[X86][SSE] Match all-of bool scalar reductions into a bitcast/movmsk + cmp.

Same as what we do for vector reductions in combineHorizontalPredicateResult, use movmsk+cmp for scalar (and(extract(x,0),extract(x,1)) reduction patterns. 

llvm-svn: 361052

2 files changed, 30 insertions, 16 deletions

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index a52d0faa35b..743e23977e5 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -37841,6 +37841,24 @@ static SDValue combineAnd(SDNode *N, SelectionDAG &DAG,
   if (SDValue V = combineParity(N, DAG, Subtarget))
     return V;
 
+  // Match all-of bool scalar reductions into a bitcast/movmsk + cmp.
+  // TODO: Support multiple SrcOps.
+  if (VT == MVT::i1) {
+    SmallVector<SDValue, 2> SrcOps;
+    if (matchBitOpReduction(SDValue(N, 0), ISD::AND, SrcOps) &&
+        SrcOps.size() == 1) {
+      SDLoc dl(N);
+      unsigned NumElts = SrcOps[0].getValueType().getVectorNumElements();
+      EVT MaskVT = EVT::getIntegerVT(*DAG.getContext(), NumElts);
+      SDValue Mask = combineBitcastvxi1(DAG, MaskVT, SrcOps[0], dl, Subtarget);
+      if (Mask) {
+        APInt AllBits = APInt::getAllOnesValue(NumElts);
+        return DAG.getSetCC(dl, MVT::i1, Mask,
+                            DAG.getConstant(AllBits, dl, MaskVT), ISD::SETEQ);
+      }
+    }
+  }
+
   if (DCI.isBeforeLegalizeOps())
     return SDValue();
 
diff --git a/llvm/test/CodeGen/X86/movmsk-cmp.ll b/llvm/test/CodeGen/X86/movmsk-cmp.ll
index c8b3488af16..eeb523c3800 100644
--- a/llvm/test/CodeGen/X86/movmsk-cmp.ll
+++ b/llvm/test/CodeGen/X86/movmsk-cmp.ll
@@ -4487,10 +4487,9 @@ define i1 @movmsk_v2i64(<2 x i64> %x, <2 x i64> %y) {
 ; SSE2-NEXT:    pand %xmm0, %xmm1
 ; SSE2-NEXT:    pcmpeqd %xmm0, %xmm0
 ; SSE2-NEXT:    pxor %xmm1, %xmm0
-; SSE2-NEXT:    movmskpd %xmm0, %ecx
-; SSE2-NEXT:    movl %ecx, %eax
-; SSE2-NEXT:    shrb %al
-; SSE2-NEXT:    andb %cl, %al
+; SSE2-NEXT:    movmskpd %xmm0, %eax
+; SSE2-NEXT:    cmpb $3, %al
+; SSE2-NEXT:    sete %al
 ; SSE2-NEXT:    retq
 ;
 ; AVX-LABEL: movmsk_v2i64:
@@ -4498,10 +4497,9 @@ define i1 @movmsk_v2i64(<2 x i64> %x, <2 x i64> %y) {
 ; AVX-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
 ; AVX-NEXT:    vpxor %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vmovmskpd %xmm0, %ecx
-; AVX-NEXT:    movl %ecx, %eax
-; AVX-NEXT:    shrb %al
-; AVX-NEXT:    andb %cl, %al
+; AVX-NEXT:    vmovmskpd %xmm0, %eax
+; AVX-NEXT:    cmpb $3, %al
+; AVX-NEXT:    sete %al
 ; AVX-NEXT:    retq
 ;
 ; KNL-LABEL: movmsk_v2i64:
@@ -4596,19 +4594,17 @@ define i1 @movmsk_v2f64(<2 x double> %x, <2 x double> %y) {
 ; SSE2-LABEL: movmsk_v2f64:
 ; SSE2:       # %bb.0:
 ; SSE2-NEXT:    cmplepd %xmm0, %xmm1
-; SSE2-NEXT:    movmskpd %xmm1, %ecx
-; SSE2-NEXT:    movl %ecx, %eax
-; SSE2-NEXT:    shrb %al
-; SSE2-NEXT:    andb %cl, %al
+; SSE2-NEXT:    movmskpd %xmm1, %eax
+; SSE2-NEXT:    cmpb $3, %al
+; SSE2-NEXT:    sete %al
 ; SSE2-NEXT:    retq
 ;
 ; AVX-LABEL: movmsk_v2f64:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vcmplepd %xmm0, %xmm1, %xmm0
-; AVX-NEXT:    vmovmskpd %xmm0, %ecx
-; AVX-NEXT:    movl %ecx, %eax
-; AVX-NEXT:    shrb %al
-; AVX-NEXT:    andb %cl, %al
+; AVX-NEXT:    vmovmskpd %xmm0, %eax
+; AVX-NEXT:    cmpb $3, %al
+; AVX-NEXT:    sete %al
 ; AVX-NEXT:    retq
 ;
 ; KNL-LABEL: movmsk_v2f64: