;; ARM NEON coprocessor Machine Description ;; Copyright (C) 2006, 2007, 2008, 2009, 2010 Free Software Foundation, Inc. ;; Written by CodeSourcery. ;; ;; This file is part of GCC. ;; ;; GCC is free software; you can redistribute it and/or modify it ;; under the terms of the GNU General Public License as published by ;; the Free Software Foundation; either version 3, or (at your option) ;; any later version. ;; ;; GCC is distributed in the hope that it will be useful, but ;; WITHOUT ANY WARRANTY; without even the implied warranty of ;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU ;; General Public License for more details. ;; ;; You should have received a copy of the GNU General Public License ;; along with GCC; see the file COPYING3. If not see ;; . ;; Constants for unspecs. (define_constants [(UNSPEC_ASHIFT_SIGNED 65) (UNSPEC_ASHIFT_UNSIGNED 66) (UNSPEC_VABD 69) (UNSPEC_VABDL 70) (UNSPEC_VADD 72) (UNSPEC_VADDHN 73) (UNSPEC_VADDL 74) (UNSPEC_VADDW 75) (UNSPEC_VBSL 78) (UNSPEC_VCAGE 79) (UNSPEC_VCAGT 80) (UNSPEC_VCEQ 81) (UNSPEC_VCGE 82) (UNSPEC_VCGT 83) (UNSPEC_VCLS 84) (UNSPEC_VCVT 88) (UNSPEC_VCVT_N 89) (UNSPEC_VEXT 93) (UNSPEC_VHADD 97) (UNSPEC_VHSUB 98) (UNSPEC_VLD1 99) (UNSPEC_VLD1_DUP 100) (UNSPEC_VLD1_LANE 101) (UNSPEC_VLD2 102) (UNSPEC_VLD2_DUP 103) (UNSPEC_VLD2_LANE 104) (UNSPEC_VLD3 105) (UNSPEC_VLD3A 106) (UNSPEC_VLD3B 107) (UNSPEC_VLD3_DUP 108) (UNSPEC_VLD3_LANE 109) (UNSPEC_VLD4 110) (UNSPEC_VLD4A 111) (UNSPEC_VLD4B 112) (UNSPEC_VLD4_DUP 113) (UNSPEC_VLD4_LANE 114) (UNSPEC_VMAX 115) (UNSPEC_VMIN 116) (UNSPEC_VMLA 117) (UNSPEC_VMLAL 118) (UNSPEC_VMLA_LANE 119) (UNSPEC_VMLAL_LANE 120) (UNSPEC_VMLS 121) (UNSPEC_VMLSL 122) (UNSPEC_VMLS_LANE 123) (UNSPEC_VMLSL_LANE 124) (UNSPEC_VMOVL 125) (UNSPEC_VMOVN 126) (UNSPEC_VMUL 127) (UNSPEC_VMULL 128) (UNSPEC_VMUL_LANE 129) (UNSPEC_VMULL_LANE 130) (UNSPEC_VPADAL 135) (UNSPEC_VPADD 136) (UNSPEC_VPADDL 137) (UNSPEC_VPMAX 138) (UNSPEC_VPMIN 139) (UNSPEC_VPSMAX 140) (UNSPEC_VPSMIN 141) (UNSPEC_VPUMAX 142) (UNSPEC_VPUMIN 143) (UNSPEC_VQABS 144) (UNSPEC_VQADD 145) (UNSPEC_VQDMLAL 146) (UNSPEC_VQDMLAL_LANE 147) (UNSPEC_VQDMLSL 148) (UNSPEC_VQDMLSL_LANE 149) (UNSPEC_VQDMULH 150) (UNSPEC_VQDMULH_LANE 151) (UNSPEC_VQDMULL 152) (UNSPEC_VQDMULL_LANE 153) (UNSPEC_VQMOVN 154) (UNSPEC_VQMOVUN 155) (UNSPEC_VQNEG 156) (UNSPEC_VQSHL 157) (UNSPEC_VQSHL_N 158) (UNSPEC_VQSHLU_N 159) (UNSPEC_VQSHRN_N 160) (UNSPEC_VQSHRUN_N 161) (UNSPEC_VQSUB 162) (UNSPEC_VRECPE 163) (UNSPEC_VRECPS 164) (UNSPEC_VREV16 165) (UNSPEC_VREV32 166) (UNSPEC_VREV64 167) (UNSPEC_VRSQRTE 168) (UNSPEC_VRSQRTS 169) (UNSPEC_VSHL 171) (UNSPEC_VSHLL_N 172) (UNSPEC_VSHL_N 173) (UNSPEC_VSHR_N 174) (UNSPEC_VSHRN_N 175) (UNSPEC_VSLI 176) (UNSPEC_VSRA_N 177) (UNSPEC_VSRI 178) (UNSPEC_VST1 179) (UNSPEC_VST1_LANE 180) (UNSPEC_VST2 181) (UNSPEC_VST2_LANE 182) (UNSPEC_VST3 183) (UNSPEC_VST3A 184) (UNSPEC_VST3B 185) (UNSPEC_VST3_LANE 186) (UNSPEC_VST4 187) (UNSPEC_VST4A 188) (UNSPEC_VST4B 189) (UNSPEC_VST4_LANE 190) (UNSPEC_VSTRUCTDUMMY 191) (UNSPEC_VSUB 192) (UNSPEC_VSUBHN 193) (UNSPEC_VSUBL 194) (UNSPEC_VSUBW 195) (UNSPEC_VTBL 196) (UNSPEC_VTBX 197) (UNSPEC_VTRN1 198) (UNSPEC_VTRN2 199) (UNSPEC_VTST 200) (UNSPEC_VUZP1 201) (UNSPEC_VUZP2 202) (UNSPEC_VZIP1 203) (UNSPEC_VZIP2 204) (UNSPEC_MISALIGNED_ACCESS 205) (UNSPEC_VCLE 206) (UNSPEC_VCLT 207)]) ;; Attribute used to permit string comparisons against in ;; neon_type attribute definitions. (define_attr "vqh_mnem" "vadd,vmin,vmax" (const_string "vadd")) (define_insn "*neon_mov" [(set (match_operand:VD 0 "nonimmediate_operand" "=w,Uv,w, w, ?r,?w,?r,?r, ?Us") (match_operand:VD 1 "general_operand" " w,w, Dn,Uvi, w, r, r, Usi,r"))] "TARGET_NEON && (register_operand (operands[0], mode) || register_operand (operands[1], mode))" { if (which_alternative == 2) { int width, is_valid; static char templ[40]; is_valid = neon_immediate_valid_for_move (operands[1], mode, &operands[1], &width); gcc_assert (is_valid != 0); if (width == 0) return "vmov.f32\t%P0, %1 @ "; else sprintf (templ, "vmov.i%d\t%%P0, %%1 @ ", width); return templ; } /* FIXME: If the memory layout is changed in big-endian mode, output_move_vfp below must be changed to output_move_neon (which will use the element/structure loads/stores), and the constraint changed to 'Um' instead of 'Uv'. */ switch (which_alternative) { case 0: return "vmov\t%P0, %P1 @ "; case 1: case 3: return output_move_vfp (operands); case 2: gcc_unreachable (); case 4: return "vmov\t%Q0, %R0, %P1 @ "; case 5: return "vmov\t%P0, %Q1, %R1 @ "; default: return output_move_double (operands); } } [(set_attr "neon_type" "neon_int_1,*,neon_vmov,*,neon_mrrc,neon_mcr_2_mcrr,*,*,*") (set_attr "type" "*,f_stored,*,f_loadd,*,*,alu,load2,store2") (set_attr "insn" "*,*,*,*,*,*,mov,*,*") (set_attr "length" "4,4,4,4,4,4,8,8,8") (set_attr "pool_range" "*,*,*,1020,*,*,*,1020,*") (set_attr "neg_pool_range" "*,*,*,1008,*,*,*,1008,*")]) (define_insn "*neon_mov" [(set (match_operand:VQXMOV 0 "nonimmediate_operand" "=w,Un,w, w, ?r,?w,?r,?r, ?Us") (match_operand:VQXMOV 1 "general_operand" " w,w, Dn,Uni, w, r, r, Usi, r"))] "TARGET_NEON && (register_operand (operands[0], mode) || register_operand (operands[1], mode))" { if (which_alternative == 2) { int width, is_valid; static char templ[40]; is_valid = neon_immediate_valid_for_move (operands[1], mode, &operands[1], &width); gcc_assert (is_valid != 0); if (width == 0) return "vmov.f32\t%q0, %1 @ "; else sprintf (templ, "vmov.i%d\t%%q0, %%1 @ ", width); return templ; } switch (which_alternative) { case 0: return "vmov\t%q0, %q1 @ "; case 1: case 3: return output_move_neon (operands); case 2: gcc_unreachable (); case 4: return "vmov\t%Q0, %R0, %e1 @ \;vmov\t%J0, %K0, %f1"; case 5: return "vmov\t%e0, %Q1, %R1 @ \;vmov\t%f0, %J1, %K1"; default: return output_move_quad (operands); } } [(set_attr "neon_type" "neon_int_1,neon_stm_2,neon_vmov,neon_ldm_2,\ neon_mrrc,neon_mcr_2_mcrr,*,*,*") (set_attr "type" "*,*,*,*,*,*,alu,load4,store4") (set_attr "insn" "*,*,*,*,*,*,mov,*,*") (set_attr "length" "4,8,4,8,8,8,16,8,16") (set_attr "pool_range" "*,*,*,1020,*,*,*,1020,*") (set_attr "neg_pool_range" "*,*,*,1008,*,*,*,1008,*")]) (define_expand "movti" [(set (match_operand:TI 0 "nonimmediate_operand" "") (match_operand:TI 1 "general_operand" ""))] "TARGET_NEON" { if (can_create_pseudo_p ()) { if (GET_CODE (operands[0]) != REG) operands[1] = force_reg (TImode, operands[1]); } }) (define_expand "mov" [(set (match_operand:VSTRUCT 0 "nonimmediate_operand" "") (match_operand:VSTRUCT 1 "general_operand" ""))] "TARGET_NEON" { if (can_create_pseudo_p ()) { if (GET_CODE (operands[0]) != REG) operands[1] = force_reg (mode, operands[1]); } }) (define_insn "*neon_mov" [(set (match_operand:VSTRUCT 0 "nonimmediate_operand" "=w,Ut,w") (match_operand:VSTRUCT 1 "general_operand" " w,w, Ut"))] "TARGET_NEON && (register_operand (operands[0], mode) || register_operand (operands[1], mode))" { switch (which_alternative) { case 0: return "#"; case 1: case 2: return output_move_neon (operands); default: gcc_unreachable (); } } [(set_attr "neon_type" "neon_int_1,neon_stm_2,neon_ldm_2") (set (attr "length") (symbol_ref "arm_attr_length_move_neon (insn)"))]) (define_split [(set (match_operand:EI 0 "s_register_operand" "") (match_operand:EI 1 "s_register_operand" ""))] "TARGET_NEON && reload_completed" [(set (match_dup 0) (match_dup 1)) (set (match_dup 2) (match_dup 3))] { int rdest = REGNO (operands[0]); int rsrc = REGNO (operands[1]); rtx dest[2], src[2]; dest[0] = gen_rtx_REG (TImode, rdest); src[0] = gen_rtx_REG (TImode, rsrc); dest[1] = gen_rtx_REG (DImode, rdest + 4); src[1] = gen_rtx_REG (DImode, rsrc + 4); neon_disambiguate_copy (operands, dest, src, 2); }) (define_split [(set (match_operand:OI 0 "s_register_operand" "") (match_operand:OI 1 "s_register_operand" ""))] "TARGET_NEON && reload_completed" [(set (match_dup 0) (match_dup 1)) (set (match_dup 2) (match_dup 3))] { int rdest = REGNO (operands[0]); int rsrc = REGNO (operands[1]); rtx dest[2], src[2]; dest[0] = gen_rtx_REG (TImode, rdest); src[0] = gen_rtx_REG (TImode, rsrc); dest[1] = gen_rtx_REG (TImode, rdest + 4); src[1] = gen_rtx_REG (TImode, rsrc + 4); neon_disambiguate_copy (operands, dest, src, 2); }) (define_split [(set (match_operand:CI 0 "s_register_operand" "") (match_operand:CI 1 "s_register_operand" ""))] "TARGET_NEON && reload_completed" [(set (match_dup 0) (match_dup 1)) (set (match_dup 2) (match_dup 3)) (set (match_dup 4) (match_dup 5))] { int rdest = REGNO (operands[0]); int rsrc = REGNO (operands[1]); rtx dest[3], src[3]; dest[0] = gen_rtx_REG (TImode, rdest); src[0] = gen_rtx_REG (TImode, rsrc); dest[1] = gen_rtx_REG (TImode, rdest + 4); src[1] = gen_rtx_REG (TImode, rsrc + 4); dest[2] = gen_rtx_REG (TImode, rdest + 8); src[2] = gen_rtx_REG (TImode, rsrc + 8); neon_disambiguate_copy (operands, dest, src, 3); }) (define_split [(set (match_operand:XI 0 "s_register_operand" "") (match_operand:XI 1 "s_register_operand" ""))] "TARGET_NEON && reload_completed" [(set (match_dup 0) (match_dup 1)) (set (match_dup 2) (match_dup 3)) (set (match_dup 4) (match_dup 5)) (set (match_dup 6) (match_dup 7))] { int rdest = REGNO (operands[0]); int rsrc = REGNO (operands[1]); rtx dest[4], src[4]; dest[0] = gen_rtx_REG (TImode, rdest); src[0] = gen_rtx_REG (TImode, rsrc); dest[1] = gen_rtx_REG (TImode, rdest + 4); src[1] = gen_rtx_REG (TImode, rsrc + 4); dest[2] = gen_rtx_REG (TImode, rdest + 8); src[2] = gen_rtx_REG (TImode, rsrc + 8); dest[3] = gen_rtx_REG (TImode, rdest + 12); src[3] = gen_rtx_REG (TImode, rsrc + 12); neon_disambiguate_copy (operands, dest, src, 4); }) (define_expand "movmisalign" [(set (match_operand:VDQX 0 "nonimmediate_operand" "") (unspec:VDQX [(match_operand:VDQX 1 "general_operand" "")] UNSPEC_MISALIGNED_ACCESS))] "TARGET_NEON && !BYTES_BIG_ENDIAN" { /* This pattern is not permitted to fail during expansion: if both arguments are non-registers (e.g. memory := constant, which can be created by the auto-vectorizer), force operand 1 into a register. */ if (!s_register_operand (operands[0], mode) && !s_register_operand (operands[1], mode)) operands[1] = force_reg (mode, operands[1]); }) (define_insn "*movmisalign_neon_store" [(set (match_operand:VDX 0 "memory_operand" "=Um") (unspec:VDX [(match_operand:VDX 1 "s_register_operand" " w")] UNSPEC_MISALIGNED_ACCESS))] "TARGET_NEON && !BYTES_BIG_ENDIAN" "vst1.\t{%P1}, %A0" [(set_attr "neon_type" "neon_vst1_1_2_regs_vst2_2_regs")]) (define_insn "*movmisalign_neon_load" [(set (match_operand:VDX 0 "s_register_operand" "=w") (unspec:VDX [(match_operand:VDX 1 "memory_operand" " Um")] UNSPEC_MISALIGNED_ACCESS))] "TARGET_NEON && !BYTES_BIG_ENDIAN" "vld1.\t{%P0}, %A1" [(set_attr "neon_type" "neon_vld1_1_2_regs")]) (define_insn "*movmisalign_neon_store" [(set (match_operand:VQX 0 "memory_operand" "=Um") (unspec:VQX [(match_operand:VQX 1 "s_register_operand" " w")] UNSPEC_MISALIGNED_ACCESS))] "TARGET_NEON && !BYTES_BIG_ENDIAN" "vst1.\t{%q1}, %A0" [(set_attr "neon_type" "neon_vst1_1_2_regs_vst2_2_regs")]) (define_insn "*movmisalign_neon_load" [(set (match_operand:VQX 0 "s_register_operand" "=w") (unspec:VQX [(match_operand:VQX 1 "memory_operand" " Um")] UNSPEC_MISALIGNED_ACCESS))] "TARGET_NEON && !BYTES_BIG_ENDIAN" "vld1.\t{%q0}, %A1" [(set_attr "neon_type" "neon_vld1_1_2_regs")]) (define_insn "vec_set_internal" [(set (match_operand:VD 0 "s_register_operand" "=w") (vec_merge:VD (vec_duplicate:VD (match_operand: 1 "s_register_operand" "r")) (match_operand:VD 3 "s_register_operand" "0") (match_operand:SI 2 "immediate_operand" "i")))] "TARGET_NEON" { int elt = ffs ((int) INTVAL (operands[2])) - 1; if (BYTES_BIG_ENDIAN) elt = GET_MODE_NUNITS (mode) - 1 - elt; operands[2] = GEN_INT (elt); return "vmov%?.\t%P0[%c2], %1"; } [(set_attr "predicable" "yes") (set_attr "neon_type" "neon_mcr")]) (define_insn "vec_set_internal" [(set (match_operand:VQ 0 "s_register_operand" "=w") (vec_merge:VQ (vec_duplicate:VQ (match_operand: 1 "s_register_operand" "r")) (match_operand:VQ 3 "s_register_operand" "0") (match_operand:SI 2 "immediate_operand" "i")))] "TARGET_NEON" { HOST_WIDE_INT elem = ffs ((int) INTVAL (operands[2])) - 1; int half_elts = GET_MODE_NUNITS (mode) / 2; int elt = elem % half_elts; int hi = (elem / half_elts) * 2; int regno = REGNO (operands[0]); if (BYTES_BIG_ENDIAN) elt = half_elts - 1 - elt; operands[0] = gen_rtx_REG (mode, regno + hi); operands[2] = GEN_INT (elt); return "vmov%?.\t%P0[%c2], %1"; } [(set_attr "predicable" "yes") (set_attr "neon_type" "neon_mcr")] ) (define_insn "vec_setv2di_internal" [(set (match_operand:V2DI 0 "s_register_operand" "=w") (vec_merge:V2DI (vec_duplicate:V2DI (match_operand:DI 1 "s_register_operand" "r")) (match_operand:V2DI 3 "s_register_operand" "0") (match_operand:SI 2 "immediate_operand" "i")))] "TARGET_NEON" { HOST_WIDE_INT elem = ffs ((int) INTVAL (operands[2])) - 1; int regno = REGNO (operands[0]) + 2 * elem; operands[0] = gen_rtx_REG (DImode, regno); return "vmov%?\t%P0, %Q1, %R1"; } [(set_attr "predicable" "yes") (set_attr "neon_type" "neon_mcr_2_mcrr")] ) (define_expand "vec_set" [(match_operand:VDQ 0 "s_register_operand" "") (match_operand: 1 "s_register_operand" "") (match_operand:SI 2 "immediate_operand" "")] "TARGET_NEON" { HOST_WIDE_INT elem = (HOST_WIDE_INT) 1 << INTVAL (operands[2]); emit_insn (gen_vec_set_internal (operands[0], operands[1], GEN_INT (elem), operands[0])); DONE; }) (define_insn "vec_extract" [(set (match_operand: 0 "s_register_operand" "=r") (vec_select: (match_operand:VD 1 "s_register_operand" "w") (parallel [(match_operand:SI 2 "immediate_operand" "i")])))] "TARGET_NEON" { if (BYTES_BIG_ENDIAN) { int elt = INTVAL (operands[2]); elt = GET_MODE_NUNITS (mode) - 1 - elt; operands[2] = GEN_INT (elt); } return "vmov%?.\t%0, %P1[%c2]"; } [(set_attr "predicable" "yes") (set_attr "neon_type" "neon_bp_simple")] ) (define_insn "vec_extract" [(set (match_operand: 0 "s_register_operand" "=r") (vec_select: (match_operand:VQ 1 "s_register_operand" "w") (parallel [(match_operand:SI 2 "immediate_operand" "i")])))] "TARGET_NEON" { int half_elts = GET_MODE_NUNITS (mode) / 2; int elt = INTVAL (operands[2]) % half_elts; int hi = (INTVAL (operands[2]) / half_elts) * 2; int regno = REGNO (operands[1]); if (BYTES_BIG_ENDIAN) elt = half_elts - 1 - elt; operands[1] = gen_rtx_REG (mode, regno + hi); operands[2] = GEN_INT (elt); return "vmov%?.\t%0, %P1[%c2]"; } [(set_attr "predicable" "yes") (set_attr "neon_type" "neon_bp_simple")] ) (define_insn "vec_extractv2di" [(set (match_operand:DI 0 "s_register_operand" "=r") (vec_select:DI (match_operand:V2DI 1 "s_register_operand" "w") (parallel [(match_operand:SI 2 "immediate_operand" "i")])))] "TARGET_NEON" { int regno = REGNO (operands[1]) + 2 * INTVAL (operands[2]); operands[1] = gen_rtx_REG (DImode, regno); return "vmov%?\t%Q0, %R0, %P1 @ v2di"; } [(set_attr "predicable" "yes") (set_attr "neon_type" "neon_int_1")] ) (define_expand "vec_init" [(match_operand:VDQ 0 "s_register_operand" "") (match_operand 1 "" "")] "TARGET_NEON" { neon_expand_vector_init (operands[0], operands[1]); DONE; }) ;; Doubleword and quadword arithmetic. ;; NOTE: some other instructions also support 64-bit integer ;; element size, which we could potentially use for "long long" operations. (define_insn "*add3_neon" [(set (match_operand:VDQ 0 "s_register_operand" "=w") (plus:VDQ (match_operand:VDQ 1 "s_register_operand" "w") (match_operand:VDQ 2 "s_register_operand" "w")))] "TARGET_NEON && (! || flag_unsafe_math_optimizations)" "vadd.\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (const_string "neon_int_1")))] ) (define_insn "adddi3_neon" [(set (match_operand:DI 0 "s_register_operand" "=w,?&r,?&r") (plus:DI (match_operand:DI 1 "s_register_operand" "%w,0,0") (match_operand:DI 2 "s_register_operand" "w,r,0"))) (clobber (reg:CC CC_REGNUM))] "TARGET_NEON" { switch (which_alternative) { case 0: return "vadd.i64\t%P0, %P1, %P2"; case 1: return "#"; case 2: return "#"; default: gcc_unreachable (); } } [(set_attr "neon_type" "neon_int_1,*,*") (set_attr "conds" "*,clob,clob") (set_attr "length" "*,8,8")] ) (define_insn "*sub3_neon" [(set (match_operand:VDQ 0 "s_register_operand" "=w") (minus:VDQ (match_operand:VDQ 1 "s_register_operand" "w") (match_operand:VDQ 2 "s_register_operand" "w")))] "TARGET_NEON && (! || flag_unsafe_math_optimizations)" "vsub.\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (const_string "neon_int_2")))] ) (define_insn "subdi3_neon" [(set (match_operand:DI 0 "s_register_operand" "=w,?&r,?&r,?&r") (minus:DI (match_operand:DI 1 "s_register_operand" "w,0,r,0") (match_operand:DI 2 "s_register_operand" "w,r,0,0"))) (clobber (reg:CC CC_REGNUM))] "TARGET_NEON" { switch (which_alternative) { case 0: return "vsub.i64\t%P0, %P1, %P2"; case 1: /* fall through */ case 2: /* fall through */ case 3: return "subs\\t%Q0, %Q1, %Q2\;sbc\\t%R0, %R1, %R2"; default: gcc_unreachable (); } } [(set_attr "neon_type" "neon_int_2,*,*,*") (set_attr "conds" "*,clob,clob,clob") (set_attr "length" "*,8,8,8")] ) (define_insn "*mul3_neon" [(set (match_operand:VDQ 0 "s_register_operand" "=w") (mult:VDQ (match_operand:VDQ 1 "s_register_operand" "w") (match_operand:VDQ 2 "s_register_operand" "w")))] "TARGET_NEON && (! || flag_unsafe_math_optimizations)" "vmul.\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mul_ddd_8_16_qdd_16_8_long_32_16_long") (const_string "neon_mul_qqq_8_16_32_ddd_32")) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mul_qqq_8_16_32_ddd_32") (const_string "neon_mul_qqq_8_16_32_ddd_32")))))] ) (define_insn "mul3add_neon" [(set (match_operand:VDQ 0 "s_register_operand" "=w") (plus:VDQ (mult:VDQ (match_operand:VDQ 2 "s_register_operand" "w") (match_operand:VDQ 3 "s_register_operand" "w")) (match_operand:VDQ 1 "s_register_operand" "0")))] "TARGET_NEON && (! || flag_unsafe_math_optimizations)" "vmla.\t%0, %2, %3" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vmla_ddd") (const_string "neon_fp_vmla_qqq")) (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_8_16_qdd_16_8_long_32_16_long") (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long")) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_qqq_8_16") (const_string "neon_mla_qqq_32_qqd_32_scalar")))))] ) (define_insn "mul3negadd_neon" [(set (match_operand:VDQ 0 "s_register_operand" "=w") (minus:VDQ (match_operand:VDQ 1 "s_register_operand" "0") (mult:VDQ (match_operand:VDQ 2 "s_register_operand" "w") (match_operand:VDQ 3 "s_register_operand" "w"))))] "TARGET_NEON && (! || flag_unsafe_math_optimizations)" "vmls.