[llvm] r288578 - [X86] Fix VEX encoded VPMADDUBSW to not be marked commutable.

[Craig Topper via llvm-commits]

Author: ctopper
Date: Fri Dec  2 23:35:44 2016
New Revision: 288578

URL: http://llvm.org/viewvc/llvm-project?rev=288578&view=rev
Log:
[X86] Fix VEX encoded VPMADDUBSW to not be marked commutable.

This was accidentallly broken in r285515 when we started lowering the intrinsic to an ISD node. Should fix PR31241.

Modified:
    llvm/trunk/lib/Target/X86/X86InstrSSE.td
    llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll
    llvm/trunk/test/CodeGen/X86/avx2-intrinsics-x86.ll

Modified: llvm/trunk/lib/Target/X86/X86InstrSSE.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrSSE.td?rev=288578&r1=288577&r2=288578&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrSSE.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrSSE.td Fri Dec  2 23:35:44 2016
@@ -5496,13 +5496,13 @@ let isCommutable = 0 in {
   defm VPSHUFB    : SS3I_binop_rm<0x00, "vpshufb", X86pshufb, v16i8, v16i8,
                                   VR128, loadv2i64, i128mem,
                                   SSE_PSHUFB, 0>, VEX_4V;
+  defm VPMADDUBSW : SS3I_binop_rm<0x04, "vpmaddubsw", X86vpmaddubsw, v8i16,
+                                  v16i8, VR128, loadv2i64, i128mem,
+                                  SSE_PMADD, 0>, VEX_4V;
 }
 defm VPMULHRSW    : SS3I_binop_rm<0x0B, "vpmulhrsw", X86mulhrs, v8i16, v8i16,
                                   VR128, loadv2i64, i128mem,
                                   SSE_PMULHRSW, 0>, VEX_4V;
-defm VPMADDUBSW : SS3I_binop_rm<0x04, "vpmaddubsw", X86vpmaddubsw, v8i16,
-                                v16i8, VR128, loadv2i64, i128mem,
-                                SSE_PMADD, 0>, VEX_4V;
 }
 
 let ImmT = NoImm, Predicates = [HasAVX] in {
@@ -5542,13 +5542,13 @@ let isCommutable = 0 in {
   defm VPSHUFBY   : SS3I_binop_rm<0x00, "vpshufb", X86pshufb, v32i8, v32i8,
                                   VR256, loadv4i64, i256mem,
                                   SSE_PSHUFB, 0>, VEX_4V, VEX_L;
+  defm VPMADDUBSWY : SS3I_binop_rm<0x04, "vpmaddubsw", X86vpmaddubsw, v16i16,
+                                   v32i8, VR256, loadv4i64, i256mem,
+                                   SSE_PMADD, 0>, VEX_4V, VEX_L;
 }
 defm VPMULHRSWY   : SS3I_binop_rm<0x0B, "vpmulhrsw", X86mulhrs, v16i16, v16i16,
                                   VR256, loadv4i64, i256mem,
                                   SSE_PMULHRSW, 0>, VEX_4V, VEX_L;
-defm VPMADDUBSWY : SS3I_binop_rm<0x04, "vpmaddubsw", X86vpmaddubsw, v16i16,
-                                 v32i8, VR256, loadv4i64, i256mem,
-                                 SSE_PMADD, 0>, VEX_4V, VEX_L;
 }
 
 let ImmT = NoImm, Predicates = [HasAVX2] in {

Modified: llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll?rev=288578&r1=288577&r2=288578&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll Fri Dec  2 23:35:44 2016
@@ -2545,7 +2545,8 @@ define <8 x i16> @test_x86_ssse3_pmadd_u
 ; AVX-LABEL: test_x86_ssse3_pmadd_ub_sw_128_load_op0:
 ; AVX:       ## BB#0:
 ; AVX-NEXT:    movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04]
-; AVX-NEXT:    vpmaddubsw (%eax), %xmm0, %xmm0 ## encoding: [0xc4,0xe2,0x79,0x04,0x00]
+; AVX-NEXT:    vmovdqa (%eax), %xmm1 ## encoding: [0xc5,0xf9,0x6f,0x08]
+; AVX-NEXT:    vpmaddubsw %xmm0, %xmm1, %xmm0 ## encoding: [0xc4,0xe2,0x71,0x04,0xc0]
 ; AVX-NEXT:    retl ## encoding: [0xc3]
 ;
 ; AVX512VL-LABEL: test_x86_ssse3_pmadd_ub_sw_128_load_op0:

Modified: llvm/trunk/test/CodeGen/X86/avx2-intrinsics-x86.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx2-intrinsics-x86.ll?rev=288578&r1=288577&r2=288578&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx2-intrinsics-x86.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx2-intrinsics-x86.ll Fri Dec  2 23:35:44 2016
@@ -761,7 +761,8 @@ define <16 x i16> @test_x86_avx2_pmadd_u
 ; AVX2-LABEL: test_x86_avx2_pmadd_ub_sw_load_op0:
 ; AVX2:       ## BB#0:
 ; AVX2-NEXT:    movl {{[0-9]+}}(%esp), %eax ## encoding: [0x8b,0x44,0x24,0x04]
-; AVX2-NEXT:    vpmaddubsw (%eax), %ymm0, %ymm0 ## encoding: [0xc4,0xe2,0x7d,0x04,0x00]
+; AVX2-NEXT:    vmovdqa (%eax), %ymm1 ## encoding: [0xc5,0xfd,0x6f,0x08]
+; AVX2-NEXT:    vpmaddubsw %ymm0, %ymm1, %ymm0 ## encoding: [0xc4,0xe2,0x75,0x04,0xc0]
 ; AVX2-NEXT:    retl ## encoding: [0xc3]
 ;
 ; AVX512VL-LABEL: test_x86_avx2_pmadd_ub_sw_load_op0: