xff, 0xff, 0xf1, 0xff, 0xff, 0xff, 0xf2, 0xff, 0xff, 0xff, 0xf3, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xf1, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xff, 0xf1, 0xff, 0xf2, 0xff, 0xf3, 0xff, 0xf4, 0xff, 0xf5, 0xff, 0xf6, 0xff, 0xf7, 0xff }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xf3f2f1f0, 0xf7f6f5f4 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xfff1fff0, 0xfff3fff2 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xfffffff0, 0xffffffff }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xf3f2f1f0, 0xf7f6f5f4 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xfff1fff0, 0xfff3fff2 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xfffffff0, 0xffffffff }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xf3f2f1f0, 0xf7f6f5f4 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xfff1fff0, 0xfff3fff2 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xf3f2f1f0, 0xf7f6f5f4, 0xfbfaf9f8, 0xfffefdfc }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xfff1fff0, 0xfff3fff2, 0xfff5fff4, 0xfff7fff6 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xfffffff0, 0xffffffff, 0xfffffff1, 0xffffffff }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xf3f2f1f0, 0xf7f6f5f4, 0xfbfaf9f8, 0xfffefdfc }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xfff1fff0, 0xfff3fff2, 0xfff5fff4, 0xfff7fff6 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xfffffff0, 0xffffffff, 0xfffffff1, 0xffffffff }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xf3f2f1f0, 0xf7f6f5f4, 0xfbfaf9f8, 0xfffefdfc }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xfff1fff0, 0xfff3fff2, 0xfff5fff4, 0xfff7fff6 }; +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x80, 0xc1, 0x0, 0x0, 0x70, 0xc1 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0xc180, 0x0, 0xc170 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xc1700000c1800000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x80, 0xc1, 0x0, 0x0, 0x70, 0xc1 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0xc180, 0x0, 0xc170 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xc1700000c1800000 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x0, 0x0, 0x80, 0xc1, 0x0, 0x0, 0x70, 0xc1 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x0, 0xc180, 0x0, 0xc170 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x80, 0xc1, 0x0, 0x0, 0x70, 0xc1, 0x0, 0x0, 0x60, 0xc1, 0x0, 0x0, 0x50, 0xc1 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0xc180, 0x0, 0xc170, 0x0, 0xc160, 0x0, 0xc150 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xc1800000, 0xc1700000, 0xc1600000, 0xc1500000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xc1700000c1800000, 0xc1500000c1600000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x80, 0xc1, 0x0, 0x0, 0x70, 0xc1, 0x0, 0x0, 0x60, 0xc1, 0x0, 0x0, 0x50, 0xc1 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0xc180, 0x0, 0xc170, 0x0, 0xc160, 0x0, 0xc150 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xc1800000, 0xc1700000, 0xc1600000, 0xc1500000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xc1700000c1800000, 0xc1500000c1600000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x0, 0x0, 0x80, 0xc1, 0x0, 0x0, 0x70, 0xc1, 0x0, 0x0, 0x60, 0xc1, 0x0, 0x0, 0x50, 0xc1 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x0, 0xc180, 0x0, 0xc170, 0x0, 0xc160, 0x0, 0xc150 }; +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0xcc, 0x80, 0xcb, 0x0, 0xcb, 0x80, 0xca }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xcc00, 0xcb80, 0xcb00, 0xca80 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xcb80cc00, 0xca80cb00 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xca80cb00cb80cc00 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0xcc, 0x80, 0xcb, 0x0, 0xcb, 0x80, 0xca }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xcc00, 0xcb80, 0xcb00, 0xca80 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xcb80cc00, 0xca80cb00 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xca80cb00cb80cc00 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x0, 0xcc, 0x80, 0xcb, 0x0, 0xcb, 0x80, 0xca }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xcc00, 0xcb80, 0xcb00, 0xca80 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xcb80cc00, 0xca80cb00 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0xcc, 0x80, 0xcb, 0x0, 0xcb, 0x80, 0xca, 0x0, 0xca, 0x80, 0xc9, 0x0, 0xc9, 0x80, 0xc8 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xcc00, 0xcb80, 0xcb00, 0xca80, 0xca00, 0xc980, 0xc900, 0xc880 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xcb80cc00, 0xca80cb00, 0xc980ca00, 0xc880c900 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xca80cb00cb80cc00, 0xc880c900c980ca00 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0xcc, 0x80, 0xcb, 0x0, 0xcb, 0x80, 0xca, 0x0, 0xca, 0x80, 0xc9, 0x0, 0xc9, 0x80, 0xc8 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xcc00, 0xcb80, 0xcb00, 0xca80, 0xca00, 0xc980, 0xc900, 0xc880 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xcb80cc00, 0xca80cb00, 0xc980ca00, 0xc880c900 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xca80cb00cb80cc00, 0xc880c900c980ca00 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x0, 0xcc, 0x80, 0xcb, 0x0, 0xcb, 0x80, 0xca, 0x0, 0xca, 0x80, 0xc9, 0x0, 0xc9, 0x80, 0xc8 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xcc00, 0xcb80, 0xcb00, 0xca80, 0xca00, 0xc980, 0xc900, 0xc880 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xcb80cc00, 0xca80cb00, 0xc980ca00, 0xc880c900 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; + +VQRDMULH output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff5, 0xfff6, 0xfff7, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; + +VQRDMULH (check mul cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; + +VQRDMULH (check rounding cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; + +VQRDMULH_LANE output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; + +VQRDMULH_LANE (check mul cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; + +VQRDMULH_LANE (check rounding cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfffc, 0xfffc, 0xfffc, 0xfffd }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffffe, 0xfffffffe }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x6, 0x6, 0x6, 0x5, 0x5, 0x4, 0x4, 0x4 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffffe, 0xfffffffe, 0xfffffffe, 0xfffffffe }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; + +VQRDMULH_N (check mul cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; + +VQRDMULH_N (check rounding cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; + +VQRSHL/VQRSHLQ (with input = 0) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; + +VQRSHL/VQRSHLQ (input 0 and negative shift amount) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; + +VQRSHL/VQRSHLQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xe0, 0xe2, 0xe4, 0xe6, 0xe8, 0xea, 0xec, 0xee }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xff80, 0xff88, 0xff90, 0xff98 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffff000, 0xfffff100 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffff80 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x8000000000000000, 0x8000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; + +VQRSHL/VQRSHLQ (negative shift amount) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xfc, 0xfc, 0xfd, 0xfd, 0xfd, 0xfd, 0xfe, 0xfe }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfffc, 0xfffc, 0xfffd, 0xfffd }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffffe, 0xfffffffe }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x3c, 0x3c, 0x3d, 0x3d, 0x3d, 0x3d, 0x3e, 0x3e }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3ffc, 0x3ffc, 0x3ffd, 0x3ffd }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x1ffffffe, 0x1ffffffe }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x80000, 0x80000, 0x80000, 0x80000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x100000000000, 0x100000000000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; + +VQRSHL/VQRSHLQ (checking cumulative saturation: shift by -1) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x4000, 0x4000, 0x4000, 0x4000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x40000000, 0x40000000 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x4000000000000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x8000000000000000 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x4000, 0x4000, 0x4000, 0x4000, 0x4000, 0x4000, 0x4000, 0x4000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x40000000, 0x40000000, 0x40000000, 0x40000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x4000000000000000, 0x4000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x8000000000000000, 0x8000000000000000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; + +VQRSHL/VQRSHLQ (checking cumulative saturation: shift by -3) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x1000, 0x1000, 0x1000, 0x1000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x10000000, 0x10000000 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x1000000000000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x2000, 0x2000, 0x2000, 0x2000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x20000000, 0x20000000 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x2000000000000000 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x1000, 0x1000, 0x1000, 0x1000, 0x1000, 0x1000, 0x1000, 0x1000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x10000000, 0x10000000, 0x10000000, 0x10000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x1000000000000000, 0x1000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x2000, 0x2000, 0x2000, 0x2000, 0x2000, 0x2000, 0x2000, 0x2000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x20000000, 0x20000000, 0x20000000, 0x20000000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x2000000000000000, 0x2000000000000000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; + +VQRSHL/VQRSHLQ (checking cumulative saturation: large shift amount) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x7fffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x7fffffffffffffff, 0x7fffffffffffffff }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; + +VQRSHL/VQRSHLQ (checking cumulative saturation: large shift amount with negative