\t%0, %2, %3" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vmla_ddd") (const_string "neon_fp_vmla_qqq")) (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_8_16_qdd_16_8_long_32_16_long") (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long")) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_qqq_8_16") (const_string "neon_mla_qqq_32_qqd_32_scalar")))))] ) (define_insn "ior3" [(set (match_operand:VDQ 0 "s_register_operand" "=w,w") (ior:VDQ (match_operand:VDQ 1 "s_register_operand" "w,0") (match_operand:VDQ 2 "neon_logic_op2" "w,Dl")))] "TARGET_NEON" { switch (which_alternative) { case 0: return "vorr\t%0, %1, %2"; case 1: return neon_output_logic_immediate ("vorr", &operands[2], mode, 0, VALID_NEON_QREG_MODE (mode)); default: gcc_unreachable (); } } [(set_attr "neon_type" "neon_int_1")] ) (define_insn "iordi3_neon" [(set (match_operand:DI 0 "s_register_operand" "=w,w,?&r,?&r") (ior:DI (match_operand:DI 1 "s_register_operand" "%w,0,0,r") (match_operand:DI 2 "neon_logic_op2" "w,Dl,r,r")))] "TARGET_NEON" { switch (which_alternative) { case 0: return "vorr\t%P0, %P1, %P2"; case 1: return neon_output_logic_immediate ("vorr", &operands[2], DImode, 0, VALID_NEON_QREG_MODE (DImode)); case 2: return "#"; case 3: return "#"; default: gcc_unreachable (); } } [(set_attr "neon_type" "neon_int_1,neon_int_1,*,*") (set_attr "length" "*,*,8,8")] ) ;; The concrete forms of the Neon immediate-logic instructions are vbic and ;; vorr. We support the pseudo-instruction vand instead, because that ;; corresponds to the canonical form the middle-end expects to use for ;; immediate bitwise-ANDs. (define_insn "and3" [(set (match_operand:VDQ 0 "s_register_operand" "=w,w") (and:VDQ (match_operand:VDQ 1 "s_register_operand" "w,0") (match_operand:VDQ 2 "neon_inv_logic_op2" "w,DL")))] "TARGET_NEON" { switch (which_alternative) { case 0: return "vand\t%0, %1, %2"; case 1: return neon_output_logic_immediate ("vand", &operands[2], mode, 1, VALID_NEON_QREG_MODE (mode)); default: gcc_unreachable (); } } [(set_attr "neon_type" "neon_int_1")] ) (define_insn "anddi3_neon" [(set (match_operand:DI 0 "s_register_operand" "=w,w,?&r,?&r") (and:DI (match_operand:DI 1 "s_register_operand" "%w,0,0,r") (match_operand:DI 2 "neon_inv_logic_op2" "w,DL,r,r")))] "TARGET_NEON" { switch (which_alternative) { case 0: return "vand\t%P0, %P1, %P2"; case 1: return neon_output_logic_immediate ("vand", &operands[2], DImode, 1, VALID_NEON_QREG_MODE (DImode)); case 2: return "#"; case 3: return "#"; default: gcc_unreachable (); } } [(set_attr "neon_type" "neon_int_1,neon_int_1,*,*") (set_attr "length" "*,*,8,8")] ) (define_insn "orn3_neon" [(set (match_operand:VDQ 0 "s_register_operand" "=w") (ior:VDQ (match_operand:VDQ 1 "s_register_operand" "w") (not:VDQ (match_operand:VDQ 2 "s_register_operand" "w"))))] "TARGET_NEON" "vorn\t%0, %1, %2" [(set_attr "neon_type" "neon_int_1")] ) (define_insn "orndi3_neon" [(set (match_operand:DI 0 "s_register_operand" "=w,?=&r,?&r") (ior:DI (match_operand:DI 1 "s_register_operand" "w,r,0") (not:DI (match_operand:DI 2 "s_register_operand" "w,0,r"))))] "TARGET_NEON" "@ vorn\t%P0, %P1, %P2 # #" [(set_attr "neon_type" "neon_int_1,*,*") (set_attr "length" "*,8,8")] ) (define_insn "bic3_neon" [(set (match_operand:VDQ 0 "s_register_operand" "=w") (and:VDQ (match_operand:VDQ 1 "s_register_operand" "w") (not:VDQ (match_operand:VDQ 2 "s_register_operand" "w"))))] "TARGET_NEON" "vbic\t%0, %1, %2" [(set_attr "neon_type" "neon_int_1")] ) ;; Compare to *anddi_notdi_di. (define_insn "bicdi3_neon" [(set (match_operand:DI 0 "s_register_operand" "=w,?=&r,?&r") (and:DI (not:DI (match_operand:DI 2 "s_register_operand" "w,r,0")) (match_operand:DI 1 "s_register_operand" "w,0,r")))] "TARGET_NEON" "@ vbic\t%P0, %P1, %P2 # #" [(set_attr "neon_type" "neon_int_1,*,*") (set_attr "length" "*,8,8")] ) (define_insn "xor3" [(set (match_operand:VDQ 0 "s_register_operand" "=w") (xor:VDQ (match_operand:VDQ 1 "s_register_operand" "w") (match_operand:VDQ 2 "s_register_operand" "w")))] "TARGET_NEON" "veor\t%0, %1, %2" [(set_attr "neon_type" "neon_int_1")] ) (define_insn "xordi3_neon" [(set (match_operand:DI 0 "s_register_operand" "=w,?&r,?&r") (xor:DI (match_operand:DI 1 "s_register_operand" "%w,0,r") (match_operand:DI 2 "s_register_operand" "w,r,r")))] "TARGET_NEON" "@ veor\t%P0, %P1, %P2 # #" [(set_attr "neon_type" "neon_int_1,*,*") (set_attr "length" "*,8,8")] ) (define_insn "one_cmpl2" [(set (match_operand:VDQ 0 "s_register_operand" "=w") (not:VDQ (match_operand:VDQ 1 "s_register_operand" "w")))] "TARGET_NEON" "vmvn\t%0, %1" [(set_attr "neon_type" "neon_int_1")] ) (define_insn "abs2" [(set (match_operand:VDQW 0 "s_register_operand" "=w") (abs:VDQW (match_operand:VDQW 1 "s_register_operand" "w")))] "TARGET_NEON" "vabs.\t%0, %1" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (const_string "neon_int_3")))] ) (define_insn "neg2" [(set (match_operand:VDQW 0 "s_register_operand" "=w") (neg:VDQW (match_operand:VDQW 1 "s_register_operand" "w")))] "TARGET_NEON" "vneg.\t%0, %1" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (const_string "neon_int_3")))] ) (define_insn "*umin3_neon" [(set (match_operand:VDQIW 0 "s_register_operand" "=w") (umin:VDQIW (match_operand:VDQIW 1 "s_register_operand" "w") (match_operand:VDQIW 2 "s_register_operand" "w")))] "TARGET_NEON" "vmin.\t%0, %1, %2" [(set_attr "neon_type" "neon_int_5")] ) (define_insn "*umax3_neon" [(set (match_operand:VDQIW 0 "s_register_operand" "=w") (umax:VDQIW (match_operand:VDQIW 1 "s_register_operand" "w") (match_operand:VDQIW 2 "s_register_operand" "w")))] "TARGET_NEON" "vmax.\t%0, %1, %2" [(set_attr "neon_type" "neon_int_5")] ) (define_insn "*smin3_neon" [(set (match_operand:VDQW 0 "s_register_operand" "=w") (smin:VDQW (match_operand:VDQW 1 "s_register_operand" "w") (match_operand:VDQW 2 "s_register_operand" "w")))] "TARGET_NEON" "vmin.\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_int_5")))] ) (define_insn "*smax3_neon" [(set (match_operand:VDQW 0 "s_register_operand" "=w") (smax:VDQW (match_operand:VDQW 1 "s_register_operand" "w") (match_operand:VDQW 2 "s_register_operand" "w")))] "TARGET_NEON" "vmax.\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_int_5")))] ) ; TODO: V2DI shifts are current disabled because there are bugs in the ; generic vectorizer code. It ends up creating a V2DI constructor with ; SImode elements. (define_insn "vashl3" [(set (match_operand:VDQIW 0 "s_register_operand" "=w") (ashift:VDQIW (match_operand:VDQIW 1 "s_register_operand" "w") (match_operand:VDQIW 2 "s_register_operand" "w")))] "TARGET_NEON" "vshl.\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_vshl_ddd") (const_string "neon_shift_3")))] ) ; Used for implementing logical shift-right, which is a left-shift by a negative ; amount, with signed operands. This is essentially the same as ashl3 ; above, but using an unspec in case GCC tries anything tricky with negative ; shift amounts. (define_insn "ashl3_signed" [(set (match_operand:VDQI 0 "s_register_operand" "=w") (unspec:VDQI [(match_operand:VDQI 1 "s_register_operand" "w") (match_operand:VDQI 2 "s_register_operand" "w")] UNSPEC_ASHIFT_SIGNED))] "TARGET_NEON" "vshl.\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_vshl_ddd") (const_string "neon_shift_3")))] ) ; Used for implementing logical shift-right, which is a left-shift by a negative ; amount, with unsigned operands. (define_insn "ashl3_unsigned" [(set (match_operand:VDQI 0 "s_register_operand" "=w") (unspec:VDQI [(match_operand:VDQI 1 "s_register_operand" "w") (match_operand:VDQI 2 "s_register_operand" "w")] UNSPEC_ASHIFT_UNSIGNED))] "TARGET_NEON" "vshl.\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_vshl_ddd") (const_string "neon_shift_3")))] ) (define_expand "vashr3" [(set (match_operand:VDQIW 0 "s_register_operand" "") (ashiftrt:VDQIW (match_operand:VDQIW 1 "s_register_operand" "") (match_operand:VDQIW 2 "s_register_operand" "")))] "TARGET_NEON" { rtx neg = gen_reg_rtx (mode); emit_insn (gen_neg2 (neg, operands[2])); emit_insn (gen_ashl3_signed (operands[0], operands[1], neg)); DONE; }) (define_expand "vlshr3" [(set (match_operand:VDQIW 0 "s_register_operand" "") (lshiftrt:VDQIW (match_operand:VDQIW 1 "s_register_operand" "") (match_operand:VDQIW 2 "s_register_operand" "")))] "TARGET_NEON" { rtx neg = gen_reg_rtx (mode); emit_insn (gen_neg2 (neg, operands[2])); emit_insn (gen_ashl3_unsigned (operands[0], operands[1], neg)); DONE; }) ;; Widening operations (define_insn "widen_ssum3" [(set (match_operand: 0 "s_register_operand" "=w") (plus: (sign_extend: (match_operand:VW 1 "s_register_operand" "%w")) (match_operand: 2 "s_register_operand" "w")))] "TARGET_NEON" "vaddw.\t%q0, %q2, %P1" [(set_attr "neon_type" "neon_int_3")] ) (define_insn "widen_usum3" [(set (match_operand: 0 "s_register_operand" "=w") (plus: (zero_extend: (match_operand:VW 1 "s_register_operand" "%w")) (match_operand: 2 "s_register_operand" "w")))] "TARGET_NEON" "vaddw.\t%q0, %q2, %P1" [(set_attr "neon_type" "neon_int_3")] ) ;; VEXT can be used to synthesize coarse whole-vector shifts with 8-bit ;; shift-count granularity. That's good enough for the middle-end's current ;; needs. (define_expand "vec_shr_" [(match_operand:VDQ 0 "s_register_operand" "") (match_operand:VDQ 1 "s_register_operand" "") (match_operand:SI 2 "const_multiple_of_8_operand" "")] "TARGET_NEON" { rtx zero_reg; HOST_WIDE_INT num_bits = INTVAL (operands[2]); const int width = GET_MODE_BITSIZE (mode); const enum machine_mode bvecmode = (width == 128) ? V16QImode : V8QImode; rtx (*gen_ext) (rtx, rtx, rtx, rtx) = (width == 128) ? gen_neon_vextv16qi : gen_neon_vextv8qi; if (num_bits == width) { emit_move_insn (operands[0], operands[1]); DONE; } zero_reg = force_reg (bvecmode, CONST0_RTX (bvecmode)); operands[0] = gen_lowpart (bvecmode, operands[0]); operands[1] = gen_lowpart (bvecmode, operands[1]); emit_insn (gen_ext (operands[0], operands[1], zero_reg, GEN_INT (num_bits / BITS_PER_UNIT))); DONE; }) (define_expand "vec_shl_" [(match_operand:VDQ 0 "s_register_operand" "") (match_operand:VDQ 1 "s_register_operand" "") (match_operand:SI 2 "const_multiple_of_8_operand" "")] "TARGET_NEON" { rtx zero_reg; HOST_WIDE_INT num_bits = INTVAL (operands[2]); const int width = GET_MODE_BITSIZE (mode); const enum machine_mode bvecmode = (width == 128) ? V16QImode : V8QImode; rtx (*gen_ext) (rtx, rtx, rtx, rtx) = (width == 128) ? gen_neon_vextv16qi : gen_neon_vextv8qi; if (num_bits == 0) { emit_move_insn (operands[0], CONST0_RTX (mode)); DONE; } num_bits = width - num_bits; zero_reg = force_reg (bvecmode, CONST0_RTX (bvecmode)); operands[0] = gen_lowpart (bvecmode, operands[0]); operands[1] = gen_lowpart (bvecmode, operands[1]); emit_insn (gen_ext (operands[0], zero_reg, operands[1], GEN_INT (num_bits / BITS_PER_UNIT))); DONE; }) ;; Helpers for quad-word reduction operations ; Add (or smin, smax...) the low N/2 elements of the N-element vector ; operand[1] to the high N/2 elements of same. Put the result in operand[0], an ; N/2-element vector. (define_insn "quad_halves_v4si" [(set (match_operand:V2SI 0 "s_register_operand" "=w") (vqh_ops:V2SI (vec_select:V2SI (match_operand:V4SI 1 "s_register_operand" "w") (parallel [(const_int 0) (const_int 1)])) (vec_select:V2SI (match_dup 1) (parallel [(const_int 2) (const_int 3)]))))] "TARGET_NEON" ".32\t%P0, %e1, %f1" [(set_attr "vqh_mnem" "") (set (attr "neon_type") (if_then_else (eq_attr "vqh_mnem" "vadd") (const_string "neon_int_1") (const_string "neon_int_5")))] ) (define_insn "quad_halves_v4sf" [(set (match_operand:V2SF 0 "s_register_operand" "=w") (vqhs_ops:V2SF (vec_select:V2SF (match_operand:V4SF 1 "s_register_operand" "w") (parallel [(const_int 0) (const_int 1)])) (vec_select:V2SF (match_dup 1) (parallel [(const_int 2) (const_int 3)]))))] "TARGET_NEON && flag_unsafe_math_optimizations" ".f32\t%P0, %e1, %f1" [(set_attr "vqh_mnem" "") (set (attr "neon_type") (if_then_else (eq_attr "vqh_mnem" "vadd") (const_string "neon_int_1") (const_string "neon_int_5")))] ) (define_insn "quad_halves_v8hi" [(set (match_operand:V4HI 0 "s_register_operand" "+w") (vqh_ops:V4HI (vec_select:V4HI (match_operand:V8HI 1 "s_register_operand" "w") (parallel [(const_int 0) (const_int 1) (const_int 2) (const_int 3)])) (vec_select:V4HI (match_dup 1) (parallel [(const_int 4) (const_int 5) (const_int 6) (const_int 7)]))))] "TARGET_NEON" ".16\t%P0, %e1, %f1" [(set_attr "vqh_mnem" "") (set (attr "neon_type") (if_then_else (eq_attr "vqh_mnem" "vadd") (const_string "neon_int_1") (const_string "neon_int_5")))] ) (define_insn "quad_halves_v16qi" [(set (match_operand:V8QI 0 "s_register_operand" "+w") (vqh_ops:V8QI (vec_select:V8QI (match_operand:V16QI 1 "s_register_operand" "w") (parallel [(const_int 0) (const_int 1) (const_int 2) (const_int 3) (const_int 4) (const_int 5) (const_int 6) (const_int 7)])) (vec_select:V8QI (match_dup 1) (parallel [(const_int 8) (const_int 9) (const_int 10) (const_int 11) (const_int 12) (const_int 13) (const_int 14) (const_int 15)]))))] "TARGET_NEON" ".8\t%P0, %e1, %f1" [(set_attr "vqh_mnem" "") (set (attr "neon_type") (if_then_else (eq_attr "vqh_mnem" "vadd") (const_string "neon_int_1") (const_string "neon_int_5")))] ) ; FIXME: We wouldn't need the following insns if we could write subregs of ; vector registers. Make an attempt at removing unnecessary moves, though ; we're really at the mercy of the register allocator. (define_insn "neon_move_lo_quad_" [(set (match_operand:ANY128 0 "s_register_operand" "+w") (vec_concat:ANY128 (match_operand: 1 "s_register_operand" "w") (vec_select: (match_dup 0) (match_operand:ANY128 2 "vect_par_constant_high" ""))))] "TARGET_NEON" { int dest = REGNO (operands[0]); int src = REGNO (operands[1]); if (dest != src) return "vmov\t%e0, %P1"; else return ""; } [(set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_move_hi_quad_" [(set (match_operand:ANY128 0 "s_register_operand" "+w") (vec_concat:ANY128 (vec_select: (match_dup 0) (match_operand:ANY128 2 "vect_par_constant_low" "")) (match_operand: 1 "s_register_operand" "w")))] "TARGET_NEON" { int dest = REGNO (operands[0]); int src = REGNO (operands[1]); if (dest != src) return "vmov\t%f0, %P1"; else return ""; } [(set_attr "neon_type" "neon_bp_simple")] ) (define_expand "move_hi_quad_" [(match_operand:ANY128 0 "s_register_operand" "") (match_operand: 1 "s_register_operand" "")] "TARGET_NEON" { rtvec v = rtvec_alloc (/2); rtx t1; int i; for (i=0; i < (/2); i++) RTVEC_ELT (v, i) = GEN_INT (i); t1 = gen_rtx_PARALLEL (mode, v); emit_insn (gen_neon_move_hi_quad_ (operands[0], operands[1], t1)); DONE; }) (define_expand "move_lo_quad_" [(match_operand:ANY128 0 "s_register_operand" "") (match_operand: 1 "s_register_operand" "")] "TARGET_NEON" { rtvec v = rtvec_alloc (/2); rtx t1; int i; for (i=0; i < (/2); i++) RTVEC_ELT (v, i) = GEN_INT ((/2) + i); t1 = gen_rtx_PARALLEL (mode, v); emit_insn (gen_neon_move_lo_quad_ (operands[0], operands[1], t1)); DONE; }) ;; Reduction operations (define_expand "reduc_splus_" [(match_operand:VD 0 "s_register_operand" "") (match_operand:VD 1 "s_register_operand" "")] "TARGET_NEON && (! || flag_unsafe_math_optimizations)" { neon_pairwise_reduce (operands[0], operands[1], mode, &gen_neon_vpadd_internal); DONE; }) (define_expand "reduc_splus_" [(match_operand:VQ 0 "s_register_operand" "") (match_operand:VQ 1 "s_register_operand" "")] "TARGET_NEON && (! || flag_unsafe_math_optimizations)" { rtx step1 = gen_reg_rtx (mode); rtx res_d = gen_reg_rtx (mode); emit_insn (gen_quad_halves_plus (step1, operands[1])); emit_insn (gen_reduc_splus_ (res_d, step1)); emit_insn (gen_move_lo_quad_ (operands[0], res_d)); DONE; }) (define_insn "reduc_splus_v2di" [(set (match_operand:V2DI 0 "s_register_operand" "=w") (unspec:V2DI [(match_operand:V2DI 1 "s_register_operand" "w")] UNSPEC_VPADD))] "TARGET_NEON" "vadd.i64\t%e0, %e1, %f1" [(set_attr "neon_type" "neon_int_1")] ) ;; NEON does not distinguish between signed and unsigned addition except on ;; widening operations. (define_expand "reduc_uplus_" [(match_operand:VDQI 0 "s_register_operand" "") (match_operand:VDQI 1 "s_register_operand" "")] "TARGET_NEON" { emit_insn (gen_reduc_splus_ (operands[0], operands[1])); DONE; }) (define_expand "reduc_smin_" [(match_operand:VD 0 "s_register_operand" "") (match_operand:VD 1 "s_register_operand" "")] "TARGET_NEON && (! || flag_unsafe_math_optimizations)" { neon_pairwise_reduce (operands[0], operands[1], mode, &gen_neon_vpsmin); DONE; }) (define_expand "reduc_smin_" [(match_operand:VQ 0 "s_register_operand" "") (match_operand:VQ 1 "s_register_operand" "")] "TARGET_NEON && (! || flag_unsafe_math_optimizations)" { rtx step1 = gen_reg_rtx (mode); rtx res_d = gen_reg_rtx (mode); emit_insn (gen_quad_halves_smin (step1, operands[1])); emit_insn (gen_reduc_smin_ (res_d, step1)); emit_insn (gen_move_lo_quad_ (operands[0], res_d)); DONE; }) (define_expand "reduc_smax_" [(match_operand:VD 0 "s_register_operand" "") (match_operand:VD 1 "s_register_operand" "")] "TARGET_NEON && (! || flag_unsafe_math_optimizations)" { neon_pairwise_reduce (operands[0], operands[1], mode, &gen_neon_vpsmax); DONE; }) (define_expand "reduc_smax_" [(match_operand:VQ 0 "s_register_operand" "") (match_operand:VQ 1 "s_register_operand" "")] "TARGET_NEON && (! || flag_unsafe_math_optimizations)" { rtx step1 = gen_reg_rtx (mode); rtx res_d = gen_reg_rtx (mode); emit_insn (gen_quad_halves_smax (step1, operands[1])); emit_insn (gen_reduc_smax_ (res_d, step1)); emit_insn (gen_move_lo_quad_ (operands[0], res_d)); DONE; }) (define_expand "reduc_umin_" [(match_operand:VDI 0 "s_register_operand" "") (match_operand:VDI 1 "s_register_operand" "")] "TARGET_NEON" { neon_pairwise_reduce (operands[0], operands[1], mode, &gen_neon_vpumin); DONE; }) (define_expand "reduc_umin_" [(match_operand:VQI 0 "s_register_operand" "") (match_operand:VQI 1 "s_register_operand" "")] "TARGET_NEON" { rtx step1 = gen_reg_rtx (mode); rtx res_d = gen_reg_rtx (mode); emit_insn (gen_quad_halves_umin (step1, operands[1])); emit_insn (gen_reduc_umin_ (res_d, step1)); emit_insn (gen_move_lo_quad_ (operands[0], res_d)); DONE; }) (define_expand "reduc_umax_" [(match_operand:VDI 0 "s_register_operand" "") (match_operand:VDI 1 "s_register_operand" "")] "TARGET_NEON" { neon_pairwise_reduce (operands[0], operands[1], mode, &gen_neon_vpumax); DONE; }) (define_expand "reduc_umax_" [(match_operand:VQI 0 "s_register_operand" "") (match_operand:VQI 1 "s_register_operand" "")] "TARGET_NEON" { rtx step1 = gen_reg_rtx (mode); rtx res_d = gen_reg_rtx (mode); emit_insn (gen_quad_halves_umax (step1, operands[1])); emit_insn (gen_reduc_umax_ (res_d, step1)); emit_insn (gen_move_lo_quad_ (operands[0], res_d)); DONE; }) (define_insn "neon_vpadd_internal" [(set (match_operand:VD 0 "s_register_operand" "=w") (unspec:VD [(match_operand:VD 1 "s_register_operand" "w") (match_operand:VD 2 "s_register_operand" "w")] UNSPEC_VPADD))] "TARGET_NEON" "vpadd.