Add new output file, to make conversion to GCC dejagnu testsuite easier.

27 files changed, 141 insertions, 50 deletions

diff --git a/Makefile b/Makefile
index 981648e..78afa9c 100644
--- a/Makefile
+++ b/Makefile
@@ -85,7 +85,7 @@ compute_ref.axf: scatter.scat compute_ref.rvct.o retarget.rvct.o	\
 	$(LD.rvct) $(LDFLAGS.rvct) --scatter $^ -o $@
 
 compute_ref.rvct.o retarget.rvct.o: %.rvct.o: %.c
-	$(CC.rvct) $(CFLAGS.rvct) -c $^ -o $@ -DREFFILE=\"$(REFRVCT)\"
+	$(CC.rvct) $(CFLAGS.rvct) -c $^ -o $@ -DREFFILE=\"$(REFRVCT)\" -DGCCTESTS_FILE=\"expected_input4gcc.txt\"
 
 ref_%.rvct.o: ref_%.c stm-arm-neon-ref.h
 	$(CC.rvct) $(CFLAGS.rvct) -c $< -o $@
@@ -162,4 +162,4 @@ ref_vsubw.*.o: ref_vaddw.c
 ref_vcage.*.o ref_vcale.*.o ref_vcagt.*.o ref_vcalt.*.o: ref_v_comp_f_op.c
 
 clean:
-	rm -f *.o *.log stm-arm-neon.refrvct
+	rm -f *.o *.log stm-arm-neon.refrvct expected_input4gcc.txt
diff --git a/compute_ref.c b/compute_ref.c
index d6b364e..331fa01 100644
--- a/compute_ref.c
+++ b/compute_ref.c
@@ -35,6 +35,7 @@ THE SOFTWARE.
 
 FILE* log_file = NULL;
 FILE* ref_file = NULL;
+FILE* gcc_tests_file = NULL;
 
 #define LOGFILE "stm-arm-neon-ref.log"
 
@@ -42,6 +43,7 @@ void cleanup (void)
 {
   if (log_file) fclose (log_file);
   if (ref_file) fclose (ref_file);
+  if (gcc_tests_file) fclose (gcc_tests_file);
   exit (1);
 }
 
@@ -211,6 +213,12 @@ int main (void)
     cleanup ();
   }
 
+  gcc_tests_file = fopen (GCCTESTS_FILE, "w");
+  if (gcc_tests_file == NULL) {
+    fprintf (log_file, "Error opening GCC ref file %s\n", GCCTESTS_FILE);
+    cleanup ();
+  }
+
   fprintf (log_file, "Computing refs....\n");
 
   exec_vld1 ();
diff --git a/ref_v_binary_sat_op.c b/ref_v_binary_sat_op.c
index b2cae86..7c72e15 100644
--- a/ref_v_binary_sat_op.c
+++ b/ref_v_binary_sat_op.c
@@ -47,7 +47,8 @@ FNNAME (INSN_NAME)
 		      VECT_VAR(vector2, T1, W, N));	\
   vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N),		\
 		    VECT_VAR(vector_res, T1, W, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_##T2##W), \
+		     xSTR(T1), W, N)
 
 #define TEST_BINARY_SAT_OP(INSN, Q, T1, T2, W, N)	\
   TEST_BINARY_SAT_OP1(INSN, Q, T1, T2, W, N)
diff --git a/ref_v_unary_sat_op.c b/ref_v_unary_sat_op.c
index fdee616..37c1757 100644
--- a/ref_v_unary_sat_op.c
+++ b/ref_v_unary_sat_op.c
@@ -45,7 +45,8 @@ FNNAME (INSN_NAME)
     INSN##Q##_##T2##W(VECT_VAR(vector, T1, W, N));	\
   vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N),		\
 		    VECT_VAR(vector_res, T1, W, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_##T2##W),	\
+		     xSTR(T1), W, N)
 
 #define TEST_UNARY_SAT_OP(INSN, Q, T1, T2, W, N)	\
   TEST_UNARY_SAT_OP1(INSN, Q, T1, T2, W, N)
diff --git a/ref_vqdmlal.c b/ref_vqdmlal.c
index ff7ce56..c686c9d 100644
--- a/ref_vqdmlal.c
+++ b/ref_vqdmlal.c
@@ -52,7 +52,8 @@ FNNAME (INSN_NAME)
 		    VECT_VAR(vector4, T1, W2, N));	\
   vst1q_##T2##W(VECT_VAR(result, T1, W, N),		\
 		VECT_VAR(vector_res, T1, W, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W2))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W2),	\
+		     xSTR(T1), W, N)
 
 #define TEST_VQDMLXL(INSN, T1, T2, W, W2, N)	\
   TEST_VQDMLXL1(INSN, T1, T2, W, W2, N)
diff --git a/ref_vqdmlal_lane.c b/ref_vqdmlal_lane.c
index bdc19c9..903b822 100644
--- a/ref_vqdmlal_lane.c
+++ b/ref_vqdmlal_lane.c
@@ -52,7 +52,8 @@ FNNAME (INSN_NAME)
 		    V);					\
   vst1q_##T2##W(VECT_VAR(result, T1, W, N),		\
 		VECT_VAR(vector_res, T1, W, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W2))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W2),	\
+		     xSTR(T1), W, N)
 
 #define TEST_VQDMLXL_LANE(INSN, T1, T2, W, W2, N, V)	\
   TEST_VQDMLXL_LANE1(INSN, T1, T2, W, W2, N, V)
diff --git a/ref_vqdmlal_n.c b/ref_vqdmlal_n.c
index 6d4dede..c9a3dc1 100644
--- a/ref_vqdmlal_n.c
+++ b/ref_vqdmlal_n.c
@@ -51,7 +51,8 @@ FNNAME (INSN_NAME)
 		    V);					\
   vst1q_##T2##W(VECT_VAR(result, T1, W, N),		\
 		VECT_VAR(vector_res, T1, W, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W2))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W2),	\
+		     xSTR(T1), W, N)
 
 #define TEST_VQDMLXL_N(INSN, T1, T2, W, W2, N, V)	\
   TEST_VQDMLXL_N1(INSN, T1, T2, W, W2, N, V)
diff --git a/ref_vqdmulh.c b/ref_vqdmulh.c
index 0063ec0..a9ee93f 100644
--- a/ref_vqdmulh.c
+++ b/ref_vqdmulh.c
@@ -47,7 +47,8 @@ FNNAME (INSN)
 		      VECT_VAR(vector2, T1, W, N));	\
   vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N),		\
 		    VECT_VAR(vector_res, T1, W, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_##T2##W),	\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQDMULH1(INSN, Q, T1, T2, W, N)	\
diff --git a/ref_vqdmulh_lane.c b/ref_vqdmulh_lane.c
index 9f4c7da..2d6bb39 100644
--- a/ref_vqdmulh_lane.c
+++ b/ref_vqdmulh_lane.c
@@ -47,7 +47,8 @@ FNNAME (INSN)
 			   L);					\
   vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N),			\
 		    VECT_VAR(vector_res, T1, W, N));		\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_lane_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_lane_##T2##W),	\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQDMULH_LANE1(INSN, Q, T1, T2, W, N, N2, L)	\
diff --git a/ref_vqdmulh_n.c b/ref_vqdmulh_n.c
index d009377..847edba 100644
--- a/ref_vqdmulh_n.c
+++ b/ref_vqdmulh_n.c
@@ -41,14 +41,15 @@ FNNAME (INSN)
   int i;
 