input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x8000000000000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x8000000000000000, 0x8000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; + +VQRSHL/VQRSHLQ (checking cumulative saturation: large negative shift amount) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; + +VQRSHL/VQRSHLQ (checking cumulative saturation: large shift amount with 0 input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VABA/VABAQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x16, 0x17, 0x18, 0x19 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x20, 0x21 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59, 0x5a }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x907, 0x908, 0x909, 0x90a }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffe7, 0xffffffe8 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x5e, 0x5f, 0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x6a, 0x6b, 0x6c, 0x6d }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xb9c, 0xb9d, 0xb9e, 0xb9f, 0xba0, 0xba1, 0xba2, 0xba3 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x26e0, 0x26e1, 0x26e2, 0x26e3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff, 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff9, 0xfffa, 0xfffb, 0xfffc, 0xfffd, 0xfffe, 0xffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VABAL output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff6, 0xfff7, 0xfff8, 0xfff9, 0xfffa, 0xfffb, 0xfffc, 0xfffd }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x16, 0x17, 0x18, 0x19 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x20, 0x21 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59, 0x5a }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x907, 0x908, 0x909, 0x90a }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffe7, 0xffffffe8 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VABAL test intermediate overflow output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xef, 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffef, 0xfff0, 0xfff1, 0xfff2 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffef, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xee, 0xef, 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffe2, 0xffe3, 0xffe4, 0xffe5 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffe7, 0xffffffe8 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VABD/VABDQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x11, 0x10, 0xf, 0xe, 0xd, 0xc, 0xb, 0xa }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3, 0x2, 0x1, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x18, 0x17 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xef, 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffe3, 0xffe4, 0xffe5, 0xffe6 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffe8, 0xffffffe9 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x41c26666, 0x41ba6666 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x1a, 0x19, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0xf, 0xe, 0xd, 0xc, 0xb }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x4, 0x3, 0x2, 0x1, 0x0, 0x1, 0x2, 0x3 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x30, 0x2f, 0x2e, 0x2d }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xe6, 0xe7, 0xe8, 0xe9, 0xea, 0xeb, 0xec, 0xed, 0xee, 0xef, 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffe4, 0xffe5, 0xffe6, 0xffe7, 0xffe8, 0xffe9, 0xffea, 0xffeb }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffd0, 0xffffffd1, 0xffffffd2, 0xffffffd3 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x42407ae1, 0x423c7ae1, 0x42387ae1, 0x42347ae1 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; + +VABDL output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x11, 0x10, 0xf, 0xe, 0xd, 0xc, 0xb, 0xa }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x3, 0x2, 0x1, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x18, 0x17 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xef, 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffe3, 0xffe4, 0xffe5, 0xffe6 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffe8, 0xffffffe9 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VAND/VANDQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x2, 0x2, 0x0, 0x0, 0x2, 0x2 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x60 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x10, 0x10, 0x10, 0x10, 0x14, 0x14, 0x14, 0x14 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x10, 0x10, 0x12, 0x12 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x20, 0x20 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf0, 0xf2, 0xf2, 0xf4, 0xf4, 0xf6, 0xf6, 0xf0, 0xf0, 0xf2, 0xf2, 0xf4, 0xf4, 0xf6, 0xf6 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffe0, 0xffe0, 0xffe0, 0xffe0, 0xffe4, 0xffe4, 0xffe4, 0xffe4 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffe0, 0xffffffe0, 0xffffffe2, 0xffffffe2 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x10, 0x10 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x4, 0x4, 0x4, 0x4, 0x8, 0x8, 0x8, 0x8, 0xc, 0xc, 0xc, 0xc }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x1, 0x2, 0x3, 0x0, 0x1, 0x2, 0x3 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x30, 0x31, 0x32, 0x33 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VORR/VORRQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf2, 0xf3, 0xf2, 0xf3, 0xf6, 0xf7, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff3, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff4 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf4, 0xf5, 0xf6, 0xf7, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfffe, 0xffff, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff8, 0xfffffff9 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf6, 0xf7, 0xf6, 0xf7, 0xf6, 0xf7, 0xf6, 0xf7, 0xfe, 0xff, 0xfe, 0xff, 0xfe, 0xff, 0xfe, 0xff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfffc, 0xfffd, 0xfffe, 0xffff, 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff2, 0xfffffff3, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff8, 0xfffffffffffffff9 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xfc, 0xfd, 0xfe, 0xff, 0xfc, 0xfd, 0xfe, 0xff, 0xfc, 0xfd, 0xfe, 0xff, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff3, 0xfff3, 0xfff3, 0xfff3, 0xfff7, 0xfff7, 0xfff7, 0xfff7 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff7, 0xfffffff7, 0xfffffff7, 0xfffffff7 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff3, 0xfffffffffffffff3 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VORN/VORNQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xfd, 0xfd, 0xff, 0xff, 0xfd, 0xfd, 0xff, 0xff }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff3, 0xfff3, 0xfff3, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffffc, 0xfffffffd }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffffb }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xfb, 0xfb, 0xfb, 0xfb, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff1, 0xfff1, 0xfff3, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff7, 0xfffffff7 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffffd }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf9, 0xf9, 0xfb, 0xfb, 0xfd, 0xfd, 0xff, 0xff, 0xf9, 0xf9, 0xfb, 0xfb, 0xfd, 0xfd, 0xff, 0xff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff3, 0xfff3, 0xfff3, 0xfff3, 0xfff7, 0xfff7, 0xfff7, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffffd, 0xfffffffd, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff7, 0xfffffffffffffff7 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf3, 0xf3, 0xf3, 0xf3, 0xf7, 0xf7, 0xf7, 0xf7, 0xfb, 0xfb, 0xfb, 0xfb, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfffc, 0xfffd, 0xfffe, 0xffff, 0xfffc, 0xfffd, 0xfffe, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff8, 0xfffffff9, 0xfffffffa, 0xfffffffb }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffffc, 0xfffffffffffffffd }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VEOR/VEORQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf2, 0xf3, 0xf0, 0xf1, 0xf6, 0xf7, 0xf4, 0xf5 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xc, 0xd, 0xe, 0xf }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff3, 0xfffffff2 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffff94 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xe4, 0xe5, 0xe6, 0xe7, 0xe0, 0xe1, 0xe2, 0xe3 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffee, 0xffef, 0xffec, 0xffed }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffd8, 0xffffffd9 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x6, 0x7, 0x4, 0x5, 0x2, 0x3, 0x0, 0x1, 0xe, 0xf, 0xc, 0xd, 0xa, 0xb, 0x8, 0x9 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x1c, 0x1d, 0x1e, 0x1f, 0x18, 0x19, 0x1a, 0x1b }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x12, 0x13, 0x10, 0x11 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffffe8, 0xffffffffffffffe9 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xfc, 0xfd, 0xfe, 0xff, 0xf8, 0xf9, 0xfa, 0xfb, 0xf4, 0xf5, 0xf6, 0xf7, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff3, 0xfff2, 0xfff1, 0xfff0, 0xfff7, 0xfff6, 0xfff5, 0xfff4 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffc7, 0xffffffc6, 0xffffffc5, 0xffffffc4 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff3, 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VBIC/VBICQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf0, 0xf1, 0xf4, 0xf5, 0xf4, 0xf5 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x1, 0x2, 0x3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffff90 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xe0, 0xe1, 0xe2, 0xe3, 0xe0, 0xe1, 0xe2, 0xe3 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffe0, 0xffe1, 0xffe0, 0xffe1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffd0, 0xffffffd1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x1, 0x0, 0x1, 0x0, 0x1, 0x0, 0x1, 0x8, 0x9, 0x8, 0x9, 0x8, 0x9, 0x8, 0x9 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x10, 0x11, 0x12, 0x13, 0x10, 0x11, 0x12, 0x13 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x10, 0x11, 0x10, 0x11 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffffe0, 0xffffffffffffffe1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf0, 0xf1, 0xf2, 0xf3, 0xf0, 0xf1, 0xf2, 0xf3, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff4, 0xfff4, 0xfff4, 0xfff4 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffc0, 0xffffffc0, 0xffffffc0, 0xffffffc0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VCREATE output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xde, 0xbc, 0x9a, 0x78, 0x56, 0x34, 0x12 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xdef0, 0x9abc, 0x5678, 0x1234 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x9abcdef0, 0x12345678 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x123456789abcdef0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xde, 0xbc, 0x9a, 0x78, 0x56, 0x34, 0x12 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xdef0, 0x9abc, 0x5678, 0x1234 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x9abcdef0, 0x12345678 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x123456789abcdef0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xde, 0xbc, 0x9a, 0x78, 0x56, 0x34, 0x12 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xdef0, 0x9abc, 0x5678, 0x1234 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x9abcdef0, 0x12345678 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VLD2_LANE/VLD2Q_LANE chunk 0 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; + +VLD2_LANE/VLD2Q_LANE chunk 1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xf0, 0xf1 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xaaaa, 0xaaaa, 0xfff0, 0xfff1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xaaaa, 0xaaaa, 0xfff0, 0xfff1 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xfff0, 0xfff1, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xfff0, 0xfff1, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xfff0, 0xfff1, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, 0xaaaaaaaa, 0xaaaaaaaa }; + +VLD3_LANE/VLD3Q_LANE chunk 0 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; + +VLD3_LANE/VLD3Q_LANE chunk 1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xaaaa, 0xaaaa, 0xfff0, 0xfff1 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff2, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xf0, 0xf1, 0xf2, 0xaa }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xaaaaaaaa, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xf0, 0xf1, 0xf2, 0xaa }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1600000, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xfff0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xfff0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xc1800000, 0xc1700000 }; + +VLD3_LANE/VLD3Q_LANE chunk 2 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xf0, 0xf1, 0xf2 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff2, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xaaaa, 0xfff0, 0xfff1, 0xfff2 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff1, 0xfffffff2 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xaaaa, 0xfff0, 0xfff1, 0xfff2 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xfff0, 0xfff1, 0xfff2, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff2, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff1, 0xfff2, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff1, 0xfff2, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1600000, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; + +VLD4_LANE/VLD4Q_LANE chunk 0 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; + +VLD4_LANE/VLD4Q_LANE chunk 1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1600000, 0xc1500000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; + +VLD4_LANE/VLD4Q_LANE chunk 2 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, 0xc1600000, 0xc1500000 }; + +VLD4_LANE/VLD4Q_LANE chunk 3 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa, 0xaa }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa, 0xaaaa }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa, 0xaaaaaaaa }; + +VLD2_DUP/VLD2Q_DUP chunk 0 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf0, 0xf1, 0xf0, 0xf1, 0xf0, 0xf1 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff0, 0xfff1 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf0, 0xf1, 0xf0, 0xf1, 0xf0, 0xf1 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff0, 0xfff1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf0, 0xf1, 0xf0, 0xf1, 0xf0, 0xf1 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff0, 0xfff1 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VLD2_DUP/VLD2Q_DUP chunk 1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf0, 0xf1, 0xf0, 0xf1, 0xf0, 0xf1 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff0, 0xfff1 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf0, 0xf1, 0xf0, 0xf1, 0xf0, 0xf1 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff0, 0xfff1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf0, 0xf1, 0xf0, 0xf1, 0xf0, 0xf1 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff0, 0xfff1 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VLD3_DUP/VLD3Q_DUP chunk 0 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf0, 0xf1, 0xf2, 0xf0, 0xf1 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf0, 0xf1, 0xf2, 0xf0, 0xf1 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf0, 0xf1, 0xf2, 0xf0, 0xf1 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VLD3_DUP/VLD3Q_DUP chunk 1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf2, 0xf0, 0xf1, 0xf2, 0xf0, 0xf1, 0xf2, 0xf0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff1, 0xfff2, 0xfff0, 0xfff1 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff2, 0xfffffff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf2, 0xf0, 0xf1, 0xf2, 0xf0, 0xf1, 0xf2, 0xf0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff1, 0xfff2, 0xfff0, 0xfff1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff2, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf2, 0xf0, 0xf1, 0xf2, 0xf0, 0xf1, 0xf2, 0xf0 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff1, 0xfff2, 0xfff0, 0xfff1 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1600000, 0xc1800000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VLD3_DUP/VLD3Q_DUP chunk 2 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf1, 0xf2, 0xf0, 0xf1, 0xf2, 0xf0, 0xf1, 0xf2 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff2, 0xfff0, 0xfff1, 0xfff2 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff1, 0xfffffff2 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf1, 0xf2, 0xf0, 0xf1, 0xf2, 0xf0, 0xf1, 0xf2 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff2, 0xfff0, 0xfff1, 0xfff2 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff1, 0xfffffff2 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf1, 0xf2, 0xf0, 0xf1, 0xf2, 0xf0, 0xf1, 0xf2 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff2, 0xfff0, 0xfff1, 0xfff2 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1700000, 0xc1600000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VLD4_DUP/VLD4Q_DUP chunk 0 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VLD4_DUP/VLD4Q_DUP chunk 1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1600000, 0xc1500000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VLD4_DUP/VLD4Q_DUP chunk 2 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff2 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VLD4_DUP/VLD4Q_DUP chunk 3 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff3 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff3 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf0, 0xf1, 0xf2, 0xf3 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1600000, 0xc1500000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VMLA output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xdf, 0xe0, 0xe1, 0xe2, 0xe3, 0xe4, 0xe5, 0xe6 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x1f8c, 0x1f8d, 0x1f8e, 0x1f8f }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x2bf7, 0x2bf8 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3e07, 0x3e08, 0x3e09, 0x3e0a }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x43ac, 0x43ad }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x43a14e76, 0x43a1ce76 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf, 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x4830, 0x4831, 0x4832, 0x4833, 0x4834, 0x4835, 0x4836, 0x4837 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x470f, 0x4710, 0x4711, 0x4712 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xac, 0xad, 0xae, 0xaf, 0xb0, 0xb1, 0xb2, 0xb3, 0xb4, 0xb5, 0xb6, 0xb7, 0xb8, 0xb9, 0xba, 0xbb }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3e07, 0x3e08, 0x3e09, 0x3e0a, 0x3e0b, 0x3e0c, 0x3e0d, 0x3e0e }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x3620, 0x3621, 0x3622, 0x3623 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x45f0ae15, 0x45f0b615, 0x45f0be15, 0x45f0c615 }; + +VMLS output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xe054, 0xe055, 0xe056, 0xe057 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xffffd3e9, 0xffffd3ea }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xc0, 0xc1, 0xc2, 0xc3, 0xc4, 0xc5, 0xc6, 0xc7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xc1d9, 0xc1da, 0xc1db, 0xc1dc }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffbc34, 0xffffbc35 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc3b14e76, 0xc3b0ce76 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xd1, 0xd2, 0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xdb, 0xdc, 0xdd, 0xde, 0xdf, 0xe0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xb7b0, 0xb7b1, 0xb7b2, 0xb7b3, 0xb7b4, 0xb7b5, 0xb7b6, 0xb7b7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffb8d1, 0xffffb8d2, 0xffffb8d3, 0xffffb8d4 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3a, 0x3b, 0x3c, 0x3d, 0x3e, 0x3f, 0x40, 0x41, 0x42, 0x43 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xc1d9, 0xc1da, 0xc1db, 0xc1dc, 0xc1dd, 0xc1de, 0xc1df, 0xc1e0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffc9c0, 0xffffc9c1, 0xffffc9c2, 0xffffc9c3 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc5f1ae15, 0xc5f1a615, 0xc5f19e15, 0xc5f19615 }; + +VMUL output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0x1, 0x12, 0x23, 0x34, 0x45, 0x56, 0x67 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfde0, 0xfe02, 0xfe24, 0xfe46 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffcd0, 0xfffffd03 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xc0, 0x4, 0x48, 0x8c, 0xd0, 0x14, 0x58, 0x9c }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfab0, 