\t%P0, %P1, %P2" ;; Assume this schedules like vadd. [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (const_string "neon_int_1")))] ) (define_insn "neon_vpsmin" [(set (match_operand:VD 0 "s_register_operand" "=w") (unspec:VD [(match_operand:VD 1 "s_register_operand" "w") (match_operand:VD 2 "s_register_operand" "w")] UNSPEC_VPSMIN))] "TARGET_NEON" "vpmin.\t%P0, %P1, %P2" ;; Assume this schedules like vmin. [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_int_5")))] ) (define_insn "neon_vpsmax" [(set (match_operand:VD 0 "s_register_operand" "=w") (unspec:VD [(match_operand:VD 1 "s_register_operand" "w") (match_operand:VD 2 "s_register_operand" "w")] UNSPEC_VPSMAX))] "TARGET_NEON" "vpmax.\t%P0, %P1, %P2" ;; Assume this schedules like vmax. [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_int_5")))] ) (define_insn "neon_vpumin" [(set (match_operand:VDI 0 "s_register_operand" "=w") (unspec:VDI [(match_operand:VDI 1 "s_register_operand" "w") (match_operand:VDI 2 "s_register_operand" "w")] UNSPEC_VPUMIN))] "TARGET_NEON" "vpmin.\t%P0, %P1, %P2" ;; Assume this schedules like umin. [(set_attr "neon_type" "neon_int_5")] ) (define_insn "neon_vpumax" [(set (match_operand:VDI 0 "s_register_operand" "=w") (unspec:VDI [(match_operand:VDI 1 "s_register_operand" "w") (match_operand:VDI 2 "s_register_operand" "w")] UNSPEC_VPUMAX))] "TARGET_NEON" "vpmax.\t%P0, %P1, %P2" ;; Assume this schedules like umax. [(set_attr "neon_type" "neon_int_5")] ) ;; Saturating arithmetic ; NOTE: Neon supports many more saturating variants of instructions than the ; following, but these are all GCC currently understands. ; FIXME: Actually, GCC doesn't know how to create saturating add/sub by itself ; yet either, although these patterns may be used by intrinsics when they're ; added. (define_insn "*ss_add_neon" [(set (match_operand:VD 0 "s_register_operand" "=w") (ss_plus:VD (match_operand:VD 1 "s_register_operand" "w") (match_operand:VD 2 "s_register_operand" "w")))] "TARGET_NEON" "vqadd.\t%P0, %P1, %P2" [(set_attr "neon_type" "neon_int_4")] ) (define_insn "*us_add_neon" [(set (match_operand:VD 0 "s_register_operand" "=w") (us_plus:VD (match_operand:VD 1 "s_register_operand" "w") (match_operand:VD 2 "s_register_operand" "w")))] "TARGET_NEON" "vqadd.\t%P0, %P1, %P2" [(set_attr "neon_type" "neon_int_4")] ) (define_insn "*ss_sub_neon" [(set (match_operand:VD 0 "s_register_operand" "=w") (ss_minus:VD (match_operand:VD 1 "s_register_operand" "w") (match_operand:VD 2 "s_register_operand" "w")))] "TARGET_NEON" "vqsub.\t%P0, %P1, %P2" [(set_attr "neon_type" "neon_int_5")] ) (define_insn "*us_sub_neon" [(set (match_operand:VD 0 "s_register_operand" "=w") (us_minus:VD (match_operand:VD 1 "s_register_operand" "w") (match_operand:VD 2 "s_register_operand" "w")))] "TARGET_NEON" "vqsub.\t%P0, %P1, %P2" [(set_attr "neon_type" "neon_int_5")] ) ;; Conditional instructions. These are comparisons with conditional moves for ;; vectors. They perform the assignment: ;; ;; Vop0 = (Vop4 Vop5) ? Vop1 : Vop2; ;; ;; where op3 is <, <=, ==, !=, >= or >. Operations are performed ;; element-wise. (define_expand "vcond" [(set (match_operand:VDQW 0 "s_register_operand" "") (if_then_else:VDQW (match_operator 3 "arm_comparison_operator" [(match_operand:VDQW 4 "s_register_operand" "") (match_operand:VDQW 5 "nonmemory_operand" "")]) (match_operand:VDQW 1 "s_register_operand" "") (match_operand:VDQW 2 "s_register_operand" "")))] "TARGET_NEON && (! || flag_unsafe_math_optimizations)" { rtx mask; int inverse = 0, immediate_zero = 0; /* See the description of "magic" bits in the 'T' case of arm_print_operand. */ HOST_WIDE_INT magic_word = (mode == V2SFmode || mode == V4SFmode) ? 3 : 1; rtx magic_rtx = GEN_INT (magic_word); mask = gen_reg_rtx (mode); if (operands[5] == CONST0_RTX (mode)) immediate_zero = 1; else if (!REG_P (operands[5])) operands[5] = force_reg (mode, operands[5]); switch (GET_CODE (operands[3])) { case GE: emit_insn (gen_neon_vcge (mask, operands[4], operands[5], magic_rtx)); break; case GT: emit_insn (gen_neon_vcgt (mask, operands[4], operands[5], magic_rtx)); break; case EQ: emit_insn (gen_neon_vceq (mask, operands[4], operands[5], magic_rtx)); break; case LE: if (immediate_zero) emit_insn (gen_neon_vcle (mask, operands[4], operands[5], magic_rtx)); else emit_insn (gen_neon_vcge (mask, operands[5], operands[4], magic_rtx)); break; case LT: if (immediate_zero) emit_insn (gen_neon_vclt (mask, operands[4], operands[5], magic_rtx)); else emit_insn (gen_neon_vcgt (mask, operands[5], operands[4], magic_rtx)); break; case NE: emit_insn (gen_neon_vceq (mask, operands[4], operands[5], magic_rtx)); inverse = 1; break; default: gcc_unreachable (); } if (inverse) emit_insn (gen_neon_vbsl (operands[0], mask, operands[2], operands[1])); else emit_insn (gen_neon_vbsl (operands[0], mask, operands[1], operands[2])); DONE; }) (define_expand "vcondu" [(set (match_operand:VDQIW 0 "s_register_operand" "") (if_then_else:VDQIW (match_operator 3 "arm_comparison_operator" [(match_operand:VDQIW 4 "s_register_operand" "") (match_operand:VDQIW 5 "s_register_operand" "")]) (match_operand:VDQIW 1 "s_register_operand" "") (match_operand:VDQIW 2 "s_register_operand" "")))] "TARGET_NEON" { rtx mask; int inverse = 0, immediate_zero = 0; mask = gen_reg_rtx (mode); if (operands[5] == CONST0_RTX (mode)) immediate_zero = 1; else if (!REG_P (operands[5])) operands[5] = force_reg (mode, operands[5]); switch (GET_CODE (operands[3])) { case GEU: emit_insn (gen_neon_vcge (mask, operands[4], operands[5], const0_rtx)); break; case GTU: emit_insn (gen_neon_vcgt (mask, operands[4], operands[5], const0_rtx)); break; case EQ: emit_insn (gen_neon_vceq (mask, operands[4], operands[5], const0_rtx)); break; case LEU: if (immediate_zero) emit_insn (gen_neon_vcle (mask, operands[4], operands[5], const0_rtx)); else emit_insn (gen_neon_vcge (mask, operands[5], operands[4], const0_rtx)); break; case LTU: if (immediate_zero) emit_insn (gen_neon_vclt (mask, operands[4], operands[5], const0_rtx)); else emit_insn (gen_neon_vcgt (mask, operands[5], operands[4], const0_rtx)); break; case NE: emit_insn (gen_neon_vceq (mask, operands[4], operands[5], const0_rtx)); inverse = 1; break; default: gcc_unreachable (); } if (inverse) emit_insn (gen_neon_vbsl (operands[0], mask, operands[2], operands[1])); else emit_insn (gen_neon_vbsl (operands[0], mask, operands[1], operands[2])); DONE; }) ;; Patterns for builtins. ; good for plain vadd, vaddq. (define_expand "neon_vadd" [(match_operand:VDQX 0 "s_register_operand" "=w") (match_operand:VDQX 1 "s_register_operand" "w") (match_operand:VDQX 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] "TARGET_NEON" { if (! || flag_unsafe_math_optimizations) emit_insn (gen_add3 (operands[0], operands[1], operands[2])); else emit_insn (gen_neon_vadd_unspec (operands[0], operands[1], operands[2])); DONE; }) ; Note that NEON operations don't support the full IEEE 754 standard: in ; particular, denormal values are flushed to zero. This means that GCC cannot ; use those instructions for autovectorization, etc. unless ; -funsafe-math-optimizations is in effect (in which case flush-to-zero ; behaviour is permissible). Intrinsic operations (provided by the arm_neon.h ; header) must work in either case: if -funsafe-math-optimizations is given, ; intrinsics expand to "canonical" RTL where possible, otherwise intrinsics ; expand to unspecs (which may potentially limit the extent to which they might ; be optimized by generic code). ; Used for intrinsics when flag_unsafe_math_optimizations is false. (define_insn "neon_vadd_unspec" [(set (match_operand:VDQX 0 "s_register_operand" "=w") (unspec:VDQX [(match_operand:VDQX 1 "s_register_operand" "w") (match_operand:VDQX 2 "s_register_operand" "w")] UNSPEC_VADD))] "TARGET_NEON" "vadd.\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (const_string "neon_int_1")))] ) ; operand 3 represents in bits: ; bit 0: signed (vs unsigned). ; bit 1: rounding (vs none). (define_insn "neon_vaddl" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VDI 1 "s_register_operand" "w") (match_operand:VDI 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VADDL))] "TARGET_NEON" "vaddl.%T3%#\t%q0, %P1, %P2" [(set_attr "neon_type" "neon_int_3")] ) (define_insn "neon_vaddw" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand: 1 "s_register_operand" "w") (match_operand:VDI 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VADDW))] "TARGET_NEON" "vaddw.%T3%#\t%q0, %q1, %P2" [(set_attr "neon_type" "neon_int_2")] ) ; vhadd and vrhadd. (define_insn "neon_vhadd" [(set (match_operand:VDQIW 0 "s_register_operand" "=w") (unspec:VDQIW [(match_operand:VDQIW 1 "s_register_operand" "w") (match_operand:VDQIW 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VHADD))] "TARGET_NEON" "v%O3hadd.%T3%#\t%0, %1, %2" [(set_attr "neon_type" "neon_int_4")] ) (define_insn "neon_vqadd" [(set (match_operand:VDQIX 0 "s_register_operand" "=w") (unspec:VDQIX [(match_operand:VDQIX 1 "s_register_operand" "w") (match_operand:VDQIX 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VQADD))] "TARGET_NEON" "vqadd.%T3%#\t%0, %1, %2" [(set_attr "neon_type" "neon_int_4")] ) (define_insn "neon_vaddhn" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VN 1 "s_register_operand" "w") (match_operand:VN 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VADDHN))] "TARGET_NEON" "v%O3addhn.\t%P0, %q1, %q2" [(set_attr "neon_type" "neon_int_4")] ) ;; We cannot replace this unspec with mul3 because of the odd ;; polynomial multiplication case that can specified by operand 3. (define_insn "neon_vmul" [(set (match_operand:VDQW 0 "s_register_operand" "=w") (unspec:VDQW [(match_operand:VDQW 1 "s_register_operand" "w") (match_operand:VDQW 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VMUL))] "TARGET_NEON" "vmul.%F3%#\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mul_ddd_8_16_qdd_16_8_long_32_16_long") (const_string "neon_mul_qqq_8_16_32_ddd_32")) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mul_qqq_8_16_32_ddd_32") (const_string "neon_mul_qqq_8_16_32_ddd_32")))))] ) (define_expand "neon_vmla" [(match_operand:VDQW 0 "s_register_operand" "=w") (match_operand:VDQW 1 "s_register_operand" "0") (match_operand:VDQW 2 "s_register_operand" "w") (match_operand:VDQW 3 "s_register_operand" "w") (match_operand:SI 4 "immediate_operand" "i")] "TARGET_NEON" { if (! || flag_unsafe_math_optimizations) emit_insn (gen_mul3add_neon (operands[0], operands[1], operands[2], operands[3])); else emit_insn (gen_neon_vmla_unspec (operands[0], operands[1], operands[2], operands[3])); DONE; }) ; Used for intrinsics when flag_unsafe_math_optimizations is false. (define_insn "neon_vmla_unspec" [(set (match_operand:VDQ 0 "s_register_operand" "=w") (unspec:VDQ [(match_operand:VDQ 1 "s_register_operand" "0") (match_operand:VDQ 2 "s_register_operand" "w") (match_operand:VDQ 3 "s_register_operand" "w")] UNSPEC_VMLA))] "TARGET_NEON" "vmla.\t%0, %2, %3" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vmla_ddd") (const_string "neon_fp_vmla_qqq")) (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_8_16_qdd_16_8_long_32_16_long") (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long")) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_qqq_8_16") (const_string "neon_mla_qqq_32_qqd_32_scalar")))))] ) (define_insn "neon_vmlal" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand: 1 "s_register_operand" "0") (match_operand:VW 2 "s_register_operand" "w") (match_operand:VW 3 "s_register_operand" "w") (match_operand:SI 4 "immediate_operand" "i")] UNSPEC_VMLAL))] "TARGET_NEON" "vmlal.%T4%#\t%q0, %P2, %P3" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_8_16_qdd_16_8_long_32_16_long") (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long")))] ) (define_expand "neon_vmls" [(match_operand:VDQW 0 "s_register_operand" "=w") (match_operand:VDQW 1 "s_register_operand" "0") (match_operand:VDQW 2 "s_register_operand" "w") (match_operand:VDQW 3 "s_register_operand" "w") (match_operand:SI 4 "immediate_operand" "i")] "TARGET_NEON" { if (! || flag_unsafe_math_optimizations) emit_insn (gen_mul3negadd_neon (operands[0], operands[1], operands[2], operands[3])); else emit_insn (gen_neon_vmls_unspec (operands[0], operands[1], operands[2], operands[3])); DONE; }) ; Used for intrinsics when flag_unsafe_math_optimizations is false. (define_insn "neon_vmls_unspec" [(set (match_operand:VDQ 0 "s_register_operand" "=w") (unspec:VDQ [(match_operand:VDQ 1 "s_register_operand" "0") (match_operand:VDQ 2 "s_register_operand" "w") (match_operand:VDQ 3 "s_register_operand" "w")] UNSPEC_VMLS))] "TARGET_NEON" "vmls.\t%0, %2, %3" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vmla_ddd") (const_string "neon_fp_vmla_qqq")) (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_8_16_qdd_16_8_long_32_16_long") (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long")) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_qqq_8_16") (const_string "neon_mla_qqq_32_qqd_32_scalar")))))] ) (define_insn "neon_vmlsl" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand: 1 "s_register_operand" "0") (match_operand:VW 2 "s_register_operand" "w") (match_operand:VW 3 "s_register_operand" "w") (match_operand:SI 4 "immediate_operand" "i")] UNSPEC_VMLSL))] "TARGET_NEON" "vmlsl.%T4%#\t%q0, %P2, %P3" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_8_16_qdd_16_8_long_32_16_long") (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long")))] ) (define_insn "neon_vqdmulh" [(set (match_operand:VMDQI 0 "s_register_operand" "=w") (unspec:VMDQI [(match_operand:VMDQI 1 "s_register_operand" "w") (match_operand:VMDQI 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VQDMULH))] "TARGET_NEON" "vq%O3dmulh.\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mul_ddd_8_16_qdd_16_8_long_32_16_long") (const_string "neon_mul_qqq_8_16_32_ddd_32")) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mul_qqq_8_16_32_ddd_32") (const_string "neon_mul_qqq_8_16_32_ddd_32"))))] ) (define_insn "neon_vqdmlal" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand: 1 "s_register_operand" "0") (match_operand:VMDI 2 "s_register_operand" "w") (match_operand:VMDI 3 "s_register_operand" "w") (match_operand:SI 4 "immediate_operand" "i")] UNSPEC_VQDMLAL))] "TARGET_NEON" "vqdmlal.\t%q0, %P2, %P3" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_8_16_qdd_16_8_long_32_16_long") (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long")))] ) (define_insn "neon_vqdmlsl" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand: 1 "s_register_operand" "0") (match_operand:VMDI 2 "s_register_operand" "w") (match_operand:VMDI 3 "s_register_operand" "w") (match_operand:SI 4 "immediate_operand" "i")] UNSPEC_VQDMLSL))] "TARGET_NEON" "vqdmlsl.\t%q0, %P2, %P3" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_8_16_qdd_16_8_long_32_16_long") (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long")))] ) (define_insn "neon_vmull" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VW 1 "s_register_operand" "w") (match_operand:VW 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VMULL))] "TARGET_NEON" "vmull.%T3%#\t%q0, %P1, %P2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mul_ddd_8_16_qdd_16_8_long_32_16_long") (const_string "neon_mul_qdd_64_32_long_qqd_16_ddd_32_scalar_64_32_long_scalar")))] ) (define_insn "neon_vqdmull" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VMDI 1 "s_register_operand" "w") (match_operand:VMDI 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VQDMULL))] "TARGET_NEON" "vqdmull.\t%q0, %P1, %P2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mul_ddd_8_16_qdd_16_8_long_32_16_long") (const_string "neon_mul_qdd_64_32_long_qqd_16_ddd_32_scalar_64_32_long_scalar")))] ) (define_expand "neon_vsub" [(match_operand:VDQX 0 "s_register_operand" "=w") (match_operand:VDQX 1 "s_register_operand" "w") (match_operand:VDQX 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] "TARGET_NEON" { if (! || flag_unsafe_math_optimizations) emit_insn (gen_sub3 (operands[0], operands[1], operands[2])); else emit_insn (gen_neon_vsub_unspec (operands[0], operands[1], operands[2])); DONE; }) ; Used for intrinsics when flag_unsafe_math_optimizations is false. (define_insn "neon_vsub_unspec" [(set (match_operand:VDQX 0 "s_register_operand" "=w") (unspec:VDQX [(match_operand:VDQX 1 "s_register_operand" "w") (match_operand:VDQX 2 "s_register_operand" "w")] UNSPEC_VSUB))] "TARGET_NEON" "vsub.\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (const_string "neon_int_2")))] ) (define_insn "neon_vsubl" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VDI 1 "s_register_operand" "w") (match_operand:VDI 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VSUBL))] "TARGET_NEON" "vsubl.%T3%#\t%q0, %P1, %P2" [(set_attr "neon_type" "neon_int_2")] ) (define_insn "neon_vsubw" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand: 1 "s_register_operand" "w") (match_operand:VDI 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VSUBW))] "TARGET_NEON" "vsubw.%T3%#\t%q0, %q1, %P2" [(set_attr "neon_type" "neon_int_2")] ) (define_insn "neon_vqsub" [(set (match_operand:VDQIX 0 "s_register_operand" "=w") (unspec:VDQIX [(match_operand:VDQIX 1 "s_register_operand" "w") (match_operand:VDQIX 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VQSUB))] "TARGET_NEON" "vqsub.