[llvm] r270843 - [AVX512] Fix intrinsic cmp{sd|ss} lowering.

[Igor Breger via llvm-commits]

Author: ibreger
Date: Thu May 26 07:42:25 2016
New Revision: 270843

URL: http://llvm.org/viewvc/llvm-project?rev=270843&view=rev
Log:
[AVX512] Fix intrinsic cmp{sd|ss} lowering.

Differential Revision: http://reviews.llvm.org/D20615

Modified:
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
    llvm/trunk/test/CodeGen/X86/avx512-intrinsics.ll

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=270843&r1=270842&r2=270843&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Thu May 26 07:42:25 2016
@@ -17455,9 +17455,7 @@ static SDValue LowerINTRINSIC_WO_CHAIN(S
                                                                    MVT::i1),
                                              Subtarget, DAG);
 
-      return DAG.getNode(ISD::SIGN_EXTEND_INREG, dl, MVT::i8,
-                         DAG.getNode(ISD::ANY_EXTEND, dl, MVT::i8, CmpMask),
-                         DAG.getValueType(MVT::i1));
+      return DAG.getNode(ISD::ZERO_EXTEND, dl, MVT::i8, CmpMask);
     }
     case COMI: { // Comparison intrinsics
       ISD::CondCode CC = (ISD::CondCode)IntrData->Opc1;

Modified: llvm/trunk/test/CodeGen/X86/avx512-intrinsics.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx512-intrinsics.ll?rev=270843&r1=270842&r2=270843&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx512-intrinsics.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx512-intrinsics.ll Thu May 26 07:42:25 2016
@@ -5659,8 +5659,7 @@ define i8 at test_int_x86_avx512_mask_cmp_s
 ; CHECK-NEXT:    kmovw %edi, %k1
 ; CHECK-NEXT:    vcmpnltsd {sae}, %xmm1, %xmm0, %k0 {%k1}
 ; CHECK-NEXT:    kmovw %k0, %eax
-; CHECK-NEXT:    shlb $7, %al
-; CHECK-NEXT:    sarb $7, %al
+; CHECK-NEXT:    andl $1, %eax
 ; CHECK-NEXT:    retq
 
   %res4 = call i8 @llvm.x86.avx512.mask.cmp.sd(<2 x double> %x0, <2 x double> %x1, i32 5, i8 %x3, i32 8)
@@ -5681,8 +5680,7 @@ define i8 at test_int_x86_avx512_mask_cmp_s
 ; CHECK-NEXT:    kandw %k2, %k1, %k1
 ; CHECK-NEXT:    korw %k1, %k0, %k0
 ; CHECK-NEXT:    kmovw %k0, %eax
-; CHECK-NEXT:    shlb $7, %al
-; CHECK-NEXT:    sarb $7, %al
+; CHECK-NEXT:    andl $1, %eax
 ; CHECK-NEXT:    retq
 
   %res1 = call i8 @llvm.x86.avx512.mask.cmp.sd(<2 x double> %x0, <2 x double> %x1, i32 2, i8 -1, i32 4)
@@ -5705,8 +5703,7 @@ define i8 at test_int_x86_avx512_mask_cmp_s
 ; CHECK-NEXT:    kmovw %edi, %k1
 ; CHECK-NEXT:    vcmpunordss %xmm1, %xmm0, %k0 {%k1}
 ; CHECK-NEXT:    kmovw %k0, %eax
-; CHECK-NEXT:    shlb $7, %al
-; CHECK-NEXT:    sarb $7, %al
+; CHECK-NEXT:    andl $1, %eax
 ; CHECK-NEXT:    retq
 
   %res2 = call i8 @llvm.x86.avx512.mask.cmp.ss(<4 x float> %x0, <4 x float> %x1, i32 3, i8 %x3, i32 4)
@@ -5719,15 +5716,16 @@ define i8 at test_int_x86_avx512_mask_cmp_s
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vcmpless %xmm1, %xmm0, %k1
 ; CHECK-NEXT:    vcmpunordss {sae}, %xmm1, %xmm0, %k0 {%k1}
-; CHECK-NEXT:    vcmpneqss %xmm1, %xmm0, %k1
-; CHECK-NEXT:    vcmpnltss {sae}, %xmm1, %xmm0, %k1 {%k1}
 ; CHECK-NEXT:    andl $1, %edi
-; CHECK-NEXT:    kmovw %edi, %k2
-; CHECK-NEXT:    kandw %k2, %k1, %k1
-; CHECK-NEXT:    kandw %k1, %k0, %k0
+; CHECK-NEXT:    kmovw %edi, %k1
+; CHECK-NEXT:    vcmpneqss %xmm1, %xmm0, %k2 {%k1}
+; CHECK-NEXT:    kmovw %k2, %ecx
+; CHECK-NEXT:    vcmpnltss {sae}, %xmm1, %xmm0, %k1 {%k1}
+; CHECK-NEXT:    kmovw %k1, %edx
+; CHECK-NEXT:    andl $1, %edx
 ; CHECK-NEXT:    kmovw %k0, %eax
-; CHECK-NEXT:    shlb $7, %al
-; CHECK-NEXT:    sarb $7, %al
+; CHECK-NEXT:    andb %cl, %al
+; CHECK-NEXT:    andb %dl, %al
 ; CHECK-NEXT:    retq
   %res1 = call i8 @llvm.x86.avx512.mask.cmp.ss(<4 x float> %x0, <4 x float> %x1, i32 2, i8 -1, i32 4)
   %res2 = call i8 @llvm.x86.avx512.mask.cmp.ss(<4 x float> %x0, <4 x float> %x1, i32 3, i8 -1, i32 8)