0xfb05, 0xfb5a, 0xfbaf }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffff9a0, 0xfffffa06 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xc0, 0x84, 0x48, 0xc, 0xd0, 0x94, 0x58, 0x1c }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc4053333, 0xc3f9c000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x90, 0x7, 0x7e, 0xf5, 0x6c, 0xe3, 0x5a, 0xd1, 0x48, 0xbf, 0x36, 0xad, 0x24, 0x9b, 0x12, 0x89 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xf780, 0xf808, 0xf890, 0xf918, 0xf9a0, 0xfa28, 0xfab0, 0xfb38 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffff670, 0xfffff709, 0xfffff7a2, 0xfffff83b }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x60, 0xa, 0xb4, 0x5e, 0x8, 0xb2, 0x5c, 0x6, 0xb0, 0x5a, 0x4, 0xae, 0x58, 0x2, 0xac, 0x56 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xf450, 0xf50b, 0xf5c6, 0xf681, 0xf73c, 0xf7f7, 0xf8b2, 0xf96d }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffff340, 0xfffff40c, 0xfffff4d8, 0xfffff5a4 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x60, 0xca, 0x34, 0x9e, 0xc8, 0x62, 0x9c, 0x36, 0x30, 0x9a, 0x64, 0xce, 0x98, 0x32, 0xcc, 0x66 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc4c73333, 0xc4bac000, 0xc4ae4ccd, 0xc4a1d999 }; + +VMUL_LANE output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xffc0, 0xffc4, 0xffc8, 0xffcc }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffde0, 0xfffffe02 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xbbc0, 0xc004, 0xc448, 0xc88c }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffface0, 0xffffb212 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc3b66666, 0xc3ab0000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffc0, 0xffc4, 0xffc8, 0xffcc, 0xffd0, 0xffd4, 0xffd8, 0xffdc }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffde0, 0xfffffe02, 0xfffffe24, 0xfffffe46 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xbbc0, 0xc004, 0xc448, 0xc88c, 0xccd0, 0xd114, 0xd558, 0xd99c }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffface0, 0xffffb212, 0xffffb744, 0xffffbc76 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc3b66666, 0xc3ab0000, 0xc39f9999, 0xc3943333 }; + +VMUL_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfef0, 0xff01, 0xff12, 0xff23 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffde0, 0xfffffe02 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfcd0, 0xfd03, 0xfd36, 0xfd69 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffbc0, 0xfffffc04 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc3b26666, 0xc3a74000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfab0, 0xfb05, 0xfb5a, 0xfbaf, 0xfc04, 0xfc59, 0xfcae, 0xfd03 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffff9a0, 0xfffffa06, 0xfffffa6c, 0xfffffad2 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xf890, 0xf907, 0xf97e, 0xf9f5, 0xfa6c, 0xfae3, 0xfb5a, 0xfbd1 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffff780, 0xfffff808, 0xfffff890, 0xfffff918 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc4b1cccd, 0xc4a6b000, 0xc49b9333, 0xc4907667 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x11000, 0x11000, 0x11000, 0x11000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x22000, 0x22000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33000, 0x33000, 0x33000, 0x33000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x44000, 0x44000 }; + +VMLA_LANE output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3e07, 0x3e08, 0x3e09, 0x3e0a }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x3e07, 0x3e08 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3e07, 0x3e08, 0x3e09, 0x3e0a }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x3e07, 0x3e08 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x4418c687, 0x44190687 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3e07, 0x3e08, 0x3e09, 0x3e0a, 0x3e0b, 0x3e0c, 0x3e0d, 0x3e0e }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x3e07, 0x3e08, 0x3e09, 0x3e0a }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3e07, 0x3e08, 0x3e09, 0x3e0a, 0x3e0b, 0x3e0c, 0x3e0d, 0x3e0e }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x3e07, 0x3e08, 0x3e09, 0x3e0a }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x441a3168, 0x441a7168, 0x441ab168, 0x441af168 }; + +VMLS_LANE output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xc1d9, 0xc1da, 0xc1db, 0xc1dc }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xffffc1d9, 0xffffc1da }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xc1d9, 0xc1da, 0xc1db, 0xc1dc }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffc1d9, 0xffffc1da }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc420c687, 0xc4208687 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xc1d9, 0xc1da, 0xc1db, 0xc1dc, 0xc1dd, 0xc1de, 0xc1df, 0xc1e0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffc1d9, 0xffffc1da, 0xffffc1db, 0xffffc1dc }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xc1d9, 0xc1da, 0xc1db, 0xc1dc, 0xc1dd, 0xc1de, 0xc1df, 0xc1e0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffc1d9, 0xffffc1da, 0xffffc1db, 0xffffc1dc }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc4223168, 0xc421f168, 0xc421b168, 0xc4217168 }; + +VMLA_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x595, 0x596, 0x597, 0x598 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xb3a, 0xb3b }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x10df, 0x10e0, 0x10e1, 0x10e2 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x1684, 0x1685 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x4497deb8, 0x4497feb8 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x1c29, 0x1c2a, 0x1c2b, 0x1c2c, 0x1c2d, 0x1c2e, 0x1c2f, 0x1c30 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x21ce, 0x21cf, 0x21d0, 0x21d1 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x2773, 0x2774, 0x2775, 0x2776, 0x2777, 0x2778, 0x2779, 0x277a }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x2d18, 0x2d19, 0x2d1a, 0x2d1b }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x4568087b, 0x4568187b, 0x4568287b, 0x4568387b }; + +VMLS_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfa4b, 0xfa4c, 0xfa4d, 0xfa4e }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffff4a6, 0xfffff4a7 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xef01, 0xef02, 0xef03, 0xef04 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffe95c, 0xffffe95d }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc49bdeb8, 0xc49bbeb8 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xe3b7, 0xe3b8, 0xe3b9, 0xe3ba, 0xe3bb, 0xe3bc, 0xe3bd, 0xe3be }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffde12, 0xffffde13, 0xffffde14, 0xffffde15 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xd86d, 0xd86e, 0xd86f, 0xd870, 0xd871, 0xd872, 0xd873, 0xd874 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffd2c8, 0xffffd2c9, 0xffffd2ca, 0xffffd2cb }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc56a087b, 0xc569f87b, 0xc569e87b, 0xc569d87b }; + +VSLI_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xffe0, 0xffe1, 0xffe2, 0xffe3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x6, 0x7 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x64fffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x50, 0x51, 0x52, 0x53, 0x50, 0x51, 0x52, 0x53 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x7bf0, 0x7bf1, 0x7bf2, 0x7bf3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x3ffffff0, 0x3ffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x10 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x50, 0x51, 0x52, 0x53, 0x50, 0x51, 0x52, 0x53 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x7bf0, 0x7bf1, 0x7bf2, 0x7bf3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xd0, 0xd1, 0xd2, 0xd3, 0xd4, 0xd5, 0xd6, 0xd7, 0xd8, 0xd9, 0xda, 0xdb, 0xdc, 0xdd, 0xde, 0xdf }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xff60, 0xff61, 0xff62, 0xff63, 0xff64, 0xff65, 0xff66, 0xff67 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfe2ffff0, 0xfe2ffff1, 0xfe2ffff2, 0xfe2ffff3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x18fff0, 0x18fff1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3ff0, 0x3ff1, 0x3ff2, 0x3ff3, 0x3ff4, 0x3ff5, 0x3ff6, 0x3ff7 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x1bfffff0, 0x1bfffff1, 0x1bfffff2, 0x1bfffff3 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x7ffffffffffff0, 0x7ffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3ff0, 0x3ff1, 0x3ff2, 0x3ff3, 0x3ff4, 0x3ff5, 0x3ff6, 0x3ff7 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VSLI_Nmax shift amount output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7ff0, 0x7ff1, 0x7ff2, 0x7ff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x7ffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x7ff0, 0x7ff1, 0x7ff2, 0x7ff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x7ffffff0, 0x7ffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x7ffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x7ff0, 0x7ff1, 0x7ff2, 0x7ff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x7ff0, 0x7ff1, 0x7ff2, 0x7ff3, 0x7ff4, 0x7ff5, 0x7ff6, 0x7ff7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7ffffff0, 0x7ffffff1, 0x7ffffff2, 0x7ffffff3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x7ffffffffffffff0, 0x7ffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7a, 0x7b, 0x7c, 0x7d, 0x7e, 0x7f }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VSRI_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x80000001, 0x80000001 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffff00000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xc5, 0xc5, 0xc5, 0xc5, 0xc5, 0xc5, 0xc5, 0xc5 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffc0, 0xffc0, 0xffc0, 0xffc0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xe000000000000000 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xc5, 0xc5, 0xc5, 0xc5, 0xc5, 0xc5, 0xc5, 0xc5 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xffc0, 0xffc0, 0xffc0, 0xffc0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf7, 0xf7, 0xf7, 0xf7, 0xf7, 0xf7, 0xf7, 0xf7, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd, 0xfffd }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffff000000000000, 0xffff000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffe00, 0xfffffe00, 0xfffffe00, 0xfffffe00 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffff800, 