%T3%#\t%0, %1, %2" [(set_attr "neon_type" "neon_int_5")] ) (define_insn "neon_vhsub" [(set (match_operand:VDQIW 0 "s_register_operand" "=w") (unspec:VDQIW [(match_operand:VDQIW 1 "s_register_operand" "w") (match_operand:VDQIW 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VHSUB))] "TARGET_NEON" "vhsub.%T3%#\t%0, %1, %2" [(set_attr "neon_type" "neon_int_5")] ) (define_insn "neon_vsubhn" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VN 1 "s_register_operand" "w") (match_operand:VN 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VSUBHN))] "TARGET_NEON" "v%O3subhn.\t%P0, %q1, %q2" [(set_attr "neon_type" "neon_int_4")] ) (define_insn "neon_vceq" [(set (match_operand: 0 "s_register_operand" "=w,w") (unspec: [(match_operand:VDQW 1 "s_register_operand" "w,w") (match_operand:VDQW 2 "nonmemory_operand" "w,Dz") (match_operand:SI 3 "immediate_operand" "i,i")] UNSPEC_VCEQ))] "TARGET_NEON" "@ vceq.\t%0, %1, %2 vceq.\t%0, %1, #0" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (const_string "neon_int_5")))] ) (define_insn "neon_vcge" [(set (match_operand: 0 "s_register_operand" "=w,w") (unspec: [(match_operand:VDQW 1 "s_register_operand" "w,w") (match_operand:VDQW 2 "nonmemory_operand" "w,Dz") (match_operand:SI 3 "immediate_operand" "i,i")] UNSPEC_VCGE))] "TARGET_NEON" "@ vcge.%T3%#\t%0, %1, %2 vcge.%T3%#\t%0, %1, #0" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (const_string "neon_int_5")))] ) (define_insn "neon_vcgt" [(set (match_operand: 0 "s_register_operand" "=w,w") (unspec: [(match_operand:VDQW 1 "s_register_operand" "w,w") (match_operand:VDQW 2 "nonmemory_operand" "w,Dz") (match_operand:SI 3 "immediate_operand" "i,i")] UNSPEC_VCGT))] "TARGET_NEON" "@ vcgt.%T3%#\t%0, %1, %2 vcgt.%T3%#\t%0, %1, #0" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (const_string "neon_int_5")))] ) ;; VCLE and VCLT only support comparisons with immediate zero (register ;; variants are VCGE and VCGT with operands reversed). (define_insn "neon_vcle" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VDQW 1 "s_register_operand" "w") (match_operand:VDQW 2 "nonmemory_operand" "Dz") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VCLE))] "TARGET_NEON" "vcle.%T3%#\t%0, %1, #0" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (const_string "neon_int_5")))] ) (define_insn "neon_vclt" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VDQW 1 "s_register_operand" "w") (match_operand:VDQW 2 "nonmemory_operand" "Dz") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VCLT))] "TARGET_NEON" "vclt.%T3%#\t%0, %1, #0" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (const_string "neon_int_5")))] ) (define_insn "neon_vcage" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VCVTF 1 "s_register_operand" "w") (match_operand:VCVTF 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VCAGE))] "TARGET_NEON" "vacge.\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")))] ) (define_insn "neon_vcagt" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VCVTF 1 "s_register_operand" "w") (match_operand:VCVTF 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VCAGT))] "TARGET_NEON" "vacgt.\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")))] ) (define_insn "neon_vtst" [(set (match_operand:VDQIW 0 "s_register_operand" "=w") (unspec:VDQIW [(match_operand:VDQIW 1 "s_register_operand" "w") (match_operand:VDQIW 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VTST))] "TARGET_NEON" "vtst.\t%0, %1, %2" [(set_attr "neon_type" "neon_int_4")] ) (define_insn "neon_vabd" [(set (match_operand:VDQW 0 "s_register_operand" "=w") (unspec:VDQW [(match_operand:VDQW 1 "s_register_operand" "w") (match_operand:VDQW 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VABD))] "TARGET_NEON" "vabd.%T3%#\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (const_string "neon_int_5")))] ) (define_insn "neon_vabdl" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VW 1 "s_register_operand" "w") (match_operand:VW 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VABDL))] "TARGET_NEON" "vabdl.%T3%#\t%q0, %P1, %P2" [(set_attr "neon_type" "neon_int_5")] ) (define_insn "neon_vaba" [(set (match_operand:VDQIW 0 "s_register_operand" "=w") (plus:VDQIW (match_operand:VDQIW 1 "s_register_operand" "0") (unspec:VDQIW [(match_operand:VDQIW 2 "s_register_operand" "w") (match_operand:VDQIW 3 "s_register_operand" "w") (match_operand:SI 4 "immediate_operand" "i")] UNSPEC_VABD)))] "TARGET_NEON" "vaba.%T4%#\t%0, %2, %3" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_vaba") (const_string "neon_vaba_qqq")))] ) (define_insn "neon_vabal" [(set (match_operand: 0 "s_register_operand" "=w") (plus: (match_operand: 1 "s_register_operand" "0") (unspec: [(match_operand:VW 2 "s_register_operand" "w") (match_operand:VW 3 "s_register_operand" "w") (match_operand:SI 4 "immediate_operand" "i")] UNSPEC_VABDL)))] "TARGET_NEON" "vabal.%T4%#\t%q0, %P2, %P3" [(set_attr "neon_type" "neon_vaba")] ) (define_insn "neon_vmax" [(set (match_operand:VDQW 0 "s_register_operand" "=w") (unspec:VDQW [(match_operand:VDQW 1 "s_register_operand" "w") (match_operand:VDQW 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VMAX))] "TARGET_NEON" "vmax.%T3%#\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (const_string "neon_int_5")))] ) (define_insn "neon_vmin" [(set (match_operand:VDQW 0 "s_register_operand" "=w") (unspec:VDQW [(match_operand:VDQW 1 "s_register_operand" "w") (match_operand:VDQW 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VMIN))] "TARGET_NEON" "vmin.%T3%#\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")) (const_string "neon_int_5")))] ) (define_expand "neon_vpadd" [(match_operand:VD 0 "s_register_operand" "=w") (match_operand:VD 1 "s_register_operand" "w") (match_operand:VD 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] "TARGET_NEON" { emit_insn (gen_neon_vpadd_internal (operands[0], operands[1], operands[2])); DONE; }) (define_insn "neon_vpaddl" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VDQIW 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] UNSPEC_VPADDL))] "TARGET_NEON" "vpaddl.%T2%#\t%0, %1" ;; Assume this schedules like vaddl. [(set_attr "neon_type" "neon_int_3")] ) (define_insn "neon_vpadal" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand: 1 "s_register_operand" "0") (match_operand:VDQIW 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VPADAL))] "TARGET_NEON" "vpadal.%T3%#\t%0, %2" ;; Assume this schedules like vpadd. [(set_attr "neon_type" "neon_int_1")] ) (define_insn "neon_vpmax" [(set (match_operand:VD 0 "s_register_operand" "=w") (unspec:VD [(match_operand:VD 1 "s_register_operand" "w") (match_operand:VD 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VPMAX))] "TARGET_NEON" "vpmax.%T3%#\t%0, %1, %2" ;; Assume this schedules like vmax. [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_int_5")))] ) (define_insn "neon_vpmin" [(set (match_operand:VD 0 "s_register_operand" "=w") (unspec:VD [(match_operand:VD 1 "s_register_operand" "w") (match_operand:VD 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VPMIN))] "TARGET_NEON" "vpmin.%T3%#\t%0, %1, %2" ;; Assume this schedules like vmin. [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_int_5")))] ) (define_insn "neon_vrecps" [(set (match_operand:VCVTF 0 "s_register_operand" "=w") (unspec:VCVTF [(match_operand:VCVTF 1 "s_register_operand" "w") (match_operand:VCVTF 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VRECPS))] "TARGET_NEON" "vrecps.\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vrecps_vrsqrts_ddd") (const_string "neon_fp_vrecps_vrsqrts_qqq")))] ) (define_insn "neon_vrsqrts" [(set (match_operand:VCVTF 0 "s_register_operand" "=w") (unspec:VCVTF [(match_operand:VCVTF 1 "s_register_operand" "w") (match_operand:VCVTF 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VRSQRTS))] "TARGET_NEON" "vrsqrts.\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vrecps_vrsqrts_ddd") (const_string "neon_fp_vrecps_vrsqrts_qqq")))] ) (define_expand "neon_vabs" [(match_operand:VDQW 0 "s_register_operand" "") (match_operand:VDQW 1 "s_register_operand" "") (match_operand:SI 2 "immediate_operand" "")] "TARGET_NEON" { emit_insn (gen_abs2 (operands[0], operands[1])); DONE; }) (define_insn "neon_vqabs" [(set (match_operand:VDQIW 0 "s_register_operand" "=w") (unspec:VDQIW [(match_operand:VDQIW 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] UNSPEC_VQABS))] "TARGET_NEON" "vqabs.\t%0, %1" [(set_attr "neon_type" "neon_vqneg_vqabs")] ) (define_expand "neon_vneg" [(match_operand:VDQW 0 "s_register_operand" "") (match_operand:VDQW 1 "s_register_operand" "") (match_operand:SI 2 "immediate_operand" "")] "TARGET_NEON" { emit_insn (gen_neg2 (operands[0], operands[1])); DONE; }) (define_insn "neon_vqneg" [(set (match_operand:VDQIW 0 "s_register_operand" "=w") (unspec:VDQIW [(match_operand:VDQIW 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] UNSPEC_VQNEG))] "TARGET_NEON" "vqneg.\t%0, %1" [(set_attr "neon_type" "neon_vqneg_vqabs")] ) (define_insn "neon_vcls" [(set (match_operand:VDQIW 0 "s_register_operand" "=w") (unspec:VDQIW [(match_operand:VDQIW 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] UNSPEC_VCLS))] "TARGET_NEON" "vcls.\t%0, %1" [(set_attr "neon_type" "neon_int_1")] ) (define_insn "clz2" [(set (match_operand:VDQIW 0 "s_register_operand" "=w") (clz:VDQIW (match_operand:VDQIW 1 "s_register_operand" "w")))] "TARGET_NEON" "vclz.\t%0, %1" [(set_attr "neon_type" "neon_int_1")] ) (define_expand "neon_vclz" [(match_operand:VDQIW 0 "s_register_operand" "") (match_operand:VDQIW 1 "s_register_operand" "") (match_operand:SI 2 "immediate_operand" "")] "TARGET_NEON" { emit_insn (gen_clz2 (operands[0], operands[1])); DONE; }) (define_insn "popcount2" [(set (match_operand:VE 0 "s_register_operand" "=w") (popcount:VE (match_operand:VE 1 "s_register_operand" "w")))] "TARGET_NEON" "vcnt.\t%0, %1" [(set_attr "neon_type" "neon_int_1")] ) (define_expand "neon_vcnt" [(match_operand:VE 0 "s_register_operand" "=w") (match_operand:VE 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] "TARGET_NEON" { emit_insn (gen_popcount2 (operands[0], operands[1])); DONE; }) (define_insn "neon_vrecpe" [(set (match_operand:V32 0 "s_register_operand" "=w") (unspec:V32 [(match_operand:V32 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] UNSPEC_VRECPE))] "TARGET_NEON" "vrecpe.\t%0, %1" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")))] ) (define_insn "neon_vrsqrte" [(set (match_operand:V32 0 "s_register_operand" "=w") (unspec:V32 [(match_operand:V32 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] UNSPEC_VRSQRTE))] "TARGET_NEON" "vrsqrte.\t%0, %1" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")))] ) (define_expand "neon_vmvn" [(match_operand:VDQIW 0 "s_register_operand" "") (match_operand:VDQIW 1 "s_register_operand" "") (match_operand:SI 2 "immediate_operand" "")] "TARGET_NEON" { emit_insn (gen_one_cmpl2 (operands[0], operands[1])); DONE; }) (define_insn "neon_vget_lane_sext_internal" [(set (match_operand:SI 0 "s_register_operand" "=r") (sign_extend:SI (vec_select: (match_operand:VD 1 "s_register_operand" "w") (parallel [(match_operand:SI 2 "immediate_operand" "i")]))))] "TARGET_NEON" { if (BYTES_BIG_ENDIAN) { int elt = INTVAL (operands[2]); elt = GET_MODE_NUNITS (mode) - 1 - elt; operands[2] = GEN_INT (elt); } return "vmov%?.s\t%0, %P1[%c2]"; } [(set_attr "predicable" "yes") (set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vget_lane_zext_internal" [(set (match_operand:SI 0 "s_register_operand" "=r") (zero_extend:SI (vec_select: (match_operand:VD 1 "s_register_operand" "w") (parallel [(match_operand:SI 2 "immediate_operand" "i")]))))] "TARGET_NEON" { if (BYTES_BIG_ENDIAN) { int elt = INTVAL (operands[2]); elt = GET_MODE_NUNITS (mode) - 1 - elt; operands[2] = GEN_INT (elt); } return "vmov%?.u\t%0, %P1[%c2]"; } [(set_attr "predicable" "yes") (set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vget_lane_sext_internal" [(set (match_operand:SI 0 "s_register_operand" "=r") (sign_extend:SI (vec_select: (match_operand:VQ 1 "s_register_operand" "w") (parallel [(match_operand:SI 2 "immediate_operand" "i")]))))] "TARGET_NEON" { rtx ops[3]; int regno = REGNO (operands[1]); unsigned int halfelts = GET_MODE_NUNITS (mode) / 2; unsigned int elt = INTVAL (operands[2]); unsigned int elt_adj = elt % halfelts; if (BYTES_BIG_ENDIAN) elt_adj = halfelts - 1 - elt_adj; ops[0] = operands[0]; ops[1] = gen_rtx_REG (mode, regno + 2 * (elt / halfelts)); ops[2] = GEN_INT (elt_adj); output_asm_insn ("vmov%?.s\t%0, %P1[%c2]", ops); return ""; } [(set_attr "predicable" "yes") (set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vget_lane_zext_internal" [(set (match_operand:SI 0 "s_register_operand" "=r") (zero_extend:SI (vec_select: (match_operand:VQ 1 "s_register_operand" "w") (parallel [(match_operand:SI 2 "immediate_operand" "i")]))))] "TARGET_NEON" { rtx ops[3]; int regno = REGNO (operands[1]); unsigned int halfelts = GET_MODE_NUNITS (mode) / 2; unsigned int elt = INTVAL (operands[2]); unsigned int elt_adj = elt % halfelts; if (BYTES_BIG_ENDIAN) elt_adj = halfelts - 1 - elt_adj; ops[0] = operands[0]; ops[1] = gen_rtx_REG (mode, regno + 2 * (elt / halfelts)); ops[2] = GEN_INT (elt_adj); output_asm_insn ("vmov%?.u\t%0, %P1[%c2]", ops); return ""; } [(set_attr "predicable" "yes") (set_attr "neon_type" "neon_bp_simple")] ) (define_expand "neon_vget_lane" [(match_operand: 0 "s_register_operand" "") (match_operand:VDQW 1 "s_register_operand" "") (match_operand:SI 2 "immediate_operand" "") (match_operand:SI 3 "immediate_operand" "")] "TARGET_NEON" { HOST_WIDE_INT magic = INTVAL (operands[3]); rtx insn; neon_lane_bounds (operands[2], 0, GET_MODE_NUNITS (mode)); if (BYTES_BIG_ENDIAN) { /* The intrinsics are defined in terms of a model where the element ordering in memory is vldm order, whereas the generic RTL is defined in terms of a model where the element ordering in memory is array order. Convert the lane number to conform to this model. */ unsigned int elt = INTVAL (operands[2]); unsigned int reg_nelts = 64 / GET_MODE_BITSIZE (GET_MODE_INNER (mode)); elt ^= reg_nelts - 1; operands[2] = GEN_INT (elt); } if ((magic & 3) == 3 || GET_MODE_BITSIZE (GET_MODE_INNER (mode)) == 32) insn = gen_vec_extract (operands[0], operands[1], operands[2]); else { if ((magic & 1) != 0) insn = gen_neon_vget_lane_sext_internal (operands[0], operands[1], operands[2]); else insn = gen_neon_vget_lane_zext_internal (operands[0], operands[1], operands[2]); } emit_insn (insn); DONE; }) ; Operand 3 (info word) is ignored because it does nothing useful with 64-bit ; elements. (define_expand "neon_vget_lanedi" [(match_operand:DI 0 "s_register_operand" "=r") (match_operand:DI 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (match_operand:SI 3 "immediate_operand" "i")] "TARGET_NEON" { neon_lane_bounds (operands[2], 0, 1); emit_move_insn (operands[0], operands[1]); DONE; }) (define_expand "neon_vget_lanev2di" [(match_operand:DI 0 "s_register_operand" "=r") (match_operand:V2DI 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (match_operand:SI 3 "immediate_operand" "i")] "TARGET_NEON" { neon_lane_bounds (operands[2], 0, 2); emit_insn (gen_vec_extractv2di (operands[0], operands[1], operands[2])); DONE; }) (define_expand "neon_vset_lane" [(match_operand:VDQ 0 "s_register_operand" "=w") (match_operand: 1 "s_register_operand" "r") (match_operand:VDQ 2 "s_register_operand" "0") (match_operand:SI 3 "immediate_operand" "i")] "TARGET_NEON" { unsigned int elt = INTVAL (operands[3]); neon_lane_bounds (operands[3], 0, GET_MODE_NUNITS (mode)); if (BYTES_BIG_ENDIAN) { unsigned int reg_nelts = 64 / GET_MODE_BITSIZE (GET_MODE_INNER (mode)); elt ^= reg_nelts - 1; } emit_insn (gen_vec_set_internal (operands[0], operands[1], GEN_INT (1 << elt), operands[2])); DONE; }) ; See neon_vget_lanedi comment for reasons operands 2 & 3 are ignored. (define_expand "neon_vset_lanedi" [(match_operand:DI 0 "s_register_operand" "=w") (match_operand:DI 1 "s_register_operand" "r") (match_operand:DI 2 "s_register_operand" "0") (match_operand:SI 3 "immediate_operand" "i")] "TARGET_NEON" { neon_lane_bounds (operands[3], 0, 1); emit_move_insn (operands[0], operands[1]); DONE; }) (define_expand "neon_vcreate" [(match_operand:VDX 0 "s_register_operand" "") (match_operand:DI 1 "general_operand" "")] "TARGET_NEON" { rtx src = gen_lowpart (mode, operands[1]); emit_move_insn (operands[0], src); DONE; }) (define_insn "neon_vdup_n" [(set (match_operand:VX 0 "s_register_operand" "=w") (vec_duplicate:VX (match_operand: 1 "s_register_operand" "r")))] "TARGET_NEON" "vdup%?.\t%0, %1" ;; Assume this schedules like vmov. [(set_attr "predicable" "yes") (set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vdup_n" [(set (match_operand:V32 0 "s_register_operand" "=w,w") (vec_duplicate:V32 (match_operand: 1 "s_register_operand" "r,t")))] "TARGET_NEON" "@ vdup%?.\t%0, %1 vdup%?.\t%0, %y1" ;; Assume this schedules like vmov. [(set_attr "predicable" "yes") (set_attr "neon_type" "neon_bp_simple")] ) (define_expand "neon_vdup_ndi" [(match_operand:DI 0 "s_register_operand" "=w") (match_operand:DI 1 "s_register_operand" "r")] "TARGET_NEON" { emit_move_insn (operands[0], operands[1]); DONE; } ) (define_insn "neon_vdup_nv2di" [(set (match_operand:V2DI 0 "s_register_operand" "=w,w") (vec_duplicate:V2DI (match_operand:DI 1 "s_register_operand" "r,w")))] "TARGET_NEON" "@ vmov%?\t%e0, %Q1, %R1\;vmov%?\t%f0, %Q1, %R1 vmov%?\t%e0, %P1\;vmov%?\t%f0, %P1" [(set_attr "predicable" "yes") (set_attr "length" "8") (set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vdup_lane_internal" [(set (match_operand:VDQW 0 "s_register_operand" "=w") (vec_duplicate:VDQW (vec_select: (match_operand: 1 "s_register_operand" "w") (parallel [(match_operand:SI 2 "immediate_operand" "i")]))))] "TARGET_NEON" { if (BYTES_BIG_ENDIAN) { int elt = INTVAL (operands[2]); elt = GET_MODE_NUNITS (mode) - 1 - elt; operands[2] = GEN_INT (elt); } if () return "vdup.\t%P0, %P1[%c2]"; else return "vdup.\t%q0, %P1[%c2]"; } ;; Assume this schedules like vmov. [(set_attr "neon_type" "neon_bp_simple")] ) (define_expand "neon_vdup_lane" [(match_operand:VDQW 0 "s_register_operand" "=w") (match_operand: 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] "TARGET_NEON" { neon_lane_bounds (operands[2], 0, GET_MODE_NUNITS (mode)); if (BYTES_BIG_ENDIAN) { unsigned int elt = INTVAL (operands[2]); unsigned int reg_nelts = 64 / GET_MODE_BITSIZE (GET_MODE_INNER (mode)); elt ^= reg_nelts - 1; operands[2] = GEN_INT (elt); } emit_insn (gen_neon_vdup_lane_internal (operands[0], operands[1], operands[2])); DONE; }) ; Scalar index is ignored, since only zero is valid here. (define_expand "neon_vdup_lanedi" [(match_operand:DI 0 "s_register_operand" "=w") (match_operand:DI 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] "TARGET_NEON" { neon_lane_bounds (operands[2], 0, 1); emit_move_insn (operands[0], operands[1]); DONE; }) ; Likewise for v2di, as the DImode second operand has only a single element. (define_expand "neon_vdup_lanev2di" [(match_operand:V2DI 0 "s_register_operand" "=w") (match_operand:DI 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] "TARGET_NEON" { neon_lane_bounds (operands[2], 0, 1); emit_insn (gen_neon_vdup_nv2di (operands[0], operands[1])); DONE; }) ;; In this insn, operand 1 should be low, and operand 2 the high part of the ;; dest vector. ;; FIXME: A different implementation of this builtin could make it much ;; more likely that we wouldn't actually need to output anything (we could make ;; it so that the reg allocator puts things in the right places magically ;; instead). Lack of subregs for vectors makes that tricky though, I think. (define_insn "neon_vcombine" [(set (match_operand: 0 "s_register_operand" "=w") (vec_concat: (match_operand:VDX 1 "s_register_operand" "w") (match_operand:VDX 2 "s_register_operand" "w")))] "TARGET_NEON" { int dest = REGNO (operands[0]); int src1 = REGNO (operands[1]); int src2 = REGNO (operands[2]); rtx destlo; if (src1 == dest && src2 == dest + 2) return ""; else if (src2 == dest && src1 == dest + 2) /* Special case of reversed high/low parts. */ return "vswp\t%P1, %P2"; destlo = gen_rtx_REG (mode, dest); if (!reg_overlap_mentioned_p (operands[2], destlo)) { /* Try to avoid unnecessary moves if part of the result is in the right place already. */ if (src1 != dest) output_asm_insn ("vmov\t%e0, %P1", operands); if (src2 != dest + 2) output_asm_insn ("vmov\t%f0, %P2", operands); } else { if (src2 != dest + 2) output_asm_insn ("vmov\t%f0, %P2", operands); if (src1 != dest) output_asm_insn ("vmov\t%e0, %P1", operands); } return ""; } ;; We set the neon_type attribute based on the vmov instructions above. [(set_attr "length" "8") (set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vget_highv16qi" [(set (match_operand:V8QI 0 "s_register_operand" "=w") (vec_select:V8QI (match_operand:V16QI 1 "s_register_operand" "w") (parallel [(const_int 8) (const_int 9) (const_int 10) (const_int 11) (const_int 12) (const_int 13) (const_int 14) (const_int 15)])))] "TARGET_NEON" { int dest = REGNO (operands[0]); int src = REGNO (operands[1]); if (dest != src + 2) return "vmov\t%P0, %f1"; else return ""; } [(set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vget_highv8hi" [(set (match_operand:V4HI 0 "s_register_operand" "=w") (vec_select:V4HI (match_operand:V8HI 1 "s_register_operand" "w") (parallel [(const_int 4) (const_int 5) (const_int 6) (const_int 7)])))] "TARGET_NEON" { int dest = REGNO (operands[0]); int src = REGNO (operands[1]); if (dest != src + 2) return "vmov\t%P0, %f1"; else return ""; } [(set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vget_highv4si" [(set (match_operand:V2SI 0 "s_register_operand" "=w") (vec_select:V2SI (match_operand:V4SI 1 "s_register_operand" "w") (parallel [(const_int 2) (const_int 3)])))] "TARGET_NEON" { int dest = REGNO (operands[0]); int src = REGNO (operands[1]); if (dest != src + 2) return "vmov\t%P0, %f1"; else return ""; } [(set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vget_highv4sf" [(set (match_operand:V2SF 0 "s_register_operand" "=w") (vec_select:V2SF (match_operand:V4SF 1 "s_register_operand" "w") (parallel [(const_int 2) (const_int 3)])))] "TARGET_NEON" { int dest = REGNO (operands[0]); int src = REGNO (operands[1]); if (dest != src + 2) return "vmov\t%P0, %f1"; else return ""; } [(set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vget_highv2di" [(set (match_operand:DI 0 "s_register_operand" "=w") (vec_select:DI (match_operand:V2DI 1 "s_register_operand" "w") (parallel [(const_int 1)])))] "TARGET_NEON" { int dest = REGNO (operands[0]); int src = REGNO (operands[1]); if (dest != src + 2) return "vmov\t%P0, %f1"; else return ""; } [(set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vget_lowv16qi" [(set (match_operand:V8QI 0 "s_register_operand" "=w") (vec_select:V8QI (match_operand:V16QI 1 "s_register_operand" "w") (parallel [(const_int 0) (const_int 1) (const_int 2) (const_int 3) (const_int 4) (const_int 5) (const_int 6) (const_int 7)])))] "TARGET_NEON" { int dest = REGNO (operands[0]); int src = REGNO (operands[1]); if (dest != src) return "vmov\t%P0, %e1"; else return ""; } [(set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vget_lowv8hi" [(set (match_operand:V4HI 0 "s_register_operand" "=w") (vec_select:V4HI (match_operand:V8HI 1 "s_register_operand" "w") (parallel [(const_int 0) (const_int 1) (const_int 2) (const_int 3)])))] "TARGET_NEON" { int dest = REGNO (operands[0]); int src = REGNO (operands[1]); if (dest != src) return "vmov\t%P0, %e1"; else return ""; } [(set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vget_lowv4si" [(set (match_operand:V2SI 0 "s_register_operand" "=w") (vec_select:V2SI (match_operand:V4SI 1 "s_register_operand" "w") (parallel [(const_int 0) (const_int 1)])))] "TARGET_NEON" { int dest = REGNO (operands[0]); int src = REGNO (operands[1]); if (dest != src) return "vmov\t%P0, %e1"; else return ""; } [(set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vget_lowv4sf" [(set (match_operand:V2SF 0 "s_register_operand" "=w") (vec_select:V2SF (match_operand:V4SF 1 "s_register_operand" "w") (parallel [(const_int 0) (const_int 1)])))] "TARGET_NEON" { int dest = REGNO (operands[0]); int src = REGNO (operands[1]); if (dest != src) return "vmov\t%P0, %e1"; else return ""; } [(set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vget_lowv2di" [(set (match_operand:DI 0 "s_register_operand" "=w") (vec_select:DI (match_operand:V2DI 1 "s_register_operand" "w") (parallel [(const_int 0)])))] "TARGET_NEON" { int dest = REGNO (operands[0]); int src = REGNO (operands[1]); if (dest != src) return "vmov\t%P0, %e1"; else return ""; } [(set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vcvt" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VCVTF 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] UNSPEC_VCVT))] "TARGET_NEON" "vcvt.%T2%#32.f32\t%0, %1" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")))] ) (define_insn "neon_vcvt" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VCVTI 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] UNSPEC_VCVT))] "TARGET_NEON" "vcvt.f32.%T2%#32\t%0, %1" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")))] ) (define_insn "neon_vcvt_n" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VCVTF 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VCVT_N))] "TARGET_NEON" { neon_const_bounds (operands[2], 1, 33); return "vcvt.%T3%#32.f32\t%0, %1, %2"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")))] ) (define_insn "neon_vcvt_n" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VCVTI 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VCVT_N))] "TARGET_NEON" { neon_const_bounds (operands[2], 1, 33); return "vcvt.f32.%T3%#32\t%0, %1, %2"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vadd_ddd_vabs_dd") (const_string "neon_fp_vadd_qqq_vabs_qq")))] ) (define_insn "neon_vmovn" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VN 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] UNSPEC_VMOVN))] "TARGET_NEON" "vmovn.\t%P0, %q1" [(set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vqmovn" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VN 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] UNSPEC_VQMOVN))] "TARGET_NEON" "vqmovn.%T2%#\t%P0, %q1" [(set_attr "neon_type" "neon_shift_2")] ) (define_insn "neon_vqmovun" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VN 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] UNSPEC_VQMOVUN))] "TARGET_NEON" "vqmovun.\t%P0, %q1" [(set_attr "neon_type" "neon_shift_2")] ) (define_insn "neon_vmovl" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VW 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] UNSPEC_VMOVL))] "TARGET_NEON" "vmovl.%T2%#\t%q0, %P1" [(set_attr "neon_type" "neon_shift_1")] ) (define_insn "neon_vmul_lane" [(set (match_operand:VMD 0 "s_register_operand" "=w") (unspec:VMD [(match_operand:VMD 1 "s_register_operand" "w") (match_operand:VMD 2 "s_register_operand" "") (match_operand:SI 3 "immediate_operand" "i") (match_operand:SI 4 "immediate_operand" "i")] UNSPEC_VMUL_LANE))] "TARGET_NEON" { neon_lane_bounds (operands[3], 0, GET_MODE_NUNITS (mode)); return "vmul.\t%P0, %P1, %P2[%c3]"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vmul_ddd") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mul_ddd_16_scalar_32_16_long_scalar") (const_string "neon_mul_qdd_64_32_long_qqd_16_ddd_32_scalar_64_32_long_scalar"))))] ) (define_insn "neon_vmul_lane" [(set (match_operand:VMQ 0 "s_register_operand" "=w") (unspec:VMQ [(match_operand:VMQ 1 "s_register_operand" "w") (match_operand: 2 "s_register_operand" "") (match_operand:SI 3 "immediate_operand" "i") (match_operand:SI 4 "immediate_operand" "i")] UNSPEC_VMUL_LANE))] "TARGET_NEON" { neon_lane_bounds (operands[3], 0, GET_MODE_NUNITS (mode)); return "vmul.\t%q0, %q1, %P2[%c3]"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vmul_qqd") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mul_qdd_64_32_long_qqd_16_ddd_32_scalar_64_32_long_scalar") (const_string "neon_mul_qqd_32_scalar"))))] ) (define_insn "neon_vmull_lane" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VMDI 1 "s_register_operand" "w") (match_operand:VMDI 2 "s_register_operand" "") (match_operand:SI 3 "immediate_operand" "i") (match_operand:SI 4 "immediate_operand" "i")] UNSPEC_VMULL_LANE))] "TARGET_NEON" { neon_lane_bounds (operands[3], 0, GET_MODE_NUNITS (mode)); return "vmull.%T4%#\t%q0, %P1, %P2[%c3]"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mul_ddd_16_scalar_32_16_long_scalar") (const_string "neon_mul_qdd_64_32_long_qqd_16_ddd_32_scalar_64_32_long_scalar")))] ) (define_insn "neon_vqdmull_lane" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VMDI 1 "s_register_operand" "w") (match_operand:VMDI 2 "s_register_operand" "") (match_operand:SI 3 "immediate_operand" "i") (match_operand:SI 4 "immediate_operand" "i")] UNSPEC_VQDMULL_LANE))] "TARGET_NEON" { neon_lane_bounds (operands[3], 0, GET_MODE_NUNITS (mode)); return "vqdmull.\t%q0, %P1, %P2[%c3]"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mul_ddd_16_scalar_32_16_long_scalar") (const_string "neon_mul_qdd_64_32_long_qqd_16_ddd_32_scalar_64_32_long_scalar")))] ) (define_insn "neon_vqdmulh_lane" [(set (match_operand:VMQI 0 "s_register_operand" "=w") (unspec:VMQI [(match_operand:VMQI 1 "s_register_operand" "w") (match_operand: 2 "s_register_operand" "") (match_operand:SI 3 "immediate_operand" "i") (match_operand:SI 4 "immediate_operand" "i")] UNSPEC_VQDMULH_LANE))] "TARGET_NEON" { neon_lane_bounds (operands[3], 0, GET_MODE_NUNITS (mode)); return "vq%O4dmulh.%T4%#\t%q0, %q1, %P2[%c3]"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mul_qdd_64_32_long_qqd_16_ddd_32_scalar_64_32_long_scalar") (const_string "neon_mul_qqd_32_scalar")))] ) (define_insn "neon_vqdmulh_lane" [(set (match_operand:VMDI 0 "s_register_operand" "=w") (unspec:VMDI [(match_operand:VMDI 1 "s_register_operand" "w") (match_operand:VMDI 2 "s_register_operand" "") (match_operand:SI 3 "immediate_operand" "i") (match_operand:SI 4 "immediate_operand" "i")] UNSPEC_VQDMULH_LANE))] "TARGET_NEON" { neon_lane_bounds (operands[3], 0, GET_MODE_NUNITS (mode)); return "vq%O4dmulh.%T4%#\t%P0, %P1, %P2[%c3]"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mul_ddd_16_scalar_32_16_long_scalar") (const_string "neon_mul_qdd_64_32_long_qqd_16_ddd_32_scalar_64_32_long_scalar")))] ) (define_insn "neon_vmla_lane" [(set (match_operand:VMD 0 "s_register_operand" "=w") (unspec:VMD [(match_operand:VMD 1 "s_register_operand" "0") (match_operand:VMD 2 "s_register_operand" "w") (match_operand:VMD 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "i") (match_operand:SI 5 "immediate_operand" "i")] UNSPEC_VMLA_LANE))] "TARGET_NEON" { neon_lane_bounds (operands[4], 0, GET_MODE_NUNITS (mode)); return "vmla.\t%P0, %P2, %P3[%c4]"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vmla_ddd_scalar") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_16_scalar_qdd_32_16_long_scalar") (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long"))))] ) (define_insn "neon_vmla_lane" [(set (match_operand:VMQ 0 "s_register_operand" "=w") (unspec:VMQ [(match_operand:VMQ 1 "s_register_operand" "0") (match_operand:VMQ 2 "s_register_operand" "w") (match_operand: 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "i") (match_operand:SI 5 "immediate_operand" "i")] UNSPEC_VMLA_LANE))] "TARGET_NEON" { neon_lane_bounds (operands[4], 0, GET_MODE_NUNITS (mode)); return "vmla.\t%q0, %q2, %P3[%c4]"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vmla_qqq_scalar") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long") (const_string "neon_mla_qqq_32_qqd_32_scalar"))))] ) (define_insn "neon_vmlal_lane" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand: 1 "s_register_operand" "0") (match_operand:VMDI 2 "s_register_operand" "w") (match_operand:VMDI 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "i") (match_operand:SI 5 "immediate_operand" "i")] UNSPEC_VMLAL_LANE))] "TARGET_NEON" { neon_lane_bounds (operands[4], 0, GET_MODE_NUNITS (mode)); return "vmlal.%T5%#\t%q0, %P2, %P3[%c4]"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_16_scalar_qdd_32_16_long_scalar") (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long")))] ) (define_insn "neon_vqdmlal_lane" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand: 1 "s_register_operand" "0") (match_operand:VMDI 2 "s_register_operand" "w") (match_operand:VMDI 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "i") (match_operand:SI 5 "immediate_operand" "i")] UNSPEC_VQDMLAL_LANE))] "TARGET_NEON" { neon_lane_bounds (operands[4], 0, GET_MODE_NUNITS (mode)); return "vqdmlal.\t%q0, %P2, %P3[%c4]"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_16_scalar_qdd_32_16_long_scalar") (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long")))] ) (define_insn "neon_vmls_lane" [(set (match_operand:VMD 0 "s_register_operand" "=w") (unspec:VMD [(match_operand:VMD 1 "s_register_operand" "0") (match_operand:VMD 2 "s_register_operand" "w") (match_operand:VMD 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "i") (match_operand:SI 5 "immediate_operand" "i")] UNSPEC_VMLS_LANE))] "TARGET_NEON" { neon_lane_bounds (operands[4], 0, GET_MODE_NUNITS (mode)); return "vmls.\t%P0, %P2, %P3[%c4]"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vmla_ddd_scalar") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_16_scalar_qdd_32_16_long_scalar") (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long"))))] ) (define_insn "neon_vmls_lane" [(set (match_operand:VMQ 0 "s_register_operand" "=w") (unspec:VMQ [(match_operand:VMQ 1 "s_register_operand" "0") (match_operand:VMQ 2 "s_register_operand" "w") (match_operand: 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "i") (match_operand:SI 5 "immediate_operand" "i")] UNSPEC_VMLS_LANE))] "TARGET_NEON" { neon_lane_bounds (operands[4], 0, GET_MODE_NUNITS (mode)); return "vmls.\t%q0, %q2, %P3[%c4]"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_fp_vmla_qqq_scalar") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long") (const_string "neon_mla_qqq_32_qqd_32_scalar"))))] ) (define_insn "neon_vmlsl_lane" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand: 1 "s_register_operand" "0") (match_operand:VMDI 2 "s_register_operand" "w") (match_operand:VMDI 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "i") (match_operand:SI 5 "immediate_operand" "i")] UNSPEC_VMLSL_LANE))] "TARGET_NEON" { neon_lane_bounds (operands[4], 0, GET_MODE_NUNITS (mode)); return "vmlsl.%T5%#\t%q0, %P2, %P3[%c4]"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_16_scalar_qdd_32_16_long_scalar") (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long")))] ) (define_insn "neon_vqdmlsl_lane" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand: 1 "s_register_operand" "0") (match_operand:VMDI 2 "s_register_operand" "w") (match_operand:VMDI 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "i") (match_operand:SI 5 "immediate_operand" "i")] UNSPEC_VQDMLSL_LANE))] "TARGET_NEON" { neon_lane_bounds (operands[4], 0, GET_MODE_NUNITS (mode)); return "vqdmlsl.