re PR target/90991 (_mm_loadu_ps instrinsic translates to vmovaps in combination with _mm512_insertf32x4)

	PR target/90991
	* config/i386/sse.md
	(*<extract_type>_vinsert<shuffletype><extract_suf>_0): Use vmovupd,
	vmovups, vmovdqu, vmovdqu32 or vmovdqu64 instead of the aligned
	insns if operands[2] is misaligned_operand.

	* gcc.target/i386/avx512dq-pr90991-1.c: New test.

From-SVN: r272674

4 files changed, 80 insertions, 6 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index f8d3574..bd65d2a 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,11 @@
+2019-06-26  Jakub Jelinek  <jakub@redhat.com>
+
+	PR target/90991
+	* config/i386/sse.md
+	(*<extract_type>_vinsert<shuffletype><extract_suf>_0): Use vmovupd,
+	vmovups, vmovdqu, vmovdqu32 or vmovdqu64 instead of the aligned
+	insns if operands[2] is misaligned_operand.
+
 2019-06-26  Li Jia He  <helijia@linux.ibm.com>
 
 	* config/rs6000/rs6000.h (TARGET_MADDLD): Remove the restriction of
diff --git a/gcc/config/i386/sse.md b/gcc/config/i386/sse.md
index abf7d98..29f16bc 100644
--- a/gcc/config/i386/sse.md
+++ b/gcc/config/i386/sse.md
@@ -13747,15 +13747,29 @@
   switch (<MODE>mode)
     {
     case E_V8DFmode:
-      return "vmovapd\t{%2, %x0|%x0, %2}";
+      if (misaligned_operand (operands[2], <ssequartermode>mode))
+	return "vmovupd\t{%2, %x0|%x0, %2}";
+      else
+	return "vmovapd\t{%2, %x0|%x0, %2}";
     case E_V16SFmode:
-      return "vmovaps\t{%2, %x0|%x0, %2}";
+      if (misaligned_operand (operands[2], <ssequartermode>mode))
+	return "vmovups\t{%2, %x0|%x0, %2}";
+      else
+	return "vmovaps\t{%2, %x0|%x0, %2}";
     case E_V8DImode:
-      return which_alternative == 2 ? "vmovdqa64\t{%2, %x0|%x0, %2}"
-				    : "vmovdqa\t{%2, %x0|%x0, %2}";
+      if (misaligned_operand (operands[2], <ssequartermode>mode))
+	return which_alternative == 2 ? "vmovdqu64\t{%2, %x0|%x0, %2}"
+				      : "vmovdqu\t{%2, %x0|%x0, %2}";
+      else
+	return which_alternative == 2 ? "vmovdqa64\t{%2, %x0|%x0, %2}"
+				      : "vmovdqa\t{%2, %x0|%x0, %2}";
     case E_V16SImode:
-      return which_alternative == 2 ? "vmovdqa32\t{%2, %x0|%x0, %2}"
-				    : "vmovdqa\t{%2, %x0|%x0, %2}";
+      if (misaligned_operand (operands[2], <ssequartermode>mode))
+	return which_alternative == 2 ? "vmovdqu32\t{%2, %x0|%x0, %2}"
+				      : "vmovdqu\t{%2, %x0|%x0, %2}";
+      else
+	return which_alternative == 2 ? "vmovdqa32\t{%2, %x0|%x0, %2}"
+				      : "vmovdqa\t{%2, %x0|%x0, %2}";
     default:
       gcc_unreachable ();
     }
diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index 5b54f28..a77eb7e 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,8 @@
+2019-06-26  Jakub Jelinek  <jakub@redhat.com>
+
+	PR target/90991
+	* gcc.target/i386/avx512dq-pr90991-1.c: New test.
+
 2019-06-26  Li Jia He  <helijia@linux.ibm.com>
 
 	* gcc.target/powerpc/maddld-1.c: New testcase.
diff --git a/gcc/testsuite/gcc.target/i386/avx512dq-pr90991-1.c b/gcc/testsuite/gcc.target/i386/avx512dq-pr90991-1.c
new file mode 100644
index 0000000..6c96812
--- /dev/null
+++ b/gcc/testsuite/gcc.target/i386/avx512dq-pr90991-1.c
@@ -0,0 +1,47 @@
+/* PR target/90991 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -mavx512dq -masm=att" } */
+/* { dg-final { scan-assembler-times "vmovaps\[ \t]\+\\(\[^\n\r]*\\), %xmm0" 1 } } */
+/* { dg-final { scan-assembler-times "vmovapd\[ \t]\+\\(\[^\n\r]*\\), %xmm0" 1 } } */
+/* { dg-final { scan-assembler-times "vmovdqa\[ \t]\+\\(\[^\n\r]*\\), %xmm0" 1 } } */
+/* { dg-final { scan-assembler-times "vmovups\[ \t]\+\\(\[^\n\r]*\\), %xmm0" 1 } } */
+/* { dg-final { scan-assembler-times "vmovupd\[ \t]\+\\(\[^\n\r]*\\), %xmm0" 1 } } */
+/* { dg-final { scan-assembler-times "vmovdqu\[ \t]\+\\(\[^\n\r]*\\), %xmm0" 1 } } */
+
+#include <x86intrin.h>
+
+__m512
+f1 (void *a)
+{
+  return _mm512_insertf32x4 (_mm512_set1_ps (0.0f), _mm_load_ps (a), 0);
+}
+
+__m512d
+f2 (void *a)
+{
+  return _mm512_insertf64x2 (_mm512_set1_pd (0.0), _mm_load_pd (a), 0);
+}
+
+__m512i
+f3 (void *a)
+{
+  return _mm512_inserti32x4 (_mm512_set1_epi32 (0), _mm_load_si128 (a), 0);
+}
+
+__m512
+f4 (void *a)
+{
+  return _mm512_insertf32x4 (_mm512_set1_ps (0.0f), _mm_loadu_ps (a), 0);
+}
+
+__m512d
+f5 (void *a)
+{
+  return _mm512_insertf64x2 (_mm512_set1_pd (0.0), _mm_loadu_pd (a), 0);
+}
+
+__m512i
+f6 (void *a)
+{
+  return _mm512_inserti32x4 (_mm512_set1_epi32 (0), _mm_loadu_si128 (a), 0);
+}