0xfffffffffffff800 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1, 0xe1 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0, 0xfff0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VSRI_N max shift amount output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7, 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VTST/VTSTQ (signed input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0xffff, 0x0, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0xffff, 0x0, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0xffffffff, 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VTST/VTSTQ (unsigned input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0xffff, 0x0, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0xffff, 0x0, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0xffffffff, 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VADDHN output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x32, 0x32, 0x32, 0x32, 0x32, 0x32, 0x32, 0x32 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x32, 0x32, 0x32, 0x32 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x18, 0x18 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x37, 0x37, 0x37, 0x37 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x3, 0x3 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRADDHN output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x19, 0x19 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x4, 0x4, 0x4, 0x4, 0x4, 0x4, 0x4, 0x4 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x38, 0x38, 0x38, 0x38 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x4, 0x4 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VADDL output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffe3, 0xffe4, 0xffe5, 0xffe6, 0xffe7, 0xffe8, 0xffe9, 0xffea }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffe2, 0xffffffe3, 0xffffffe4, 0xffffffe5 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffffe0, 0xffffffffffffffe1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x1e3, 0x1e4, 0x1e5, 0x1e6, 0x1e7, 0x1e8, 0x1e9, 0x1ea }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x1ffe1, 0x1ffe2, 0x1ffe3, 0x1ffe4 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x1ffffffe0, 0x1ffffffe1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VADDW output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffe3, 0xffe4, 0xffe5, 0xffe6, 0xffe7, 0xffe8, 0xffe9, 0xffea }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffe2, 0xffffffe3, 0xffffffe4, 0xffffffe5 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffffe0, 0xffffffffffffffe1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xe3, 0xe4, 0xe5, 0xe6, 0xe7, 0xe8, 0xe9, 0xea }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffe1, 0xffe2, 0xffe3, 0xffe4 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffe0, 0xffffffe1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VHADD/VHADDQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf1, 0xf2, 0xf2, 0xf3, 0xf3, 0xf4, 0xf4, 0xf5 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff1, 0xfff1, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf1, 0xf2, 0xf2, 0xf3, 0xf3, 0xf4, 0xf4, 0xf5 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff1, 0xfff2 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf2, 0xf2, 0xf3, 0xf3, 0xf4, 0xf4, 0xf5, 0xf5, 0xf6, 0xf6, 0xf7, 0xf7, 0xf8, 0xf8, 0xf9, 0xf9 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff1, 0xfff2, 0xfff2, 0xfff3, 0xfff3, 0xfff4, 0xfff4, 0xfff5 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff1, 0xfffffff2 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf4, 0xf5, 0xf5, 0xf6, 0xf6, 0xf7, 0xf7, 0xf8, 0xf8, 0xf9, 0xf9, 0xfa, 0xfa, 0xfb, 0xfb, 0xfc }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff1, 0xfff1, 0xfff2, 0xfff2, 0xfff3, 0xfff3, 0xfff4, 0xfff4 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff1, 0xfffffff2 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRHADD/VRHADDQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf2, 0xf2, 0xf3, 0xf3, 0xf4, 0xf4, 0xf5, 0xf5 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff1, 0xfff2, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf2, 0xf2, 0xf3, 0xf3, 0xf4, 0xf4, 0xf5, 0xf5 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff1, 0xfff1, 0xfff2, 0xfff2 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf2, 0xf3, 0xf3, 0xf4, 0xf4, 0xf5, 0xf5, 0xf6, 0xf6, 0xf7, 0xf7, 0xf8, 0xf8, 0xf9, 0xf9, 0xfa }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff2, 0xfff2, 0xfff3, 0xfff3, 0xfff4, 0xfff4, 0xfff5, 0xfff5 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff1, 0xfffffff1, 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf5, 0xf5, 0xf6, 0xf6, 0xf7, 0xf7, 0xf8, 0xf8, 0xf9, 0xf9, 0xfa, 0xfa, 0xfb, 0xfb, 0xfc, 0xfc }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff1, 0xfff2, 0xfff2, 0xfff3, 0xfff3, 0xfff4, 0xfff4, 0xfff5 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff1, 0xfffffff1, 0xfffffff2, 0xfffffff2 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VHSUB/VHSUBQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xfe, 0xff, 0xff, 0x0, 0x0, 0x1, 0x1, 0x2 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xffff, 0xffff, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xfe, 0xff, 0xff, 0x0, 0x0, 0x1, 0x1, 0x2 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0x0, 0x0, 0x1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xfe, 0xfe, 0xff, 0xff, 0x0, 0x0, 0x1, 0x1, 0x2, 0x2, 0x3, 0x3, 0x4, 0x4, 0x5, 0x5 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfffe, 0xffff, 0xffff, 0x0, 0x0, 0x1, 0x1, 0x2 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffff, 0x0, 0x0, 0x1 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xfb, 0xfc, 0xfc, 0xfd, 0xfd, 0xfe, 0xfe, 0xff, 0xff, 0x0, 0x0, 0x1, 0x1, 0x2, 0x2, 0x3 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0x0, 0x0, 0x1, 0x1, 0x2, 0x2 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0x0, 0x0, 0x1 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VSUBL output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfffd, 0xfffe, 0xffff, 0x0, 0x1, 0x2, 0x3, 0x4 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffffe, 0xffffffff, 0x0, 0x1 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfffd, 0xfffe, 0xffff, 0x0, 0x1, 0x2, 0x3, 0x4 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0x0, 0x1, 0x2 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VSUBW output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfffd, 0xfffe, 0xffff, 0x0, 0x1, 0x2, 0x3, 0x4 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffffe, 0xffffffff, 0x0, 0x1 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfefd, 0xfefe, 0xfeff, 0xff00, 0xff01, 0xff02, 0xff03, 0xff04 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffeffff, 0xffff0000, 0xffff0001, 0xffff0002 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffff00000000, 0xffffffff00000001 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VSUBHN output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x31, 0x31, 0x31, 0x31, 0x31, 0x31, 0x31, 0x31 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x31, 0x31, 0x31, 0x31 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x17, 0x17 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x36, 0x36, 0x36, 0x36 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x2, 0x2 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSUBHN output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x31, 0x31, 0x31, 0x31, 0x31, 0x31, 0x31, 0x31 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x31, 0x31, 0x31, 0x31 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x17, 0x17 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x36, 0x36, 0x36, 0x36 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x2, 0x2 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VMVN/VMVNQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9, 0x8 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xf, 0xe, 0xd, 0xc }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xf, 0xe }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9, 0x8 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xf, 0xe, 0xd, 0xc }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xf, 0xe }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9, 0x8 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9, 0x8 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xf, 0xe, 0xd, 0xc }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9, 0x8 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xf, 0xe, 0xd, 0xc }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0xf, 0xe, 0xd, 0xc, 0xb, 0xa, 0x9, 0x8, 0x7, 0x6, 0x5, 0x4, 0x3, 0x2, 0x1, 0x0 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 0; + +VQMOVN output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x12, 0x12, 0x12, 0x12, 0x12, 0x12, 0x12, 0x12 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x1278, 0x1278, 0x1278, 0x1278 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x12345678, 0x12345678 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x82, 0x82, 0x82, 0x82, 0x82, 0x82, 0x82, 0x82 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x8765, 0x8765, 0x8765, 0x8765 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x87654321, 0x87654321 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 1; + +VQMOVN output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 0; + +VQMOVUN output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x34, 0x34, 0x34, 0x34, 0x34, 0x34, 0x34, 0x34 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x5678, 0x5678, 0x5678, 0x5678 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x12345678, 0x12345678 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,uint,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,2) = 1; + +VQMOVUN (negative input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHR_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf9, 0xf9, 0xfa, 0xfa, 0xfb, 0xfb, 0xfc }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffffc, 0xfffffffc }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x3c, 0x3c, 0x3d, 0x3d, 0x3d, 0x3d, 0x3e, 0x3e }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x1ffe, 0x1ffe, 0x1ffe, 0x1ffe }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x8000000, 0x8000000 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x80000000 