\t%q0, %P2, %P3[%c4]"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_mla_ddd_16_scalar_qdd_32_16_long_scalar") (const_string "neon_mla_ddd_32_qqd_16_ddd_32_scalar_qdd_64_32_long_scalar_qdd_64_32_long")))] ) ; FIXME: For the "_n" multiply/multiply-accumulate insns, we copy a value in a ; core register into a temp register, then use a scalar taken from that. This ; isn't an optimal solution if e.g. the scalar has just been read from memory ; or extracted from another vector. The latter case it's currently better to ; use the "_lane" variant, and the former case can probably be implemented ; using vld1_lane, but that hasn't been done yet. (define_expand "neon_vmul_n" [(match_operand:VMD 0 "s_register_operand" "") (match_operand:VMD 1 "s_register_operand" "") (match_operand: 2 "s_register_operand" "") (match_operand:SI 3 "immediate_operand" "")] "TARGET_NEON" { rtx tmp = gen_reg_rtx (mode); emit_insn (gen_neon_vset_lane (tmp, operands[2], tmp, const0_rtx)); emit_insn (gen_neon_vmul_lane (operands[0], operands[1], tmp, const0_rtx, const0_rtx)); DONE; }) (define_expand "neon_vmul_n" [(match_operand:VMQ 0 "s_register_operand" "") (match_operand:VMQ 1 "s_register_operand" "") (match_operand: 2 "s_register_operand" "") (match_operand:SI 3 "immediate_operand" "")] "TARGET_NEON" { rtx tmp = gen_reg_rtx (mode); emit_insn (gen_neon_vset_lane (tmp, operands[2], tmp, const0_rtx)); emit_insn (gen_neon_vmul_lane (operands[0], operands[1], tmp, const0_rtx, const0_rtx)); DONE; }) (define_expand "neon_vmull_n" [(match_operand: 0 "s_register_operand" "") (match_operand:VMDI 1 "s_register_operand" "") (match_operand: 2 "s_register_operand" "") (match_operand:SI 3 "immediate_operand" "")] "TARGET_NEON" { rtx tmp = gen_reg_rtx (mode); emit_insn (gen_neon_vset_lane (tmp, operands[2], tmp, const0_rtx)); emit_insn (gen_neon_vmull_lane (operands[0], operands[1], tmp, const0_rtx, operands[3])); DONE; }) (define_expand "neon_vqdmull_n" [(match_operand: 0 "s_register_operand" "") (match_operand:VMDI 1 "s_register_operand" "") (match_operand: 2 "s_register_operand" "") (match_operand:SI 3 "immediate_operand" "")] "TARGET_NEON" { rtx tmp = gen_reg_rtx (mode); emit_insn (gen_neon_vset_lane (tmp, operands[2], tmp, const0_rtx)); emit_insn (gen_neon_vqdmull_lane (operands[0], operands[1], tmp, const0_rtx, const0_rtx)); DONE; }) (define_expand "neon_vqdmulh_n" [(match_operand:VMDI 0 "s_register_operand" "") (match_operand:VMDI 1 "s_register_operand" "") (match_operand: 2 "s_register_operand" "") (match_operand:SI 3 "immediate_operand" "")] "TARGET_NEON" { rtx tmp = gen_reg_rtx (mode); emit_insn (gen_neon_vset_lane (tmp, operands[2], tmp, const0_rtx)); emit_insn (gen_neon_vqdmulh_lane (operands[0], operands[1], tmp, const0_rtx, operands[3])); DONE; }) (define_expand "neon_vqdmulh_n" [(match_operand:VMQI 0 "s_register_operand" "") (match_operand:VMQI 1 "s_register_operand" "") (match_operand: 2 "s_register_operand" "") (match_operand:SI 3 "immediate_operand" "")] "TARGET_NEON" { rtx tmp = gen_reg_rtx (mode); emit_insn (gen_neon_vset_lane (tmp, operands[2], tmp, const0_rtx)); emit_insn (gen_neon_vqdmulh_lane (operands[0], operands[1], tmp, const0_rtx, operands[3])); DONE; }) (define_expand "neon_vmla_n" [(match_operand:VMD 0 "s_register_operand" "") (match_operand:VMD 1 "s_register_operand" "") (match_operand:VMD 2 "s_register_operand" "") (match_operand: 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "")] "TARGET_NEON" { rtx tmp = gen_reg_rtx (mode); emit_insn (gen_neon_vset_lane (tmp, operands[3], tmp, const0_rtx)); emit_insn (gen_neon_vmla_lane (operands[0], operands[1], operands[2], tmp, const0_rtx, operands[4])); DONE; }) (define_expand "neon_vmla_n" [(match_operand:VMQ 0 "s_register_operand" "") (match_operand:VMQ 1 "s_register_operand" "") (match_operand:VMQ 2 "s_register_operand" "") (match_operand: 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "")] "TARGET_NEON" { rtx tmp = gen_reg_rtx (mode); emit_insn (gen_neon_vset_lane (tmp, operands[3], tmp, const0_rtx)); emit_insn (gen_neon_vmla_lane (operands[0], operands[1], operands[2], tmp, const0_rtx, operands[4])); DONE; }) (define_expand "neon_vmlal_n" [(match_operand: 0 "s_register_operand" "") (match_operand: 1 "s_register_operand" "") (match_operand:VMDI 2 "s_register_operand" "") (match_operand: 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "")] "TARGET_NEON" { rtx tmp = gen_reg_rtx (mode); emit_insn (gen_neon_vset_lane (tmp, operands[3], tmp, const0_rtx)); emit_insn (gen_neon_vmlal_lane (operands[0], operands[1], operands[2], tmp, const0_rtx, operands[4])); DONE; }) (define_expand "neon_vqdmlal_n" [(match_operand: 0 "s_register_operand" "") (match_operand: 1 "s_register_operand" "") (match_operand:VMDI 2 "s_register_operand" "") (match_operand: 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "")] "TARGET_NEON" { rtx tmp = gen_reg_rtx (mode); emit_insn (gen_neon_vset_lane (tmp, operands[3], tmp, const0_rtx)); emit_insn (gen_neon_vqdmlal_lane (operands[0], operands[1], operands[2], tmp, const0_rtx, operands[4])); DONE; }) (define_expand "neon_vmls_n" [(match_operand:VMD 0 "s_register_operand" "") (match_operand:VMD 1 "s_register_operand" "") (match_operand:VMD 2 "s_register_operand" "") (match_operand: 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "")] "TARGET_NEON" { rtx tmp = gen_reg_rtx (mode); emit_insn (gen_neon_vset_lane (tmp, operands[3], tmp, const0_rtx)); emit_insn (gen_neon_vmls_lane (operands[0], operands[1], operands[2], tmp, const0_rtx, operands[4])); DONE; }) (define_expand "neon_vmls_n" [(match_operand:VMQ 0 "s_register_operand" "") (match_operand:VMQ 1 "s_register_operand" "") (match_operand:VMQ 2 "s_register_operand" "") (match_operand: 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "")] "TARGET_NEON" { rtx tmp = gen_reg_rtx (mode); emit_insn (gen_neon_vset_lane (tmp, operands[3], tmp, const0_rtx)); emit_insn (gen_neon_vmls_lane (operands[0], operands[1], operands[2], tmp, const0_rtx, operands[4])); DONE; }) (define_expand "neon_vmlsl_n" [(match_operand: 0 "s_register_operand" "") (match_operand: 1 "s_register_operand" "") (match_operand:VMDI 2 "s_register_operand" "") (match_operand: 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "")] "TARGET_NEON" { rtx tmp = gen_reg_rtx (mode); emit_insn (gen_neon_vset_lane (tmp, operands[3], tmp, const0_rtx)); emit_insn (gen_neon_vmlsl_lane (operands[0], operands[1], operands[2], tmp, const0_rtx, operands[4])); DONE; }) (define_expand "neon_vqdmlsl_n" [(match_operand: 0 "s_register_operand" "") (match_operand: 1 "s_register_operand" "") (match_operand:VMDI 2 "s_register_operand" "") (match_operand: 3 "s_register_operand" "") (match_operand:SI 4 "immediate_operand" "")] "TARGET_NEON" { rtx tmp = gen_reg_rtx (mode); emit_insn (gen_neon_vset_lane (tmp, operands[3], tmp, const0_rtx)); emit_insn (gen_neon_vqdmlsl_lane (operands[0], operands[1], operands[2], tmp, const0_rtx, operands[4])); DONE; }) (define_insn "neon_vext" [(set (match_operand:VDQX 0 "s_register_operand" "=w") (unspec:VDQX [(match_operand:VDQX 1 "s_register_operand" "w") (match_operand:VDQX 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VEXT))] "TARGET_NEON" { neon_const_bounds (operands[3], 0, GET_MODE_NUNITS (mode)); return "vext.\t%0, %1, %2, %3"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_bp_simple") (const_string "neon_bp_2cycle")))] ) (define_insn "neon_vrev64" [(set (match_operand:VDQ 0 "s_register_operand" "=w") (unspec:VDQ [(match_operand:VDQ 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] UNSPEC_VREV64))] "TARGET_NEON" "vrev64.\t%0, %1" [(set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vrev32" [(set (match_operand:VX 0 "s_register_operand" "=w") (unspec:VX [(match_operand:VX 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] UNSPEC_VREV32))] "TARGET_NEON" "vrev32.\t%0, %1" [(set_attr "neon_type" "neon_bp_simple")] ) (define_insn "neon_vrev16" [(set (match_operand:VE 0 "s_register_operand" "=w") (unspec:VE [(match_operand:VE 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i")] UNSPEC_VREV16))] "TARGET_NEON" "vrev16.\t%0, %1" [(set_attr "neon_type" "neon_bp_simple")] ) ; vbsl_* intrinsics may compile to any of vbsl/vbif/vbit depending on register ; allocation. For an intrinsic of form: ; rD = vbsl_* (rS, rN, rM) ; We can use any of: ; vbsl rS, rN, rM (if D = S) ; vbit rD, rN, rS (if D = M, so 1-bits in rS choose bits from rN, else rM) ; vbif rD, rM, rS (if D = N, so 0-bits in rS choose bits from rM, else rN) (define_insn "neon_vbsl_internal" [(set (match_operand:VDQX 0 "s_register_operand" "=w,w,w") (unspec:VDQX [(match_operand:VDQX 1 "s_register_operand" " 0,w,w") (match_operand:VDQX 2 "s_register_operand" " w,w,0") (match_operand:VDQX 3 "s_register_operand" " w,0,w")] UNSPEC_VBSL))] "TARGET_NEON" "@ vbsl\t%0, %2, %3 vbit\t%0, %2, %1 vbif\t%0, %3, %1" [(set_attr "neon_type" "neon_int_1")] ) (define_expand "neon_vbsl" [(set (match_operand:VDQX 0 "s_register_operand" "") (unspec:VDQX [(match_operand: 1 "s_register_operand" "") (match_operand:VDQX 2 "s_register_operand" "") (match_operand:VDQX 3 "s_register_operand" "")] UNSPEC_VBSL))] "TARGET_NEON" { /* We can't alias operands together if they have different modes. */ operands[1] = gen_lowpart (mode, operands[1]); }) (define_insn "neon_vshl" [(set (match_operand:VDQIX 0 "s_register_operand" "=w") (unspec:VDQIX [(match_operand:VDQIX 1 "s_register_operand" "w") (match_operand:VDQIX 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VSHL))] "TARGET_NEON" "v%O3shl.%T3%#\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_vshl_ddd") (const_string "neon_shift_3")))] ) (define_insn "neon_vqshl" [(set (match_operand:VDQIX 0 "s_register_operand" "=w") (unspec:VDQIX [(match_operand:VDQIX 1 "s_register_operand" "w") (match_operand:VDQIX 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VQSHL))] "TARGET_NEON" "vq%O3shl.%T3%#\t%0, %1, %2" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_shift_2") (const_string "neon_vqshl_vrshl_vqrshl_qqq")))] ) (define_insn "neon_vshr_n" [(set (match_operand:VDQIX 0 "s_register_operand" "=w") (unspec:VDQIX [(match_operand:VDQIX 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VSHR_N))] "TARGET_NEON" { neon_const_bounds (operands[2], 1, neon_element_bits (mode) + 1); return "v%O3shr.%T3%#\t%0, %1, %2"; } [(set_attr "neon_type" "neon_shift_1")] ) (define_insn "neon_vshrn_n" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VN 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VSHRN_N))] "TARGET_NEON" { neon_const_bounds (operands[2], 1, neon_element_bits (mode) / 2 + 1); return "v%O3shrn.\t%P0, %q1, %2"; } [(set_attr "neon_type" "neon_shift_1")] ) (define_insn "neon_vqshrn_n" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VN 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VQSHRN_N))] "TARGET_NEON" { neon_const_bounds (operands[2], 1, neon_element_bits (mode) / 2 + 1); return "vq%O3shrn.%T3%#\t%P0, %q1, %2"; } [(set_attr "neon_type" "neon_shift_2")] ) (define_insn "neon_vqshrun_n" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VN 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VQSHRUN_N))] "TARGET_NEON" { neon_const_bounds (operands[2], 1, neon_element_bits (mode) / 2 + 1); return "vq%O3shrun.%T3%#\t%P0, %q1, %2"; } [(set_attr "neon_type" "neon_shift_2")] ) (define_insn "neon_vshl_n" [(set (match_operand:VDQIX 0 "s_register_operand" "=w") (unspec:VDQIX [(match_operand:VDQIX 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VSHL_N))] "TARGET_NEON" { neon_const_bounds (operands[2], 0, neon_element_bits (mode)); return "vshl.\t%0, %1, %2"; } [(set_attr "neon_type" "neon_shift_1")] ) (define_insn "neon_vqshl_n" [(set (match_operand:VDQIX 0 "s_register_operand" "=w") (unspec:VDQIX [(match_operand:VDQIX 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VQSHL_N))] "TARGET_NEON" { neon_const_bounds (operands[2], 0, neon_element_bits (mode)); return "vqshl.%T3%#\t%0, %1, %2"; } [(set_attr "neon_type" "neon_shift_2")] ) (define_insn "neon_vqshlu_n" [(set (match_operand:VDQIX 0 "s_register_operand" "=w") (unspec:VDQIX [(match_operand:VDQIX 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VQSHLU_N))] "TARGET_NEON" { neon_const_bounds (operands[2], 0, neon_element_bits (mode)); return "vqshlu.%T3%#\t%0, %1, %2"; } [(set_attr "neon_type" "neon_shift_2")] ) (define_insn "neon_vshll_n" [(set (match_operand: 0 "s_register_operand" "=w") (unspec: [(match_operand:VW 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VSHLL_N))] "TARGET_NEON" { /* The boundaries are: 0 < imm <= size. */ neon_const_bounds (operands[2], 0, neon_element_bits (mode) + 1); return "vshll.%T3%#\t%q0, %P1, %2"; } [(set_attr "neon_type" "neon_shift_1")] ) (define_insn "neon_vsra_n" [(set (match_operand:VDQIX 0 "s_register_operand" "=w") (unspec:VDQIX [(match_operand:VDQIX 1 "s_register_operand" "0") (match_operand:VDQIX 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i") (match_operand:SI 4 "immediate_operand" "i")] UNSPEC_VSRA_N))] "TARGET_NEON" { neon_const_bounds (operands[3], 1, neon_element_bits (mode) + 1); return "v%O4sra.%T4%#\t%0, %2, %3"; } [(set_attr "neon_type" "neon_vsra_vrsra")] ) (define_insn "neon_vsri_n" [(set (match_operand:VDQIX 0 "s_register_operand" "=w") (unspec:VDQIX [(match_operand:VDQIX 1 "s_register_operand" "0") (match_operand:VDQIX 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VSRI))] "TARGET_NEON" { neon_const_bounds (operands[3], 1, neon_element_bits (mode) + 1); return "vsri.\t%0, %2, %3"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_shift_1") (const_string "neon_shift_3")))] ) (define_insn "neon_vsli_n" [(set (match_operand:VDQIX 0 "s_register_operand" "=w") (unspec:VDQIX [(match_operand:VDQIX 1 "s_register_operand" "0") (match_operand:VDQIX 2 "s_register_operand" "w") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VSLI))] "TARGET_NEON" { neon_const_bounds (operands[3], 0, neon_element_bits (mode)); return "vsli.\t%0, %2, %3"; } [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_shift_1") (const_string "neon_shift_3")))] ) (define_insn "neon_vtbl1v8qi" [(set (match_operand:V8QI 0 "s_register_operand" "=w") (unspec:V8QI [(match_operand:V8QI 1 "s_register_operand" "w") (match_operand:V8QI 2 "s_register_operand" "w")] UNSPEC_VTBL))] "TARGET_NEON" "vtbl.8\t%P0, {%P1}, %P2" [(set_attr "neon_type" "neon_bp_2cycle")] ) (define_insn "neon_vtbl2v8qi" [(set (match_operand:V8QI 0 "s_register_operand" "=w") (unspec:V8QI [(match_operand:TI 1 "s_register_operand" "w") (match_operand:V8QI 2 "s_register_operand" "w")] UNSPEC_VTBL))] "TARGET_NEON" { rtx ops[4]; int tabbase = REGNO (operands[1]); ops[0] = operands[0]; ops[1] = gen_rtx_REG (V8QImode, tabbase); ops[2] = gen_rtx_REG (V8QImode, tabbase + 2); ops[3] = operands[2]; output_asm_insn ("vtbl.8\t%P0, {%P1, %P2}, %P3", ops); return ""; } [(set_attr "neon_type" "neon_bp_2cycle")] ) (define_insn "neon_vtbl3v8qi" [(set (match_operand:V8QI 0 "s_register_operand" "=w") (unspec:V8QI [(match_operand:EI 1 "s_register_operand" "w") (match_operand:V8QI 2 "s_register_operand" "w")] UNSPEC_VTBL))] "TARGET_NEON" { rtx ops[5]; int tabbase = REGNO (operands[1]); ops[0] = operands[0]; ops[1] = gen_rtx_REG (V8QImode, tabbase); ops[2] = gen_rtx_REG (V8QImode, tabbase + 2); ops[3] = gen_rtx_REG (V8QImode, tabbase + 4); ops[4] = operands[2]; output_asm_insn ("vtbl.8\t%P0, {%P1, %P2, %P3}, %P4", ops); return ""; } [(set_attr "neon_type" "neon_bp_3cycle")] ) (define_insn "neon_vtbl4v8qi" [(set (match_operand:V8QI 0 "s_register_operand" "=w") (unspec:V8QI [(match_operand:OI 1 "s_register_operand" "w") (match_operand:V8QI 2 "s_register_operand" "w")] UNSPEC_VTBL))] "TARGET_NEON" { rtx ops[6]; int tabbase = REGNO (operands[1]); ops[0] = operands[0]; ops[1] = gen_rtx_REG (V8QImode, tabbase); ops[2] = gen_rtx_REG (V8QImode, tabbase + 2); ops[3] = gen_rtx_REG (V8QImode, tabbase + 4); ops[4] = gen_rtx_REG (V8QImode, tabbase + 6); ops[5] = operands[2]; output_asm_insn ("vtbl.8\t%P0, {%P1, %P2, %P3, %P4}, %P5", ops); return ""; } [(set_attr "neon_type" "neon_bp_3cycle")] ) (define_insn "neon_vtbx1v8qi" [(set (match_operand:V8QI 0 "s_register_operand" "=w") (unspec:V8QI [(match_operand:V8QI 1 "s_register_operand" "0") (match_operand:V8QI 2 "s_register_operand" "w") (match_operand:V8QI 3 "s_register_operand" "w")] UNSPEC_VTBX))] "TARGET_NEON" "vtbx.8\t%P0, {%P2}, %P3" [(set_attr "neon_type" "neon_bp_2cycle")] ) (define_insn "neon_vtbx2v8qi" [(set (match_operand:V8QI 0 "s_register_operand" "=w") (unspec:V8QI [(match_operand:V8QI 1 "s_register_operand" "0") (match_operand:TI 2 "s_register_operand" "w") (match_operand:V8QI 3 "s_register_operand" "w")] UNSPEC_VTBX))] "TARGET_NEON" { rtx ops[4]; int tabbase = REGNO (operands[2]); ops[0] = operands[0]; ops[1] = gen_rtx_REG (V8QImode, tabbase); ops[2] = gen_rtx_REG (V8QImode, tabbase + 2); ops[3] = operands[3]; output_asm_insn ("vtbx.8\t%P0, {%P1, %P2}, %P3", ops); return ""; } [(set_attr "neon_type" "neon_bp_2cycle")] ) (define_insn "neon_vtbx3v8qi" [(set (match_operand:V8QI 0 "s_register_operand" "=w") (unspec:V8QI [(match_operand:V8QI 1 "s_register_operand" "0") (match_operand:EI 2 "s_register_operand" "w") (match_operand:V8QI 3 "s_register_operand" "w")] UNSPEC_VTBX))] "TARGET_NEON" { rtx ops[5]; int tabbase = REGNO (operands[2]); ops[0] = operands[0]; ops[1] = gen_rtx_REG (V8QImode, tabbase); ops[2] = gen_rtx_REG (V8QImode, tabbase + 2); ops[3] = gen_rtx_REG (V8QImode, tabbase + 4); ops[4] = operands[3]; output_asm_insn ("vtbx.8\t%P0, {%P1, %P2, %P3}, %P4", ops); return ""; } [(set_attr "neon_type" "neon_bp_3cycle")] ) (define_insn "neon_vtbx4v8qi" [(set (match_operand:V8QI 0 "s_register_operand" "=w") (unspec:V8QI [(match_operand:V8QI 1 "s_register_operand" "0") (match_operand:OI 2 "s_register_operand" "w") (match_operand:V8QI 3 "s_register_operand" "w")] UNSPEC_VTBX))] "TARGET_NEON" { rtx ops[6]; int tabbase = REGNO (operands[2]); ops[0] = operands[0]; ops[1] = gen_rtx_REG (V8QImode, tabbase); ops[2] = gen_rtx_REG (V8QImode, tabbase + 2); ops[3] = gen_rtx_REG (V8QImode, tabbase + 4); ops[4] = gen_rtx_REG (V8QImode, tabbase + 6); ops[5] = operands[3]; output_asm_insn ("vtbx.8\t%P0, {%P1, %P2, %P3, %P4}, %P5", ops); return ""; } [(set_attr "neon_type" "neon_bp_3cycle")] ) (define_insn "neon_vtrn_internal" [(set (match_operand:VDQW 0 "s_register_operand" "=w") (unspec:VDQW [(match_operand:VDQW 1 "s_register_operand" "0") (match_operand:VDQW 2 "s_register_operand" "w")] UNSPEC_VTRN1)) (set (match_operand:VDQW 3 "s_register_operand" "=2") (unspec:VDQW [(match_dup 1) (match_dup 2)] UNSPEC_VTRN2))] "TARGET_NEON" "vtrn.\t%0, %3" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_bp_simple") (const_string "neon_bp_3cycle")))] ) (define_expand "neon_vtrn" [(match_operand:SI 0 "s_register_operand" "r") (match_operand:VDQW 1 "s_register_operand" "w") (match_operand:VDQW 2 "s_register_operand" "w")] "TARGET_NEON" { neon_emit_pair_result_insn (mode, gen_neon_vtrn_internal, operands[0], operands[1], operands[2]); DONE; }) (define_insn "neon_vzip_internal" [(set (match_operand:VDQW 0 "s_register_operand" "=w") (unspec:VDQW [(match_operand:VDQW 1 "s_register_operand" "0") (match_operand:VDQW 2 "s_register_operand" "w")] UNSPEC_VZIP1)) (set (match_operand:VDQW 3 "s_register_operand" "=2") (unspec:VDQW [(match_dup 1) (match_dup 2)] UNSPEC_VZIP2))] "TARGET_NEON" "vzip.\t%0, %3" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_bp_simple") (const_string "neon_bp_3cycle")))] ) (define_expand "neon_vzip" [(match_operand:SI 0 "s_register_operand" "r") (match_operand:VDQW 1 "s_register_operand" "w") (match_operand:VDQW 2 "s_register_operand" "w")] "TARGET_NEON" { neon_emit_pair_result_insn (mode, gen_neon_vzip_internal, operands[0], operands[1], operands[2]); DONE; }) (define_insn "neon_vuzp_internal" [(set (match_operand:VDQW 0 "s_register_operand" "=w") (unspec:VDQW [(match_operand:VDQW 1 "s_register_operand" "0") (match_operand:VDQW 2 "s_register_operand" "w")] UNSPEC_VUZP1)) (set (match_operand:VDQW 3 "s_register_operand" "=2") (unspec:VDQW [(match_dup 1) (match_dup 2)] UNSPEC_VUZP2))] "TARGET_NEON" "vuzp.