   /* vector_res = vqdmulh_n(vector,val), then store the result.  */
-#define TEST_VQDMULH_N2(INSN, Q, T1, T2, W, N, L)	\
-  Set_Neon_Overflow(0);					\
-  VECT_VAR(vector_res, T1, W, N) =			\
-    INSN##Q##_n_##T2##W(VECT_VAR(vector, T1, W, N),	\
-			L);				\
-  vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N),		\
-		    VECT_VAR(vector_res, T1, W, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_n_##T2##W))
+#define TEST_VQDMULH_N2(INSN, Q, T1, T2, W, N, L)		\
+  Set_Neon_Overflow(0);						\
+  VECT_VAR(vector_res, T1, W, N) =				\
+    INSN##Q##_n_##T2##W(VECT_VAR(vector, T1, W, N),		\
+			L);					\
+  vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N),			\
+		    VECT_VAR(vector_res, T1, W, N));		\
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_n_##T2##W),	\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQDMULH_N1(INSN, Q, T1, T2, W, N, L)	\
diff --git a/ref_vqdmull.c b/ref_vqdmull.c
index 2d5f696..1d6bb6e 100644
--- a/ref_vqdmull.c
+++ b/ref_vqdmull.c
@@ -46,7 +46,8 @@ FNNAME (INSN)
 		   VECT_VAR(vector2, T1, W, N));	\
   vst1q_##T2##W2(VECT_VAR(result, T1, W2, N),		\
 		 VECT_VAR(vector_res, T1, W2, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W),	\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQDMULL1(INSN, T1, T2, W, W2, N)	\
diff --git a/ref_vqdmull_lane.c b/ref_vqdmull_lane.c
index b77fbd0..ae50a2e 100644
--- a/ref_vqdmull_lane.c
+++ b/ref_vqdmull_lane.c
@@ -49,7 +49,8 @@ FNNAME (INSN)
 			L);						\
   vst1q_##T2##W2(VECT_VAR(result, T1, W2, N),				\
 		 VECT_VAR(vector_res, T1, W2, N));			\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##_lane_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##_lane_##T2##W),		\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQDMULL_LANE1(INSN, T1, T2, W, W2, N, L)	\
diff --git a/ref_vqdmull_n.c b/ref_vqdmull_n.c
index 7a149cc..e41593d 100644
--- a/ref_vqdmull_n.c
+++ b/ref_vqdmull_n.c
@@ -48,7 +48,8 @@ FNNAME (INSN)
 		     L);				\
   vst1q_##T2##W2(VECT_VAR(result, T1, W2, N),		\
 		 VECT_VAR(vector_res, T1, W2, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##_n_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##_n_##T2##W),	\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQDMULL_N1(INSN, T1, T2, W, W2, N, L)	\
diff --git a/ref_vqmovn.c b/ref_vqmovn.c
index 838aa46..303aba7 100644
--- a/ref_vqmovn.c
+++ b/ref_vqmovn.c
@@ -46,7 +46,8 @@ FNNAME (INSN_NAME)
     INSN##_##T2##W2(VECT_VAR(vector, T1, W2, N));			\
   vst1##_##T2##W(VECT_VAR(result, T1, W, N),				\
 		 VECT_VAR(vector_res, T1, W, N));			\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W2))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W2),			\
+		     xSTR(T1), W, N)
 
 #define TEST_UNARY_OP(INSN, T1, T2, W, W2, N)				\
   TEST_UNARY_OP1(INSN, T1, T2, W, W2, N)				\
diff --git a/ref_vqmovun.c b/ref_vqmovun.c
index 42326ee..292d116 100644
--- a/ref_vqmovun.c
+++ b/ref_vqmovun.c
@@ -46,7 +46,8 @@ FNNAME (INSN_NAME)
     INSN##_s##W2(VECT_VAR(vector, int, W2, N));				\
   vst1##_##T2##W(VECT_VAR(result, T1, W, N),				\
 		 VECT_VAR(vector_res, T1, W, N));			\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##_s##W2))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##_s##W2),			\
+		     xSTR(T1), W, N)
 
 #define TEST_UNARY_OP(INSN, T1, T2, W, W2, N)	\
   TEST_UNARY_OP1(INSN, T1, T2, W, W2, N)	\
diff --git a/ref_vqrdmulh.c b/ref_vqrdmulh.c
index cf878aa..4d9ced3 100644
--- a/ref_vqrdmulh.c
+++ b/ref_vqrdmulh.c
@@ -47,7 +47,8 @@ FNNAME (INSN)
 		      VECT_VAR(vector2, T1, W, N));	\
   vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N),		\
 		    VECT_VAR(vector_res, T1, W, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_##T2##W),	\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQRDMULH1(INSN, Q, T1, T2, W, N)	\
diff --git a/ref_vqrdmulh_lane.c b/ref_vqrdmulh_lane.c
index 4a713aa..93890c4 100644
--- a/ref_vqrdmulh_lane.c
+++ b/ref_vqrdmulh_lane.c
@@ -48,7 +48,8 @@ FNNAME (INSN)
 			   L);					\
   vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N),			\
 		    VECT_VAR(vector_res, T1, W, N));		\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_lane_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_lane_##T2##W),	\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQRDMULH_LANE1(INSN, Q, T1, T2, W, N, N2, L)	\
diff --git a/ref_vqrdmulh_n.c b/ref_vqrdmulh_n.c
index eb79f56..3b7e8d8 100644
--- a/ref_vqrdmulh_n.c
+++ b/ref_vqrdmulh_n.c
@@ -42,14 +42,15 @@ FNNAME (INSN)
   int i;
 