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf8, 0xf9, 0xf9, 0xfa, 0xfa, 0xfb, 0xfb, 0xfc, 0xfc, 0xfd, 0xfd, 0xfe, 0xfe, 0xff, 0xff, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffffc, 0xfffffffc, 0xfffffffd, 0xfffffffd }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x3c, 0x3c, 0x3d, 0x3d, 0x3d, 0x3d, 0x3e, 0x3e, 0x3e, 0x3e, 0x3f, 0x3f, 0x3f, 0x3f, 0x40, 0x40 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x1ffe, 0x1ffe, 0x1ffe, 0x1ffe, 0x1fff, 0x1fff, 0x1fff, 0x1fff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x8000000, 0x8000000, 0x8000000, 0x8000000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHR_N (overflow test: max shift amount, positive input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x1, 0x1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHR_N (overflow test: shift by 1, with negative input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x4000, 0x4000, 0x4000, 0x4000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x40000000, 0x40000000 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x4000000000000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x8000000000000000 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x4000, 0x4000, 0x4000, 0x4000, 0x4000, 0x4000, 0x4000, 0x4000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x40000000, 0x40000000, 0x40000000, 0x40000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x4000000000000000, 0x4000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x8000000000000000, 0x8000000000000000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHR_N (overflow test: shift by 3, positive input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x1000, 0x1000, 0x1000, 0x1000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x10000000, 0x10000000 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x1000000000000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x2000, 0x2000, 0x2000, 0x2000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x20000000, 0x20000000 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x2000000000000000 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x1000, 0x1000, 0x1000, 0x1000, 0x1000, 0x1000, 0x1000, 0x1000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x10000000, 0x10000000, 0x10000000, 0x10000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x1000000000000000, 0x1000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x2000, 0x2000, 0x2000, 0x2000, 0x2000, 0x2000, 0x2000, 0x2000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x20000000, 0x20000000, 0x20000000, 0x20000000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x2000000000000000, 0x2000000000000000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHR_N (overflow test: shift by 1, with negative input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xc000, 0xc000, 0xc000, 0xc000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xc0000000, 0xc0000000 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xc000000000000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x8000000000000000 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xc000, 0xc000, 0xc000, 0xc000, 0xc000, 0xc000, 0xc000, 0xc000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xc0000000, 0xc0000000, 0xc0000000, 0xc0000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xc000000000000000, 0xc000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x8000000000000000, 0x8000000000000000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSHR_N (overflow test: shift by 3, with negative input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xf000, 0xf000, 0xf000, 0xf000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xf0000000, 0xf0000000 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xf000000000000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x2000, 0x2000, 0x2000, 0x2000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x20000000, 0x20000000 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x2000000000000000 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xf000, 0xf000, 0xf000, 0xf000, 0xf000, 0xf000, 0xf000, 0xf000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xf0000000, 0xf0000000, 0xf0000000, 0xf0000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xf000000000000000, 0xf000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x2000, 0x2000, 0x2000, 0x2000, 0x2000, 0x2000, 0x2000, 0x2000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x20000000, 0x20000000, 0x20000000, 0x20000000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x2000000000000000, 0x2000000000000000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSRA_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffffd, 0xfffffffe }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb, 0xc }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfffd, 0xfffe, 0xffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff4, 0xfffffff5 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffff0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff, 0x0, 0x1, 0x2, 0x3, 0x4, 0x5, 0x6, 0x7, 0x8 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0xfff4, 0xfff5, 0xfff6, 0xfff7 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffffd, 0xfffffffe, 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x5, 0x6, 0x7, 0x8, 0x9, 0xa, 0xb, 0xc, 0xd, 0xe, 0xf, 0x10, 0x11, 0x12, 0x13, 0x14 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfffd, 0xfffe, 0xffff, 0x0, 0x1, 0x2, 0x3, 0x4 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff4, 0xfffffff5, 0xfffffff6, 0xfffffff7 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffff0, 0xfffffffffffffff1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSRA_N (checking overflow: shift by 1, positive input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x4000, 0x4000, 0x4000, 0x4000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x40000000, 0x40000000 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x4000000000000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x8000000000000000 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40, 0x40 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x4000, 0x4000, 0x4000, 0x4000, 0x4000, 0x4000, 0x4000, 0x4000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x40000000, 0x40000000, 0x40000000, 0x40000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x4000000000000000, 0x4000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x8000000000000000, 0x8000000000000000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSRA_N (checking overflow: shift by 3, positive input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x1000, 0x1000, 0x1000, 0x1000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x10000000, 0x10000000 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x1000000000000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x2000, 0x2000, 0x2000, 0x2000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x20000000, 0x20000000 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x2000000000000000 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x1000, 0x1000, 0x1000, 0x1000, 0x1000, 0x1000, 0x1000, 0x1000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x10000000, 0x10000000, 0x10000000, 0x10000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x1000000000000000, 0x1000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x2000, 0x2000, 0x2000, 0x2000, 0x2000, 0x2000, 0x2000, 0x2000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x20000000, 0x20000000, 0x20000000, 0x20000000 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x2000000000000000, 0x2000000000000000 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSRA_N (checking overflow: shift by max, positive input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x1, 0x1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSRA_N (checking overflow: shift by 1, negative input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xc000, 0xc000, 0xc000, 0xc000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xc0000000, 0xc0000000 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xc000000000000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0, 0xc0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xc000, 0xc000, 0xc000, 0xc000, 0xc000, 0xc000, 0xc000, 0xc000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xc0000000, 0xc0000000, 0xc0000000, 0xc0000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xc000000000000000, 0xc000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x1, 0x1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSRA_N (checking overflow: shift by max, negative input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xf000, 0xf000, 0xf000, 0xf000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xf0000000, 0xf0000000 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xf000000000000000 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0, 0xf0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xf000, 0xf000, 0xf000, 0xf000, 0xf000, 0xf000, 0xf000, 0xf000 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xf0000000, 0xf0000000, 0xf0000000, 0xf0000000 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xf000000000000000, 0xf000000000000000 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x1, 0x1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VRSRA_N (checking overflow: shift by max, negative input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x1, 0x1 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VSHLL_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffe0, 0xffe2, 0xffe4, 0xffe6, 0xffe8, 0xffea, 0xffec, 0xffee }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffe0, 0xffffffe2, 0xffffffe4, 0xffffffe6 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffff80, 0xffffffffffffff88 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3c0, 0x3c4, 0x3c8, 0x3cc, 0x3d0, 0x3d4, 0x3d8, 0x3dc }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfff00, 0xfff10, 0xfff20, 0xfff30 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x7ffffff80, 0x7ffffff88 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VPADDL/VPADDLQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xffe1, 0xffe5, 0xffe9, 0xffed }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xffffffe1, 0xffffffe5 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffffe1 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x1e1, 0x1e5, 0x1e9, 0x1ed }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x1ffe1, 0x1ffe5 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x1ffffffe1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffe1, 0xffe5, 0xffe9, 0xffed, 0xfff1, 0xfff5, 0xfff9, 0xfffd }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffe1, 0xffffffe5, 0xffffffe9, 0xffffffed }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffffe1, 0xffffffffffffffe5 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x1e1, 0x1e5, 0x1e9, 0x1ed, 0x1f1, 0x1f5, 0x1f9, 0x1fd }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x1ffe1, 0x1ffe5, 0x1ffe9, 0x1ffed }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x1ffffffe1, 0x1ffffffe5 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VPADD output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xe1, 0xe5, 0xe9, 0xed, 0xe1, 0xe5, 0xe9, 0xed }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xffe1, 0xffe5, 0xffe1, 0xffe5 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xffffffe1, 0xffffffe1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xe1, 0xe5, 0xe9, 0xed, 0xe1, 0xe5, 0xe9, 0xed }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffe1, 0xffe5, 0xffe1, 0xffe5 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffe1, 0xffffffe1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1f80000, 0xc1f80000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VPADAL/VPADALQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xffd1, 0xffd6, 0xffdb, 0xffe0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xffffffd1, 0xffffffd6 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0xffffffffffffffd1 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x1d1, 0x1d6, 0x1db, 0x1e0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x1ffd1, 0x1ffd6 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x1ffffffd1 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xffd1, 0xffd6, 0xffdb, 0xffe0, 0xffe5, 0xffea, 0xffef, 0xfff4 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xffffffd1, 0xffffffd6, 0xffffffdb, 0xffffffe0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0xffffffffffffffd1, 0xffffffffffffffd6 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x1d1, 0x1d6, 0x1db, 0x1e0, 0x1e5, 0x1ea, 0x1ef, 0x1f4 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x1ffd1, 0x1ffd6, 0x1ffdb, 0x1ffe0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x1ffffffd1, 0x1ffffffd6 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; + +VQSHLU_N/VQSHLUQ_N (negative input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x0 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; + +VQSHLU_N/VQSHLUQ_N (check cumulative saturation: shift by 1) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xfe, 0xfe, 0xfe, 0xfe, 0xfe, 0xfe, 0xfe, 0xfe }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfffe, 0xfffe, 0xfffe, 0xfffe }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffffe, 0xfffffffe }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xfffffffffffffffe }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xfe, 0xfe, 0xfe, 0xfe, 0xfe, 0xfe, 0xfe, 0xfe, 0xfe, 0xfe, 0xfe, 0xfe, 0xfe, 0xfe, 0xfe, 0xfe }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfffe, 0xfffe, 0xfffe, 0xfffe, 0xfffe, 0xfffe, 0xfffe, 0xfffe }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffffe, 0xfffffffe, 0xfffffffe, 0xfffffffe }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xfffffffffffffffe, 0xfffffffffffffffe }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 1; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 1; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; + +VQSHLU_N/VQSHLUQ_N (check cumulative saturation: shift by 2) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0xffffffffffffffff, 0xffffffffffffffff }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,8,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,2) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,1) = 0; +int VECT_VAR(expected_cumulative_sat,int,8,16) = 0; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; + +VQSHLU_N/VQSHLUQ_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x8, 0x8, 0x8, 0x8 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x18, 0x18 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x40 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0xa0, 0xa0, 0xa0, 0xa0, 0xa0, 0xa0, 0xa0, 0xa0, 0xa0, 0xa0, 0xa0, 0xa0, 0xa0, 0xa0, 0xa0, 0xa0 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x180, 0x180, 0x180, 0x180, 0x180, 0x180, 0x180, 0x180 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x380, 0x380, 0x380, 0x380 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x800, 0x800 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VCLZ/VCLZQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3, 0x3, 0x3, 0x3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x11, 0x11 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x5, 0x5 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x3, 0x3, 0x3, 0x3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3, 0x3 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xd, 0xd, 0xd, 0xd, 0xd, 0xd, 0xd, 0xd }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x1f, 0x1f, 0x1f, 0x1f }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VCLZ/VCLZQ (input=0) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x10, 0x10, 0x10, 0x10 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x20, 0x20 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x10, 0x10, 0x10, 0x10 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x20, 0x20 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x20, 0x20, 0x20, 0x20 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10, 0x10 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x20, 0x20, 0x20, 0x20 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VCLS/VCLSQ (positive input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x2, 0x2, 0x2, 0x2 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x19, 0x19 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x14, 0x14, 0x14, 0x14 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VCLS/VCLSQ (negative input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7, 0x7 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x1, 0x1, 0x1, 0x1 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x1, 0x1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2, 0x2 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VCNT/VCNTQ output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8, 0x8 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x4, 0x4, 0x4, 0x4, 0x4, 0x4, 0x4, 0x4 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x4, 0x4, 0x4, 0x4, 0x4, 0x4, 0x4, 0x4 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6, 0x6 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; + +VQSHRN_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf8, 0xf9, 0xf9, 0xfa, 0xfa, 0xfb, 0xfb }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff8, 0xfff8, 0xfff9, 0xfff9 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffffc, 0xfffffffc }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 1; + +VQSHRN_N (check saturation: shift by 3) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; +int VECT_VAR(expected_cumulative_sat,uint,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,uint,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,uint,64,2) = 0; + +VQSHRN_N (check saturation: shift by max) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f, 0x7f }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x7fff, 0x7fff, 0x7fff, 0x7fff }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VPMAX output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf1, 0xf3, 0xf5, 0xf7, 0xf1, 0xf3, 0xf5, 0xf7 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff1, 0xfff3, 0xfff1, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf1, 0xf3, 0xf5, 0xf7, 0xf1, 0xf3, 0xf5, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff1, 0xfff3, 0xfff1, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff1, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1700000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VPMIN output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf2, 0xf4, 0xf6, 0xf0, 0xf2, 0xf4, 0xf6 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff2, 0xfff0, 0xfff2 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf2, 0xf4, 0xf6, 0xf0, 0xf2, 0xf4, 0xf6 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff2, 0xfff0, 0xfff2 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1800000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; + +VQSHRUN_N (negative input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; + +VQSHRUN_N (check cumulative saturation) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; + +VQSHRUN_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x48, 0x48, 0x48, 0x48, 0x48, 0x48, 0x48, 0x48 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xdeadbe, 0xdeadbe }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; + +VQRSHRUN_N (negative input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; + +VQRSHRUN_N (check cumulative saturation: shift by 1) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xffff, 0xffff, 0xffff, 0xffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 0; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; + +VQRSHRUN_N (check cumulative saturation: shift by max, positive input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x8000, 0x8000, 0x8000, 0x8000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 1; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 1; + +VQRSHRUN_N (check cumulative saturation: shift by max, negative input) output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +int VECT_VAR(expected_cumulative_sat,int,16,8) = 0; +int VECT_VAR(expected_cumulative_sat,int,32,4) = 1; +int VECT_VAR(expected_cumulative_sat,int,64,2) = 0; + +VQRSHRUN_N output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x49, 0x49, 0x49, 0x49, 0x49, 0x49, 0x49, 0x49 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xdeadbf, 0xdeadbf }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VST2_LANE/VST2Q_LANE chunk 0 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff1, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, 0x0, 0x0 }; + +VST2_LANE/VST2Q_LANE chunk 1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; + +VST3_LANE/VST3Q_LANE