\t%0, %3" [(set (attr "neon_type") (if_then_else (ne (symbol_ref "") (const_int 0)) (const_string "neon_bp_simple") (const_string "neon_bp_3cycle")))] ) (define_expand "neon_vuzp" [(match_operand:SI 0 "s_register_operand" "r") (match_operand:VDQW 1 "s_register_operand" "w") (match_operand:VDQW 2 "s_register_operand" "w")] "TARGET_NEON" { neon_emit_pair_result_insn (mode, gen_neon_vuzp_internal, operands[0], operands[1], operands[2]); DONE; }) (define_expand "neon_vreinterpretv8qi" [(match_operand:V8QI 0 "s_register_operand" "") (match_operand:VDX 1 "s_register_operand" "")] "TARGET_NEON" { neon_reinterpret (operands[0], operands[1]); DONE; }) (define_expand "neon_vreinterpretv4hi" [(match_operand:V4HI 0 "s_register_operand" "") (match_operand:VDX 1 "s_register_operand" "")] "TARGET_NEON" { neon_reinterpret (operands[0], operands[1]); DONE; }) (define_expand "neon_vreinterpretv2si" [(match_operand:V2SI 0 "s_register_operand" "") (match_operand:VDX 1 "s_register_operand" "")] "TARGET_NEON" { neon_reinterpret (operands[0], operands[1]); DONE; }) (define_expand "neon_vreinterpretv2sf" [(match_operand:V2SF 0 "s_register_operand" "") (match_operand:VDX 1 "s_register_operand" "")] "TARGET_NEON" { neon_reinterpret (operands[0], operands[1]); DONE; }) (define_expand "neon_vreinterpretdi" [(match_operand:DI 0 "s_register_operand" "") (match_operand:VDX 1 "s_register_operand" "")] "TARGET_NEON" { neon_reinterpret (operands[0], operands[1]); DONE; }) (define_expand "neon_vreinterpretv16qi" [(match_operand:V16QI 0 "s_register_operand" "") (match_operand:VQX 1 "s_register_operand" "")] "TARGET_NEON" { neon_reinterpret (operands[0], operands[1]); DONE; }) (define_expand "neon_vreinterpretv8hi" [(match_operand:V8HI 0 "s_register_operand" "") (match_operand:VQX 1 "s_register_operand" "")] "TARGET_NEON" { neon_reinterpret (operands[0], operands[1]); DONE; }) (define_expand "neon_vreinterpretv4si" [(match_operand:V4SI 0 "s_register_operand" "") (match_operand:VQX 1 "s_register_operand" "")] "TARGET_NEON" { neon_reinterpret (operands[0], operands[1]); DONE; }) (define_expand "neon_vreinterpretv4sf" [(match_operand:V4SF 0 "s_register_operand" "") (match_operand:VQX 1 "s_register_operand" "")] "TARGET_NEON" { neon_reinterpret (operands[0], operands[1]); DONE; }) (define_expand "neon_vreinterpretv2di" [(match_operand:V2DI 0 "s_register_operand" "") (match_operand:VQX 1 "s_register_operand" "")] "TARGET_NEON" { neon_reinterpret (operands[0], operands[1]); DONE; }) (define_insn "neon_vld1" [(set (match_operand:VDQX 0 "s_register_operand" "=w") (unspec:VDQX [(mem:VDQX (match_operand:SI 1 "s_register_operand" "r"))] UNSPEC_VLD1))] "TARGET_NEON" "vld1.\t%h0, [%1]" [(set_attr "neon_type" "neon_vld1_1_2_regs")] ) (define_insn "neon_vld1_lane" [(set (match_operand:VDX 0 "s_register_operand" "=w") (unspec:VDX [(mem: (match_operand:SI 1 "s_register_operand" "r")) (match_operand:VDX 2 "s_register_operand" "0") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VLD1_LANE))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[3]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); if (lane < 0 || lane >= max) error ("lane out of range"); if (max == 1) return "vld1.\t%P0, [%1]"; else return "vld1.\t{%P0[%c3]}, [%1]"; } [(set (attr "neon_type") (if_then_else (eq (const_string "") (const_int 2)) (const_string "neon_vld1_1_2_regs") (const_string "neon_vld1_vld2_lane")))] ) (define_insn "neon_vld1_lane" [(set (match_operand:VQX 0 "s_register_operand" "=w") (unspec:VQX [(mem: (match_operand:SI 1 "s_register_operand" "r")) (match_operand:VQX 2 "s_register_operand" "0") (match_operand:SI 3 "immediate_operand" "i")] UNSPEC_VLD1_LANE))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[3]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); int regno = REGNO (operands[0]); if (lane < 0 || lane >= max) error ("lane out of range"); else if (lane >= max / 2) { lane -= max / 2; regno += 2; operands[3] = GEN_INT (lane); } operands[0] = gen_rtx_REG (mode, regno); if (max == 2) return "vld1.\t%P0, [%1]"; else return "vld1.\t{%P0[%c3]}, [%1]"; } [(set (attr "neon_type") (if_then_else (eq (const_string "") (const_int 2)) (const_string "neon_vld1_1_2_regs") (const_string "neon_vld1_vld2_lane")))] ) (define_insn "neon_vld1_dup" [(set (match_operand:VDX 0 "s_register_operand" "=w") (unspec:VDX [(mem: (match_operand:SI 1 "s_register_operand" "r"))] UNSPEC_VLD1_DUP))] "TARGET_NEON" { if (GET_MODE_NUNITS (mode) > 1) return "vld1.\t{%P0[]}, [%1]"; else return "vld1.\t%h0, [%1]"; } [(set (attr "neon_type") (if_then_else (gt (const_string "") (const_string "1")) (const_string "neon_vld2_2_regs_vld1_vld2_all_lanes") (const_string "neon_vld1_1_2_regs")))] ) (define_insn "neon_vld1_dup" [(set (match_operand:VQX 0 "s_register_operand" "=w") (unspec:VQX [(mem: (match_operand:SI 1 "s_register_operand" "r"))] UNSPEC_VLD1_DUP))] "TARGET_NEON" { if (GET_MODE_NUNITS (mode) > 2) return "vld1.\t{%e0[], %f0[]}, [%1]"; else return "vld1.\t%h0, [%1]"; } [(set (attr "neon_type") (if_then_else (gt (const_string "") (const_string "1")) (const_string "neon_vld2_2_regs_vld1_vld2_all_lanes") (const_string "neon_vld1_1_2_regs")))] ) (define_insn "neon_vst1" [(set (mem:VDQX (match_operand:SI 0 "s_register_operand" "r")) (unspec:VDQX [(match_operand:VDQX 1 "s_register_operand" "w")] UNSPEC_VST1))] "TARGET_NEON" "vst1.\t%h1, [%0]" [(set_attr "neon_type" "neon_vst1_1_2_regs_vst2_2_regs")]) (define_insn "neon_vst1_lane" [(set (mem: (match_operand:SI 0 "s_register_operand" "r")) (vec_select: (match_operand:VDX 1 "s_register_operand" "w") (parallel [(match_operand:SI 2 "neon_lane_number" "i")])))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[2]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); if (lane < 0 || lane >= max) error ("lane out of range"); if (max == 1) return "vst1.\t{%P1}, [%0]"; else return "vst1.\t{%P1[%c2]}, [%0]"; } [(set (attr "neon_type") (if_then_else (eq (const_string "") (const_int 1)) (const_string "neon_vst1_1_2_regs_vst2_2_regs") (const_string "neon_vst1_vst2_lane")))]) (define_insn "neon_vst1_lane" [(set (mem: (match_operand:SI 0 "s_register_operand" "r")) (vec_select: (match_operand:VQX 1 "s_register_operand" "w") (parallel [(match_operand:SI 2 "neon_lane_number" "i")])))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[2]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); int regno = REGNO (operands[1]); if (lane < 0 || lane >= max) error ("lane out of range"); else if (lane >= max / 2) { lane -= max / 2; regno += 2; operands[2] = GEN_INT (lane); } operands[1] = gen_rtx_REG (mode, regno); if (max == 2) return "vst1.\t{%P1}, [%0]"; else return "vst1.\t{%P1[%c2]}, [%0]"; } [(set_attr "neon_type" "neon_vst1_vst2_lane")] ) (define_insn "neon_vld2" [(set (match_operand:TI 0 "s_register_operand" "=w") (unspec:TI [(mem:TI (match_operand:SI 1 "s_register_operand" "r")) (unspec:VDX [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD2))] "TARGET_NEON" { if ( == 64) return "vld1.64\t%h0, [%1]"; else return "vld2.\t%h0, [%1]"; } [(set (attr "neon_type") (if_then_else (eq (const_string "") (const_string "64")) (const_string "neon_vld1_1_2_regs") (const_string "neon_vld2_2_regs_vld1_vld2_all_lanes")))] ) (define_insn "neon_vld2" [(set (match_operand:OI 0 "s_register_operand" "=w") (unspec:OI [(mem:OI (match_operand:SI 1 "s_register_operand" "r")) (unspec:VQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD2))] "TARGET_NEON" "vld2.\t%h0, [%1]" [(set_attr "neon_type" "neon_vld2_2_regs_vld1_vld2_all_lanes")]) (define_insn "neon_vld2_lane" [(set (match_operand:TI 0 "s_register_operand" "=w") (unspec:TI [(mem: (match_operand:SI 1 "s_register_operand" "r")) (match_operand:TI 2 "s_register_operand" "0") (match_operand:SI 3 "immediate_operand" "i") (unspec:VD [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD2_LANE))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[3]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); int regno = REGNO (operands[0]); rtx ops[4]; if (lane < 0 || lane >= max) error ("lane out of range"); ops[0] = gen_rtx_REG (DImode, regno); ops[1] = gen_rtx_REG (DImode, regno + 2); ops[2] = operands[1]; ops[3] = operands[3]; output_asm_insn ("vld2.\t{%P0[%c3], %P1[%c3]}, [%2]", ops); return ""; } [(set_attr "neon_type" "neon_vld1_vld2_lane")] ) (define_insn "neon_vld2_lane" [(set (match_operand:OI 0 "s_register_operand" "=w") (unspec:OI [(mem: (match_operand:SI 1 "s_register_operand" "r")) (match_operand:OI 2 "s_register_operand" "0") (match_operand:SI 3 "immediate_operand" "i") (unspec:VMQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD2_LANE))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[3]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); int regno = REGNO (operands[0]); rtx ops[4]; if (lane < 0 || lane >= max) error ("lane out of range"); else if (lane >= max / 2) { lane -= max / 2; regno += 2; } ops[0] = gen_rtx_REG (DImode, regno); ops[1] = gen_rtx_REG (DImode, regno + 4); ops[2] = operands[1]; ops[3] = GEN_INT (lane); output_asm_insn ("vld2.\t{%P0[%c3], %P1[%c3]}, [%2]", ops); return ""; } [(set_attr "neon_type" "neon_vld1_vld2_lane")] ) (define_insn "neon_vld2_dup" [(set (match_operand:TI 0 "s_register_operand" "=w") (unspec:TI [(mem: (match_operand:SI 1 "s_register_operand" "r")) (unspec:VDX [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD2_DUP))] "TARGET_NEON" { if (GET_MODE_NUNITS (mode) > 1) return "vld2.\t{%e0[], %f0[]}, [%1]"; else return "vld1.\t%h0, [%1]"; } [(set (attr "neon_type") (if_then_else (gt (const_string "") (const_string "1")) (const_string "neon_vld2_2_regs_vld1_vld2_all_lanes") (const_string "neon_vld1_1_2_regs")))] ) (define_insn "neon_vst2" [(set (mem:TI (match_operand:SI 0 "s_register_operand" "r")) (unspec:TI [(match_operand:TI 1 "s_register_operand" "w") (unspec:VDX [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VST2))] "TARGET_NEON" { if ( == 64) return "vst1.64\t%h1, [%0]"; else return "vst2.\t%h1, [%0]"; } [(set (attr "neon_type") (if_then_else (eq (const_string "") (const_string "64")) (const_string "neon_vst1_1_2_regs_vst2_2_regs") (const_string "neon_vst1_1_2_regs_vst2_2_regs")))] ) (define_insn "neon_vst2" [(set (mem:OI (match_operand:SI 0 "s_register_operand" "r")) (unspec:OI [(match_operand:OI 1 "s_register_operand" "w") (unspec:VQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VST2))] "TARGET_NEON" "vst2.\t%h1, [%0]" [(set_attr "neon_type" "neon_vst1_1_2_regs_vst2_2_regs")] ) (define_insn "neon_vst2_lane" [(set (mem: (match_operand:SI 0 "s_register_operand" "r")) (unspec: [(match_operand:TI 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (unspec:VD [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VST2_LANE))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[2]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); int regno = REGNO (operands[1]); rtx ops[4]; if (lane < 0 || lane >= max) error ("lane out of range"); ops[0] = operands[0]; ops[1] = gen_rtx_REG (DImode, regno); ops[2] = gen_rtx_REG (DImode, regno + 2); ops[3] = operands[2]; output_asm_insn ("vst2.\t{%P1[%c3], %P2[%c3]}, [%0]", ops); return ""; } [(set_attr "neon_type" "neon_vst1_vst2_lane")] ) (define_insn "neon_vst2_lane" [(set (mem: (match_operand:SI 0 "s_register_operand" "r")) (unspec: [(match_operand:OI 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (unspec:VMQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VST2_LANE))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[2]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); int regno = REGNO (operands[1]); rtx ops[4]; if (lane < 0 || lane >= max) error ("lane out of range"); else if (lane >= max / 2) { lane -= max / 2; regno += 2; } ops[0] = operands[0]; ops[1] = gen_rtx_REG (DImode, regno); ops[2] = gen_rtx_REG (DImode, regno + 4); ops[3] = GEN_INT (lane); output_asm_insn ("vst2.\t{%P1[%c3], %P2[%c3]}, [%0]", ops); return ""; } [(set_attr "neon_type" "neon_vst1_vst2_lane")] ) (define_insn "neon_vld3" [(set (match_operand:EI 0 "s_register_operand" "=w") (unspec:EI [(mem:EI (match_operand:SI 1 "s_register_operand" "r")) (unspec:VDX [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD3))] "TARGET_NEON" { if ( == 64) return "vld1.64\t%h0, [%1]"; else return "vld3.\t%h0, [%1]"; } [(set (attr "neon_type") (if_then_else (eq (const_string "") (const_string "64")) (const_string "neon_vld1_1_2_regs") (const_string "neon_vld3_vld4")))] ) (define_expand "neon_vld3" [(match_operand:CI 0 "s_register_operand" "=w") (match_operand:SI 1 "s_register_operand" "+r") (unspec:VQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] "TARGET_NEON" { emit_insn (gen_neon_vld3qa (operands[0], operands[0], operands[1], operands[1])); emit_insn (gen_neon_vld3qb (operands[0], operands[0], operands[1], operands[1])); DONE; }) (define_insn "neon_vld3qa" [(set (match_operand:CI 0 "s_register_operand" "=w") (unspec:CI [(mem:CI (match_operand:SI 3 "s_register_operand" "2")) (match_operand:CI 1 "s_register_operand" "0") (unspec:VQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD3A)) (set (match_operand:SI 2 "s_register_operand" "=r") (plus:SI (match_dup 3) (const_int 24)))] "TARGET_NEON" { int regno = REGNO (operands[0]); rtx ops[4]; ops[0] = gen_rtx_REG (DImode, regno); ops[1] = gen_rtx_REG (DImode, regno + 4); ops[2] = gen_rtx_REG (DImode, regno + 8); ops[3] = operands[2]; output_asm_insn ("vld3.\t{%P0, %P1, %P2}, [%3]!", ops); return ""; } [(set_attr "neon_type" "neon_vld3_vld4")] ) (define_insn "neon_vld3qb" [(set (match_operand:CI 0 "s_register_operand" "=w") (unspec:CI [(mem:CI (match_operand:SI 3 "s_register_operand" "2")) (match_operand:CI 1 "s_register_operand" "0") (unspec:VQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD3B)) (set (match_operand:SI 2 "s_register_operand" "=r") (plus:SI (match_dup 3) (const_int 24)))] "TARGET_NEON" { int regno = REGNO (operands[0]); rtx ops[4]; ops[0] = gen_rtx_REG (DImode, regno + 2); ops[1] = gen_rtx_REG (DImode, regno + 6); ops[2] = gen_rtx_REG (DImode, regno + 10); ops[3] = operands[2]; output_asm_insn ("vld3.\t{%P0, %P1, %P2}, [%3]!", ops); return ""; } [(set_attr "neon_type" "neon_vld3_vld4")] ) (define_insn "neon_vld3_lane" [(set (match_operand:EI 0 "s_register_operand" "=w") (unspec:EI [(mem: (match_operand:SI 1 "s_register_operand" "r")) (match_operand:EI 2 "s_register_operand" "0") (match_operand:SI 3 "immediate_operand" "i") (unspec:VD [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD3_LANE))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[3]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); int regno = REGNO (operands[0]); rtx ops[5]; if (lane < 0 || lane >= max) error ("lane out of range"); ops[0] = gen_rtx_REG (DImode, regno); ops[1] = gen_rtx_REG (DImode, regno + 2); ops[2] = gen_rtx_REG (DImode, regno + 4); ops[3] = operands[1]; ops[4] = operands[3]; output_asm_insn ("vld3.\t{%P0[%c4], %P1[%c4], %P2[%c4]}, [%3]", ops); return ""; } [(set_attr "neon_type" "neon_vld3_vld4_lane")] ) (define_insn "neon_vld3_lane" [(set (match_operand:CI 0 "s_register_operand" "=w") (unspec:CI [(mem: (match_operand:SI 1 "s_register_operand" "r")) (match_operand:CI 2 "s_register_operand" "0") (match_operand:SI 3 "immediate_operand" "i") (unspec:VMQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD3_LANE))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[3]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); int regno = REGNO (operands[0]); rtx ops[5]; if (lane < 0 || lane >= max) error ("lane out of range"); else if (lane >= max / 2) { lane -= max / 2; regno += 2; } ops[0] = gen_rtx_REG (DImode, regno); ops[1] = gen_rtx_REG (DImode, regno + 4); ops[2] = gen_rtx_REG (DImode, regno + 8); ops[3] = operands[1]; ops[4] = GEN_INT (lane); output_asm_insn ("vld3.\t{%P0[%c4], %P1[%c4], %P2[%c4]}, [%3]", ops); return ""; } [(set_attr "neon_type" "neon_vld3_vld4_lane")] ) (define_insn "neon_vld3_dup" [(set (match_operand:EI 0 "s_register_operand" "=w") (unspec:EI [(mem: (match_operand:SI 1 "s_register_operand" "r")) (unspec:VDX [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD3_DUP))] "TARGET_NEON" { if (GET_MODE_NUNITS (mode) > 1) { int regno = REGNO (operands[0]); rtx ops[4]; ops[0] = gen_rtx_REG (DImode, regno); ops[1] = gen_rtx_REG (DImode, regno + 2); ops[2] = gen_rtx_REG (DImode, regno + 4); ops[3] = operands[1]; output_asm_insn ("vld3.\t{%P0[], %P1[], %P2[]}, [%3]", ops); return ""; } else return "vld1.\t%h0, [%1]"; } [(set (attr "neon_type") (if_then_else (gt (const_string "") (const_string "1")) (const_string "neon_vld3_vld4_all_lanes") (const_string "neon_vld1_1_2_regs")))]) (define_insn "neon_vst3" [(set (mem:EI (match_operand:SI 0 "s_register_operand" "r")) (unspec:EI [(match_operand:EI 1 "s_register_operand" "w") (unspec:VDX [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VST3))] "TARGET_NEON" { if ( == 64) return "vst1.64\t%h1, [%0]"; else return "vst3.\t%h1, [%0]"; } [(set (attr "neon_type") (if_then_else (eq (const_string "") (const_string "64")) (const_string "neon_vst1_1_2_regs_vst2_2_regs") (const_string "neon_vst2_4_regs_vst3_vst4")))]) (define_expand "neon_vst3" [(match_operand:SI 0 "s_register_operand" "+r") (match_operand:CI 1 "s_register_operand" "w") (unspec:VQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] "TARGET_NEON" { emit_insn (gen_neon_vst3qa (operands[0], operands[0], operands[1])); emit_insn (gen_neon_vst3qb (operands[0], operands[0], operands[1])); DONE; }) (define_insn "neon_vst3qa" [(set (mem:EI (match_operand:SI 1 "s_register_operand" "0")) (unspec:EI [(match_operand:CI 2 "s_register_operand" "w") (unspec:VQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VST3A)) (set (match_operand:SI 0 "s_register_operand" "=r") (plus:SI (match_dup 1) (const_int 24)))] "TARGET_NEON" { int regno = REGNO (operands[2]); rtx ops[4]; ops[0] = operands[0]; ops[1] = gen_rtx_REG (DImode, regno); ops[2] = gen_rtx_REG (DImode, regno + 4); ops[3] = gen_rtx_REG (DImode, regno + 8); output_asm_insn ("vst3.