   /* vector_res = vqrdmulh_n(vector,val), then store the result.  */
-#define TEST_VQRDMULH_N2(INSN, Q, T1, T2, W, N, L)	\
-  Set_Neon_Overflow(0);					\
-  VECT_VAR(vector_res, T1, W, N) =			\
-    INSN##Q##_n_##T2##W(VECT_VAR(vector, T1, W, N),	\
-			L);				\
-  vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N),		\
-		    VECT_VAR(vector_res, T1, W, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_n_##T2##W))
+#define TEST_VQRDMULH_N2(INSN, Q, T1, T2, W, N, L)		\
+  Set_Neon_Overflow(0);						\
+  VECT_VAR(vector_res, T1, W, N) =				\
+    INSN##Q##_n_##T2##W(VECT_VAR(vector, T1, W, N),		\
+			L);					\
+  vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N),			\
+		    VECT_VAR(vector_res, T1, W, N));		\
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_n_##T2##W),	\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQRDMULH_N1(INSN, Q, T1, T2, W, N, L)	\
diff --git a/ref_vqrshl.c b/ref_vqrshl.c
index 3af92b7..9e09619 100644
--- a/ref_vqrshl.c
+++ b/ref_vqrshl.c
@@ -47,7 +47,8 @@ FNNAME (INSN)
 		      VECT_VAR(vector_shift, T3, W, N));	\
   vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N),			\
 		    VECT_VAR(vector_res, T1, W, N));		\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_##T2##W),		\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQRSHL1(INSN, T3, Q, T1, T2, W, N)	\
diff --git a/ref_vqrshrn_n.c b/ref_vqrshrn_n.c
index 21765e5..c73da4c 100644
--- a/ref_vqrshrn_n.c
+++ b/ref_vqrshrn_n.c
@@ -47,7 +47,8 @@ FNNAME (INSN)
 		   V);					\
   vst1_##T2##W2(VECT_VAR(result, T1, W2, N),		\
 		VECT_VAR(vector_res, T1, W2, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W),	\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQRSHRN_N1(INSN, T1, T2, W, W2, N, V)	\
diff --git a/ref_vqrshrun_n.c b/ref_vqrshrun_n.c
index a80b396..c525f4a 100644
--- a/ref_vqrshrun_n.c
+++ b/ref_vqrshrun_n.c
@@ -47,7 +47,8 @@ FNNAME (INSN)
 		   V);					\
   vst1_u##W2(VECT_VAR(result, uint, W2, N),		\
 	     VECT_VAR(vector_res, uint, W2, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W),	\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQRSHRUN_N1(INSN, T1, T2, W, W2, N, V)	\
diff --git a/ref_vqshl.c b/ref_vqshl.c
index 277c2f4..7f9c02b 100644
--- a/ref_vqshl.c
+++ b/ref_vqshl.c
@@ -47,7 +47,8 @@ FNNAME (INSN)
 		      VECT_VAR(vector_shift, T3, W, N));	\
   vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N),			\