chunk 0 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, 0xc1600000, 0x0 }; + +VST3_LANE/VST3Q_LANE chunk 1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff2, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff2, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1600000, 0x0 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; + +VST3_LANE/VST3Q_LANE chunk 2 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; + +VST4_LANE/VST4Q_LANE chunk 0 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff0, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf0, 0xf1, 0xf2, 0xf3, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfffffff0, 0xfffffff1, 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0xfff0, 0xfff1, 0xfff2, 0xfff3, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, 0xc1600000, 0xc1500000 }; + +VST4_LANE/VST4Q_LANE chunk 1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xfffffff2, 0xfffffff3 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1600000, 0xc1500000 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; + +VST4_LANE/VST4Q_LANE chunk 2 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; + +VST4_LANE/VST4Q_LANE chunk 3 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; + +VTBL1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x0, 0xf2, 0xf2, 0xf2, 0x0, 0x0, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0x0, 0xf3, 0xf3, 0xf3, 0x0, 0x0, 0xf3, 0xf3 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0x0, 0xf3, 0xf3, 0xf3, 0x0, 0x0, 0xf3, 0xf3 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VTBL2 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf6, 0xf3, 0xf3, 0xf3, 0x0, 0x0, 0xf3, 0xf3 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf6, 0xf5, 0xf5, 0xf5, 0x0, 0x0, 0xf5, 0xf5 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf6, 0xf5, 0xf5, 0xf5, 0x0, 0x0, 0xf5, 0xf5 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VTBL3 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf4, 0xf4, 0xf4, 0xff, 0x0, 0xf4, 0xf4 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf8, 0xf7, 0xf7, 0xf7, 0xff, 0x0, 0xf7, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf8, 0xf7, 0xf7, 0xf7, 0xff, 0x0, 0xf7, 0xf7 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VTBL4 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xfa, 0xf5, 0xf5, 0xf5, 0x3, 0x0, 0xf5, 0xf5 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xfa, 0xf9, 0xf9, 0xf9, 0x3, 0x0, 0xf9, 0xf9 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xfa, 0xf9, 0xf9, 0xf9, 0x3, 0x0, 0xf9, 0xf9 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VTBX1 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0x33, 0xf2, 0xf2, 0xf2, 0x33, 0x33, 0xf2, 0xf2 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xcc, 0xf3, 0xf3, 0xf3, 0xcc, 0xcc, 0xf3, 0xf3 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xcc, 0xf3, 0xf3, 0xf3, 0xcc, 0xcc, 0xf3, 0xf3 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VTBX2 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf6, 0xf3, 0xf3, 0xf3, 0x33, 0x33, 0xf3, 0xf3 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf6, 0xf5, 0xf5, 0xf5, 0xcc, 0xcc, 0xf5, 0xf5 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf6, 0xf5, 0xf5, 0xf5, 0xcc, 0xcc, 0xf5, 0xf5 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VTBX3 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xf8, 0xf4, 0xf4, 0xf4, 0xff, 0x33, 0xf4, 0xf4 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xf8, 0xf7, 0xf7, 0xf7, 0xff, 0xcc, 0xf7, 0xf7 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xf8, 0xf7, 0xf7, 0xf7, 0xff, 0xcc, 0xf7, 0xf7 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; + +VTBX4 output: +VECT_VAR_DECL(expected,int,8,8) [] = { 0xfa, 0xf5, 0xf5, 0xf5, 0x3, 0x33, 0xf5, 0xf5 }; +VECT_VAR_DECL(expected,int,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,8) [] = { 0xfa, 0xf9, 0xf9, 0xf9, 0x3, 0xcc, 0xf9, 0xf9 }; +VECT_VAR_DECL(expected,uint,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,1) [] = { 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,8) [] = { 0xfa, 0xf9, 0xf9, 0xf9, 0x3, 0xcc, 0xf9, 0xf9 }; +VECT_VAR_DECL(expected,poly,16,4) [] = { 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,int,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,int,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,uint,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,uint,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,64,2) [] = { 0x3333333333333333, 0x3333333333333333 }; +VECT_VAR_DECL(expected,poly,8,16) [] = { 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33, 0x33 }; +VECT_VAR_DECL(expected,poly,16,8) [] = { 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333, 0x3333 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x33333333, 0x33333333, 0x33333333, 0x33333333 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xbf000000, 0xbf000000, 0xbf000000, 0xbf000000 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x3f068000, 0x3f068000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x3c030000, 0x3c030000, 0x3c030000, 0x3c030000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xee800000, 0xee800000, 0xee800000, 0xee800000 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xbdcc8000, 0xbdcc8000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xbc030000, 0xbc030000, 0xbc030000, 0xbc030000 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x7fc00000, 0x7fc00000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x7f800000, 0x7f800000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xff800000, 0xff800000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x80000000, 0x80000000, 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x9c800000, 0x9c800000, 0x9c800000, 0x9c800000 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x3e498000, 0x3e498000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x3e700000, 0x3e700000, 0x3e700000, 0x3e700000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x80000000, 0x80000000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xae800000, 0xae800000, 0xae800000, 0xae800000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xb4800000, 0xb4800000 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xed000000, 0xed000000, 0xed000000, 0xed000000 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x7fc00000, 0x7fc00000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x7f800000, 0x7f800000, 0x7f800000, 0x7f800000 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x7fc00000, 0x7fc00000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xff800000, 0xff800000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x7fc00000, 0x7fc00000, 0x7fc00000, 0x7fc00000 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0xffffffff }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x0 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x0, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc1800000, 0xc1700000 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x4f800000, 0x4f800000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x41720000, 0xc1720000, 0x0, 0x80000000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1800000, 0xc1700000, 0xc1600000, 0xc1500000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x4f800000, 0x4f800000, 0x4f800000, 0x4f800000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xfffffff1, 0x5 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x5 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0xf, 0xfffffff1 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0xf, 0x0 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc0800000, 0xc0700000 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x4c000000, 0x4c000000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xb2800000, 0xb2700000, 0xb2600000, 0xb2500000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x49800000, 0x49800000, 0x49800000, 0x49800000 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xff0b3333, 0x54cccd }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0x0, 0x15 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x0, 0x0, 0x1e3d7, 0xfffe1c29 }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x0, 0x0, 0x1e, 0x0 }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xa, 0xa }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xa, 0xa }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7d, 0x7d, 0x7d, 0x7d }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0x7d, 0x7d, 0x7d, 0x7d }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0xa66666, 0xa66666 }; +VECT_VAR_DECL(expected,uint,32,2) [] = { 0xa66666, 0xa66666 }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0xfbccc, 0xfbccc, 0xfbccc, 0xfbccc }; +VECT_VAR_DECL(expected,uint,32,4) [] = { 0xfbccc, 0xfbccc, 0xfbccc, 0xfbccc }; +VECT_VAR_DECL(expected,int,32,2) [] = { 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,int,32,4) [] = { 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x38640000, 0x7f800000, 0x7fc00000, 0xff800000 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc2e19eb7, 0xc2e19eb7 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc1db851f, 0xc1db851f, 0xc1db851f, 0xc1db851f }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x7fc00000, 0x7fc00000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x7fc00000, 0x7fc00000, 0x7fc00000, 0x7fc00000 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xff800000, 0xff800000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x40000000, 0x40000000, 0x40000000, 0x40000000 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x40000000, 0x40000000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x40000000, 0x40000000, 0x40000000, 0x40000000 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xc2796b84, 0xc2796b84 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0xc0e4a3d8, 0xc0e4a3d8, 0xc0e4a3d8, 0xc0e4a3d8 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x7fc00000, 0x7fc00000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x7fc00000, 0x7fc00000, 0x7fc00000, 0x7fc00000 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0xff800000, 0xff800000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x3fc00000, 0x3fc00000, 0x3fc00000, 0x3fc00000 }; +VECT_VAR_DECL(expected,hfloat,32,2) [] = { 0x3fc00000, 0x3fc00000 }; +VECT_VAR_DECL(expected,hfloat,32,4) [] = { 0x3fc00000, 0x3fc00000, 0x3fc00000, 0x3fc00000 }; |