\t{%P1, %P2, %P3}, [%0]!", ops); return ""; } [(set_attr "neon_type" "neon_vst2_4_regs_vst3_vst4")] ) (define_insn "neon_vst3qb" [(set (mem:EI (match_operand:SI 1 "s_register_operand" "0")) (unspec:EI [(match_operand:CI 2 "s_register_operand" "w") (unspec:VQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VST3B)) (set (match_operand:SI 0 "s_register_operand" "=r") (plus:SI (match_dup 1) (const_int 24)))] "TARGET_NEON" { int regno = REGNO (operands[2]); rtx ops[4]; ops[0] = operands[0]; ops[1] = gen_rtx_REG (DImode, regno + 2); ops[2] = gen_rtx_REG (DImode, regno + 6); ops[3] = gen_rtx_REG (DImode, regno + 10); output_asm_insn ("vst3.\t{%P1, %P2, %P3}, [%0]!", ops); return ""; } [(set_attr "neon_type" "neon_vst2_4_regs_vst3_vst4")] ) (define_insn "neon_vst3_lane" [(set (mem: (match_operand:SI 0 "s_register_operand" "r")) (unspec: [(match_operand:EI 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (unspec:VD [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VST3_LANE))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[2]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); int regno = REGNO (operands[1]); rtx ops[5]; if (lane < 0 || lane >= max) error ("lane out of range"); ops[0] = operands[0]; ops[1] = gen_rtx_REG (DImode, regno); ops[2] = gen_rtx_REG (DImode, regno + 2); ops[3] = gen_rtx_REG (DImode, regno + 4); ops[4] = operands[2]; output_asm_insn ("vst3.\t{%P1[%c4], %P2[%c4], %P3[%c4]}, [%0]", ops); return ""; } [(set_attr "neon_type" "neon_vst3_vst4_lane")] ) (define_insn "neon_vst3_lane" [(set (mem: (match_operand:SI 0 "s_register_operand" "r")) (unspec: [(match_operand:CI 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (unspec:VMQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VST3_LANE))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[2]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); int regno = REGNO (operands[1]); rtx ops[5]; if (lane < 0 || lane >= max) error ("lane out of range"); else if (lane >= max / 2) { lane -= max / 2; regno += 2; } ops[0] = operands[0]; ops[1] = gen_rtx_REG (DImode, regno); ops[2] = gen_rtx_REG (DImode, regno + 4); ops[3] = gen_rtx_REG (DImode, regno + 8); ops[4] = GEN_INT (lane); output_asm_insn ("vst3.\t{%P1[%c4], %P2[%c4], %P3[%c4]}, [%0]", ops); return ""; } [(set_attr "neon_type" "neon_vst3_vst4_lane")]) (define_insn "neon_vld4" [(set (match_operand:OI 0 "s_register_operand" "=w") (unspec:OI [(mem:OI (match_operand:SI 1 "s_register_operand" "r")) (unspec:VDX [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD4))] "TARGET_NEON" { if ( == 64) return "vld1.64\t%h0, [%1]"; else return "vld4.\t%h0, [%1]"; } [(set (attr "neon_type") (if_then_else (eq (const_string "") (const_string "64")) (const_string "neon_vld1_1_2_regs") (const_string "neon_vld3_vld4")))] ) (define_expand "neon_vld4" [(match_operand:XI 0 "s_register_operand" "=w") (match_operand:SI 1 "s_register_operand" "+r") (unspec:VQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] "TARGET_NEON" { emit_insn (gen_neon_vld4qa (operands[0], operands[0], operands[1], operands[1])); emit_insn (gen_neon_vld4qb (operands[0], operands[0], operands[1], operands[1])); DONE; }) (define_insn "neon_vld4qa" [(set (match_operand:XI 0 "s_register_operand" "=w") (unspec:XI [(mem:XI (match_operand:SI 3 "s_register_operand" "2")) (match_operand:XI 1 "s_register_operand" "0") (unspec:VQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD4A)) (set (match_operand:SI 2 "s_register_operand" "=r") (plus:SI (match_dup 3) (const_int 32)))] "TARGET_NEON" { int regno = REGNO (operands[0]); rtx ops[5]; ops[0] = gen_rtx_REG (DImode, regno); ops[1] = gen_rtx_REG (DImode, regno + 4); ops[2] = gen_rtx_REG (DImode, regno + 8); ops[3] = gen_rtx_REG (DImode, regno + 12); ops[4] = operands[2]; output_asm_insn ("vld4.\t{%P0, %P1, %P2, %P3}, [%4]!", ops); return ""; } [(set_attr "neon_type" "neon_vld3_vld4")] ) (define_insn "neon_vld4qb" [(set (match_operand:XI 0 "s_register_operand" "=w") (unspec:XI [(mem:XI (match_operand:SI 3 "s_register_operand" "2")) (match_operand:XI 1 "s_register_operand" "0") (unspec:VQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD4B)) (set (match_operand:SI 2 "s_register_operand" "=r") (plus:SI (match_dup 3) (const_int 32)))] "TARGET_NEON" { int regno = REGNO (operands[0]); rtx ops[5]; ops[0] = gen_rtx_REG (DImode, regno + 2); ops[1] = gen_rtx_REG (DImode, regno + 6); ops[2] = gen_rtx_REG (DImode, regno + 10); ops[3] = gen_rtx_REG (DImode, regno + 14); ops[4] = operands[2]; output_asm_insn ("vld4.\t{%P0, %P1, %P2, %P3}, [%4]!", ops); return ""; } [(set_attr "neon_type" "neon_vld3_vld4")] ) (define_insn "neon_vld4_lane" [(set (match_operand:OI 0 "s_register_operand" "=w") (unspec:OI [(mem: (match_operand:SI 1 "s_register_operand" "r")) (match_operand:OI 2 "s_register_operand" "0") (match_operand:SI 3 "immediate_operand" "i") (unspec:VD [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD4_LANE))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[3]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); int regno = REGNO (operands[0]); rtx ops[6]; if (lane < 0 || lane >= max) error ("lane out of range"); ops[0] = gen_rtx_REG (DImode, regno); ops[1] = gen_rtx_REG (DImode, regno + 2); ops[2] = gen_rtx_REG (DImode, regno + 4); ops[3] = gen_rtx_REG (DImode, regno + 6); ops[4] = operands[1]; ops[5] = operands[3]; output_asm_insn ("vld4.\t{%P0[%c5], %P1[%c5], %P2[%c5], %P3[%c5]}, [%4]", ops); return ""; } [(set_attr "neon_type" "neon_vld3_vld4_lane")] ) (define_insn "neon_vld4_lane" [(set (match_operand:XI 0 "s_register_operand" "=w") (unspec:XI [(mem: (match_operand:SI 1 "s_register_operand" "r")) (match_operand:XI 2 "s_register_operand" "0") (match_operand:SI 3 "immediate_operand" "i") (unspec:VMQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD4_LANE))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[3]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); int regno = REGNO (operands[0]); rtx ops[6]; if (lane < 0 || lane >= max) error ("lane out of range"); else if (lane >= max / 2) { lane -= max / 2; regno += 2; } ops[0] = gen_rtx_REG (DImode, regno); ops[1] = gen_rtx_REG (DImode, regno + 4); ops[2] = gen_rtx_REG (DImode, regno + 8); ops[3] = gen_rtx_REG (DImode, regno + 12); ops[4] = operands[1]; ops[5] = GEN_INT (lane); output_asm_insn ("vld4.\t{%P0[%c5], %P1[%c5], %P2[%c5], %P3[%c5]}, [%4]", ops); return ""; } [(set_attr "neon_type" "neon_vld3_vld4_lane")] ) (define_insn "neon_vld4_dup" [(set (match_operand:OI 0 "s_register_operand" "=w") (unspec:OI [(mem: (match_operand:SI 1 "s_register_operand" "r")) (unspec:VDX [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VLD4_DUP))] "TARGET_NEON" { if (GET_MODE_NUNITS (mode) > 1) { int regno = REGNO (operands[0]); rtx ops[5]; ops[0] = gen_rtx_REG (DImode, regno); ops[1] = gen_rtx_REG (DImode, regno + 2); ops[2] = gen_rtx_REG (DImode, regno + 4); ops[3] = gen_rtx_REG (DImode, regno + 6); ops[4] = operands[1]; output_asm_insn ("vld4.\t{%P0[], %P1[], %P2[], %P3[]}, [%4]", ops); return ""; } else return "vld1.\t%h0, [%1]"; } [(set (attr "neon_type") (if_then_else (gt (const_string "") (const_string "1")) (const_string "neon_vld3_vld4_all_lanes") (const_string "neon_vld1_1_2_regs")))] ) (define_insn "neon_vst4" [(set (mem:OI (match_operand:SI 0 "s_register_operand" "r")) (unspec:OI [(match_operand:OI 1 "s_register_operand" "w") (unspec:VDX [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VST4))] "TARGET_NEON" { if ( == 64) return "vst1.64\t%h1, [%0]"; else return "vst4.\t%h1, [%0]"; } [(set (attr "neon_type") (if_then_else (eq (const_string "") (const_string "64")) (const_string "neon_vst1_1_2_regs_vst2_2_regs") (const_string "neon_vst2_4_regs_vst3_vst4")))] ) (define_expand "neon_vst4" [(match_operand:SI 0 "s_register_operand" "+r") (match_operand:XI 1 "s_register_operand" "w") (unspec:VQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] "TARGET_NEON" { emit_insn (gen_neon_vst4qa (operands[0], operands[0], operands[1])); emit_insn (gen_neon_vst4qb (operands[0], operands[0], operands[1])); DONE; }) (define_insn "neon_vst4qa" [(set (mem:OI (match_operand:SI 1 "s_register_operand" "0")) (unspec:OI [(match_operand:XI 2 "s_register_operand" "w") (unspec:VQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VST4A)) (set (match_operand:SI 0 "s_register_operand" "=r") (plus:SI (match_dup 1) (const_int 32)))] "TARGET_NEON" { int regno = REGNO (operands[2]); rtx ops[5]; ops[0] = operands[0]; ops[1] = gen_rtx_REG (DImode, regno); ops[2] = gen_rtx_REG (DImode, regno + 4); ops[3] = gen_rtx_REG (DImode, regno + 8); ops[4] = gen_rtx_REG (DImode, regno + 12); output_asm_insn ("vst4.\t{%P1, %P2, %P3, %P4}, [%0]!", ops); return ""; } [(set_attr "neon_type" "neon_vst2_4_regs_vst3_vst4")] ) (define_insn "neon_vst4qb" [(set (mem:OI (match_operand:SI 1 "s_register_operand" "0")) (unspec:OI [(match_operand:XI 2 "s_register_operand" "w") (unspec:VQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VST4B)) (set (match_operand:SI 0 "s_register_operand" "=r") (plus:SI (match_dup 1) (const_int 32)))] "TARGET_NEON" { int regno = REGNO (operands[2]); rtx ops[5]; ops[0] = operands[0]; ops[1] = gen_rtx_REG (DImode, regno + 2); ops[2] = gen_rtx_REG (DImode, regno + 6); ops[3] = gen_rtx_REG (DImode, regno + 10); ops[4] = gen_rtx_REG (DImode, regno + 14); output_asm_insn ("vst4.\t{%P1, %P2, %P3, %P4}, [%0]!", ops); return ""; } [(set_attr "neon_type" "neon_vst2_4_regs_vst3_vst4")] ) (define_insn "neon_vst4_lane" [(set (mem: (match_operand:SI 0 "s_register_operand" "r")) (unspec: [(match_operand:OI 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (unspec:VD [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VST4_LANE))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[2]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); int regno = REGNO (operands[1]); rtx ops[6]; if (lane < 0 || lane >= max) error ("lane out of range"); ops[0] = operands[0]; ops[1] = gen_rtx_REG (DImode, regno); ops[2] = gen_rtx_REG (DImode, regno + 2); ops[3] = gen_rtx_REG (DImode, regno + 4); ops[4] = gen_rtx_REG (DImode, regno + 6); ops[5] = operands[2]; output_asm_insn ("vst4.\t{%P1[%c5], %P2[%c5], %P3[%c5], %P4[%c5]}, [%0]", ops); return ""; } [(set_attr "neon_type" "neon_vst3_vst4_lane")] ) (define_insn "neon_vst4_lane" [(set (mem: (match_operand:SI 0 "s_register_operand" "r")) (unspec: [(match_operand:XI 1 "s_register_operand" "w") (match_operand:SI 2 "immediate_operand" "i") (unspec:VMQ [(const_int 0)] UNSPEC_VSTRUCTDUMMY)] UNSPEC_VST4_LANE))] "TARGET_NEON" { HOST_WIDE_INT lane = INTVAL (operands[2]); HOST_WIDE_INT max = GET_MODE_NUNITS (mode); int regno = REGNO (operands[1]); rtx ops[6]; if (lane < 0 || lane >= max) error ("lane out of range"); else if (lane >= max / 2) { lane -= max / 2; regno += 2; } ops[0] = operands[0]; ops[1] = gen_rtx_REG (DImode, regno); ops[2] = gen_rtx_REG (DImode, regno + 4); ops[3] = gen_rtx_REG (DImode, regno + 8); ops[4] = gen_rtx_REG (DImode, regno + 12); ops[5] = GEN_INT (lane); output_asm_insn ("vst4.\t{%P1[%c5], %P2[%c5], %P3[%c5], %P4[%c5]}, [%0]", ops); return ""; } [(set_attr "neon_type" "neon_vst3_vst4_lane")] ) (define_expand "neon_vand" [(match_operand:VDQX 0 "s_register_operand" "") (match_operand:VDQX 1 "s_register_operand" "") (match_operand:VDQX 2 "neon_inv_logic_op2" "") (match_operand:SI 3 "immediate_operand" "")] "TARGET_NEON" { emit_insn (gen_and3 (operands[0], operands[1], operands[2])); DONE; }) (define_expand "neon_vorr" [(match_operand:VDQX 0 "s_register_operand" "") (match_operand:VDQX 1 "s_register_operand" "") (match_operand:VDQX 2 "neon_logic_op2" "") (match_operand:SI 3 "immediate_operand" "")] "TARGET_NEON" { emit_insn (gen_ior3 (operands[0], operands[1], operands[2])); DONE; }) (define_expand "neon_veor" [(match_operand:VDQX 0 "s_register_operand" "") (match_operand:VDQX 1 "s_register_operand" "") (match_operand:VDQX 2 "s_register_operand" "") (match_operand:SI 3 "immediate_operand" "")] "TARGET_NEON" { emit_insn (gen_xor3 (operands[0], operands[1], operands[2])); DONE; }) (define_expand "neon_vbic" [(match_operand:VDQX 0 "s_register_operand" "") (match_operand:VDQX 1 "s_register_operand" "") (match_operand:VDQX 2 "neon_logic_op2" "") (match_operand:SI 3 "immediate_operand" "")] "TARGET_NEON" { emit_insn (gen_bic3_neon (operands[0], operands[1], operands[2])); DONE; }) (define_expand "neon_vorn" [(match_operand:VDQX 0 "s_register_operand" "") (match_operand:VDQX 1 "s_register_operand" "") (match_operand:VDQX 2 "neon_inv_logic_op2" "") (match_operand:SI 3 "immediate_operand" "")] "TARGET_NEON" { emit_insn (gen_orn3_neon (operands[0], operands[1], operands[2])); DONE; }) (define_insn "neon_vec_unpack_lo_" [(set (match_operand: 0 "register_operand" "=w") (SE: (vec_select: (match_operand:VU 1 "register_operand" "w") (match_operand:VU 2 "vect_par_constant_low" ""))))] "TARGET_NEON" "vmovl. %q0, %e1" [(set_attr "neon_type" "neon_shift_1")] ) (define_insn "neon_vec_unpack_hi_" [(set (match_operand: 0 "register_operand" "=w") (SE: (vec_select: (match_operand:VU 1 "register_operand" "w") (match_operand:VU 2 "vect_par_constant_high" ""))))] "TARGET_NEON" "vmovl. %q0, %f1" [(set_attr "neon_type" "neon_shift_1")] ) (define_expand "vec_unpack_hi_" [(match_operand: 0 "register_operand" "") (SE: (match_operand:VU 1 "register_operand"))] "TARGET_NEON" { rtvec v = rtvec_alloc (/2) ; rtx t1; int i; for (i = 0; i < (/2); i++) RTVEC_ELT (v, i) = GEN_INT ((/2) + i); t1 = gen_rtx_PARALLEL (mode, v); emit_insn (gen_neon_vec_unpack_hi_ (operands[0], operands[1], t1)); DONE; } ) (define_expand "vec_unpack_lo_" [(match_operand: 0 "register_operand" "") (SE: (match_operand:VU 1 "register_operand" ""))] "TARGET_NEON" { rtvec v = rtvec_alloc (/2) ; rtx t1; int i; for (i = 0; i < (/2) ; i++) RTVEC_ELT (v, i) = GEN_INT (i); t1 = gen_rtx_PARALLEL (mode, v); emit_insn (gen_neon_vec_unpack_lo_ (operands[0], operands[1], t1)); DONE; } ) (define_insn "neon_vec_mult_lo_" [(set (match_operand: 0 "register_operand" "=w") (mult: (SE: (vec_select: (match_operand:VU 1 "register_operand" "w") (match_operand:VU 2 "vect_par_constant_low" ""))) (SE: (vec_select: (match_operand:VU 3 "register_operand" "w") (match_dup 2)))))] "TARGET_NEON" "vmull. %q0, %e1, %e3" [(set_attr "neon_type" "neon_shift_1")] ) (define_expand "vec_widen_mult_lo_" [(match_operand: 0 "register_operand" "") (SE: (match_operand:VU 1 "register_operand" "")) (SE: (match_operand:VU 2 "register_operand" ""))] "TARGET_NEON" { rtvec v = rtvec_alloc (/2) ; rtx t1; int i; for (i = 0; i < (/2) ; i++) RTVEC_ELT (v, i) = GEN_INT (i); t1 = gen_rtx_PARALLEL (mode, v); emit_insn (gen_neon_vec_mult_lo_ (operands[0], operands[1], t1, operands[2])); DONE; } ) (define_insn "neon_vec_mult_hi_" [(set (match_operand: 0 "register_operand" "=w") (mult: (SE: (vec_select: (match_operand:VU 1 "register_operand" "w") (match_operand:VU 2 "vect_par_constant_high" ""))) (SE: (vec_select: (match_operand:VU 3 "register_operand" "w") (match_dup 2)))))] "TARGET_NEON" "vmull. %q0, %f1, %f3" [(set_attr "neon_type" "neon_shift_1")] ) (define_expand "vec_widen_mult_hi_" [(match_operand: 0 "register_operand" "") (SE: (match_operand:VU 1 "register_operand" "")) (SE: (match_operand:VU 2 "register_operand" ""))] "TARGET_NEON" { rtvec v = rtvec_alloc (/2) ; rtx t1; int i; for (i = 0; i < (/2) ; i++) RTVEC_ELT (v, i) = GEN_INT (/2 + i); t1 = gen_rtx_PARALLEL (mode, v); emit_insn (gen_neon_vec_mult_hi_ (operands[0], operands[1], t1, operands[2])); DONE; } ) ;; Vectorize for non-neon-quad case (define_insn "neon_unpack_" [(set (match_operand: 0 "register_operand" "=w") (SE: (match_operand:VDI 1 "register_operand" "w")))] "TARGET_NEON" "vmovl. %q0, %P1" [(set_attr "neon_type" "neon_shift_1")] ) (define_expand "vec_unpack_lo_" [(match_operand: 0 "register_operand" "") (SE:(match_operand:VDI 1 "register_operand"))] "TARGET_NEON" { rtx tmpreg = gen_reg_rtx (mode); emit_insn (gen_neon_unpack_ (tmpreg, operands[1])); emit_insn (gen_neon_vget_low (operands[0], tmpreg)); DONE; } ) (define_expand "vec_unpack_hi_" [(match_operand: 0 "register_operand" "") (SE:(match_operand:VDI 1 "register_operand"))] "TARGET_NEON" { rtx tmpreg = gen_reg_rtx (mode); emit_insn (gen_neon_unpack_ (tmpreg, operands[1])); emit_insn (gen_neon_vget_high (operands[0], tmpreg)); DONE; } ) (define_insn "neon_vec_mult_" [(set (match_operand: 0 "register_operand" "=w") (mult: (SE: (match_operand:VDI 1 "register_operand" "w")) (SE: (match_operand:VDI 2 "register_operand" "w"))))] "TARGET_NEON" "vmull. %q0, %P1, %P2" [(set_attr "neon_type" "neon_shift_1")] ) (define_expand "vec_widen_mult_hi_" [(match_operand: 0 "register_operand" "") (SE: (match_operand:VDI 1 "register_operand" "")) (SE: (match_operand:VDI 2 "register_operand" ""))] "TARGET_NEON" { rtx tmpreg = gen_reg_rtx (mode); emit_insn (gen_neon_vec_mult_ (tmpreg, operands[1], operands[2])); emit_insn (gen_neon_vget_high (operands[0], tmpreg)); DONE; } ) (define_expand "vec_widen_mult_lo_" [(match_operand: 0 "register_operand" "") (SE: (match_operand:VDI 1 "register_operand" "")) (SE: (match_operand:VDI 2 "register_operand" ""))] "TARGET_NEON" { rtx tmpreg = gen_reg_rtx (mode); emit_insn (gen_neon_vec_mult_ (tmpreg, operands[1], operands[2])); emit_insn (gen_neon_vget_low (operands[0], tmpreg)); DONE; } ) ;; The case when using all quad registers. (define_insn "vec_pack_trunc_" [(set (match_operand: 0 "register_operand" "=&w") (vec_concat: (truncate: (match_operand:VN 1 "register_operand" "w")) (truncate: (match_operand:VN 2 "register_operand" "w"))))] "TARGET_NEON" "vmovn.i\t%e0, %q1\;vmovn.i\t%f0, %q2" [(set_attr "neon_type" "neon_shift_1") (set_attr "length" "8")] ) ;; For the non-quad case. (define_insn "neon_vec_pack_trunc_" [(set (match_operand: 0 "register_operand" "=w") (truncate: (match_operand:VN 1 "register_operand" "w")))] "TARGET_NEON" "vmovn.i\t%P0, %q1" [(set_attr "neon_type" "neon_shift_1")] ) (define_expand "vec_pack_trunc_" [(match_operand: 0 "register_operand" "") (match_operand:VSHFT 1 "register_operand" "") (match_operand:VSHFT 2 "register_operand")] "TARGET_NEON" { rtx tempreg = gen_reg_rtx (mode); emit_insn (gen_move_lo_quad_ (tempreg, operands[1])); emit_insn (gen_move_hi_quad_ (tempreg, operands[2])); emit_insn (gen_neon_vec_pack_trunc_ (operands[0], tempreg)); DONE; })