 		    VECT_VAR(vector_res, T1, W, N));		\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_##T2##W),		\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQSHL1(INSN, T3, Q, T1, T2, W, N)	\
diff --git a/ref_vqshl_n.c b/ref_vqshl_n.c
index 7f56980..731c4ca 100644
--- a/ref_vqshl_n.c
+++ b/ref_vqshl_n.c
@@ -40,14 +40,15 @@ THE SOFTWARE.
 FNNAME (INSN)
 {
   /* Basic test: v2=vqshl_n(v1,v), then store the result.  */
-#define TEST_VQSHL_N2(INSN, Q, T1, T2, W, N, V)		\
-  Set_Neon_Overflow(0);					\
-  VECT_VAR(vector_res, T1, W, N) =			\
-    INSN##Q##_n_##T2##W(VECT_VAR(vector, T1, W, N),	\
-			V);				\
-  vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N),		\
-		    VECT_VAR(vector_res, T1, W, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_n_##T2##W))
+#define TEST_VQSHL_N2(INSN, Q, T1, T2, W, N, V)			\
+  Set_Neon_Overflow(0);						\
+  VECT_VAR(vector_res, T1, W, N) =				\
+    INSN##Q##_n_##T2##W(VECT_VAR(vector, T1, W, N),		\
+			V);					\
+  vst1##Q##_##T2##W(VECT_VAR(result, T1, W, N),			\
+		    VECT_VAR(vector_res, T1, W, N));		\
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_n_##T2##W),	\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQSHL_N1(INSN, T3, Q, T1, T2, W, N)	\
diff --git a/ref_vqshlu_n.c b/ref_vqshlu_n.c
index 1f3f1b3..c8b14a6 100644
--- a/ref_vqshlu_n.c
+++ b/ref_vqshlu_n.c
@@ -47,7 +47,8 @@ FNNAME (INSN)
 			V);					\
   vst1##Q##_##T4##W(VECT_VAR(result, T3, W, N),			\
 		    VECT_VAR(vector_res, T3, W, N));		\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_n_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##Q##_n_##T2##W),	\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQSHLU_N1(INSN, Q, T1, T2, T3, T4, W, N, V)	\
diff --git a/ref_vqshrn_n.c b/ref_vqshrn_n.c
index 8c0ccf8..21c080c 100644
--- a/ref_vqshrn_n.c
+++ b/ref_vqshrn_n.c
@@ -47,7 +47,8 @@ FNNAME (INSN)
 		   V);					\
   vst1_##T2##W2(VECT_VAR(result, T1, W2, N),		\
 		VECT_VAR(vector_res, T1, W2, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W),	\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQSHRN_N1(INSN, T1, T2, W, W2, N, V)	\
diff --git a/ref_vqshrun_n.c b/ref_vqshrun_n.c
index 151fbec..bdf47bb 100644
--- a/ref_vqshrun_n.c
+++ b/ref_vqshrun_n.c
@@ -47,7 +47,8 @@ FNNAME (INSN)
 		   V);					\
   vst1_u##W2(VECT_VAR(result, uint, W2, N),		\
 	     VECT_VAR(vector_res, uint, W2, N));	\
-  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W))
+  dump_neon_overflow(TEST_MSG, xSTR(INSN##_##T2##W),	\
+		     xSTR(T1), W, N)
 
   /* Two auxliary macros are necessary to expand INSN */
 #define TEST_VQSHRUN_N1(INSN, T1, T2, W, W2, N, V)	\
diff --git a/stm-arm-neon-ref.h b/stm-arm-neon-ref.h
index 3ea931f..ba29f10 100644
--- a/stm-arm-neon-ref.h
+++ b/stm-arm-neon-ref.h
@@ -82,7 +82,8 @@ static int result_idx = 0;
     {									\
       fprintf(ref_file, "%" FMT ", ", VECT_VAR(result, T, W, N)[i]);	\
     }									\
-  fprintf(ref_file, " }\n");
+  fprintf(ref_file, " }\n");						\
+  DUMP4GCC(MSG,T,W,N,FMT);
 
 #define DUMP_FP(MSG,T,W,N,FMT)						\
   fprintf(ref_file, "%s:%d:%s [] = { ", MSG, result_idx++,		\
@@ -96,7 +97,46 @@ static int result_idx = 0;
       tmp.f = VECT_VAR(result, T, W, N)[i];				\
       fprintf(ref_file, "%" FMT " %.7a %.7g, ", tmp.i, tmp.f, tmp.f);	\
     }									\
-  fprintf(ref_file, " }\n");
+  fprintf(ref_file, " }\n");						\
+  DUMP4GCC_FP(MSG,T,W,N,FMT);
+
+#define DUMP4GCC(MSG,T,W,N,FMT)						\
+  fprintf(gcc_tests_file, "VECT_VAR_DECL(expected,%s,%d,%d) [] = { ",	\
+	  STR(T), W, N);						\
+  for(i=0; i<(N-1) ; i++)						\
+    {									\
+      if (W < 32) {							\
+	uint32_t tmp = (uint##W##_t) VECT_VAR(result, T, W, N)[i];	\
+	fprintf(gcc_tests_file, "0x%" FMT ", ", tmp);			\
+      } else {								\
+	fprintf(gcc_tests_file, "0x%" FMT ", ", VECT_VAR(result, T, W, N)[i]); \
+      }									\
+    }									\
+  if (W < 32) {								\
+    uint32_t tmp = (uint##W##_t) VECT_VAR(result, T, W, N)[i];		\
+    fprintf(gcc_tests_file, "0x%" FMT, tmp);				\
+  } else {								\
+    fprintf(gcc_tests_file, "0x%" FMT, VECT_VAR(result, T, W, N)[i]);	\
+  }									\
+  fprintf(gcc_tests_file, " };\n");
+
+#define DUMP4GCC_FP(MSG,T,W,N,FMT)					\
+  {									\
+    union fp_operand {							\
+      uint##W##_t i;							\
+      float##W##_t f;							\
+    } tmp;								\
+    fprintf(gcc_tests_file, "VECT_VAR_DECL(expected,%s,%d,%d) [] = { ",	\
+	    "hfloat", W, N);						\
+    for(i=0; i<(N-1) ; i++)						\
+      {									\
+	tmp.f = VECT_VAR(result, T, W, N)[i];				\
+	fprintf(gcc_tests_file, "0x%" FMT ", ", tmp.i);			\
+      }									\
+    tmp.f = VECT_VAR(result, T, W, N)[i];				\
+    fprintf(gcc_tests_file, "0x%" FMT, tmp.i);				\
+    fprintf(gcc_tests_file, " };\n");					\
+  }
 
 /* ARMCC has internal knowledge of half-precision type. Define this
    alias to avoid having to duplicate declarations.  */
@@ -115,6 +155,19 @@ static int result_idx = 0;
     }									\
   fprintf(ref_file, " }\n");
 
+#define DUMP4GCC_FP16(MSG,T,W,N,FMT)					\
+  if (0) {								\
+    fprintf(gcc_tests_file, "%s:%d:%s [] = { ", MSG, result_idx++,	\
+	    STR(VECT_VAR(result, T, W, N)));				\
+    for(i=0; i<N ; i++)							\
+      {									\
+	uint##W##_t tmp;						\
+	tmp = (uint##W##_t)VECT_VAR(result, T, W, N)[i];		\
+	fprintf(gcc_tests_file, "%" FMT ", ", tmp);			\
+      }									\
+    fprintf(gcc_tests_file, " }\n");					\
+  }
+
 #define CLEAN_PATTERN_8  0x33
 #define CLEAN_PATTERN_16 0x3333
 #define CLEAN_PATTERN_32 0x33333333
@@ -146,6 +199,7 @@ static int result_idx = 0;
 /* Generic declarations: */
 extern FILE* log_file;
 extern FILE* ref_file;
+extern FILE* gcc_tests_file;
 
 /* Input buffers, one of each size */
 extern ARRAY(buffer, int, 8, 8);
@@ -383,6 +437,7 @@ static void dump_results (char *test_name)
   int i;
 
   fprintf(ref_file, "\n%s output:\n", test_name);
+  fprintf(gcc_tests_file, "\n%s output:\n", test_name);
 
   DUMP(test_name, int, 8, 8, PRId8);
   DUMP(test_name, int, 16, 4, PRId16);
@@ -421,6 +476,7 @@ static void dump_results_hex2 (const char *test_name, const char* comment)
   int i;
 
   fprintf(ref_file, "\n%s%s output:\n", test_name, comment);
+  fprintf(gcc_tests_file, "\n%s%s output:\n", test_name, comment);
 
   DUMP(test_name, int, 8, 8, PRIx8);
   DUMP(test_name, int, 16, 4, PRIx16);
@@ -529,10 +585,13 @@ static void __set_neon_overflow(int x) {
 
 #endif /* STM_ARM_NEON_MODELS */
 
-static void dump_neon_overflow(const char* msg, const char *name)
+static void dump_neon_overflow(const char* msg, const char *name,
+			       const char* t1, int w, int n)
 {
   fprintf(ref_file, "%s:%d:%s Neon overflow %d\n", msg, result_idx++,
 	  name, Neon_Overflow);
+  fprintf(gcc_tests_file, "int VECT_VAR(expected_overflow,%s,%d,%d) = %d;\n", \
+	  t1, w, n, Neon_Overflow);
 }
 
 /* Clean output buffers before execution */