[DAGCombine] Generalize distributeTruncateThroughAnd to work with any non-opaque...

llvm-svn: 284574

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index aa63889..6a047ac 100644 (file)
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -4445,19 +4445,15 @@ SDValue DAGCombiner::distributeTruncateThroughAnd(SDNode *N) {
   // (truncate:TruncVT (and N00, N01C)) -> (and (truncate:TruncVT N00), TruncC)
   if (N->hasOneUse() && N->getOperand(0).hasOneUse()) {
     SDValue N01 = N->getOperand(0).getOperand(1);
-
-    if (ConstantSDNode *N01C = isConstOrConstSplat(N01)) {
-      if (!N01C->isOpaque()) {
-        EVT TruncVT = N->getValueType(0);
-        SDValue N00 = N->getOperand(0).getOperand(0);
-        APInt TruncC = N01C->getAPIntValue();
-        TruncC = TruncC.trunc(TruncVT.getScalarSizeInBits());
-        SDLoc DL(N);
-
-        return DAG.getNode(ISD::AND, DL, TruncVT,
-                           DAG.getNode(ISD::TRUNCATE, DL, TruncVT, N00),
-                           DAG.getConstant(TruncC, DL, TruncVT));
-      }
+    if (isConstantOrConstantVector(N01, /* NoOpaques */ true)) {
+      SDLoc DL(N);
+      EVT TruncVT = N->getValueType(0);
+      SDValue N00 = N->getOperand(0).getOperand(0);
+      SDValue Trunc00 = DAG.getNode(ISD::TRUNCATE, DL, TruncVT, N00);
+      SDValue Trunc01 = DAG.getNode(ISD::TRUNCATE, DL, TruncVT, N01);
+      AddToWorklist(Trunc00.getNode());
+      AddToWorklist(Trunc01.getNode());
+      return DAG.getNode(ISD::AND, DL, TruncVT, Trunc00, Trunc01);
     }
   }
 

diff --git a/llvm/test/CodeGen/X86/combine-shl.ll b/llvm/test/CodeGen/X86/combine-shl.ll
index 7d97060..ff5d4f0 100644 (file)
--- a/llvm/test/CodeGen/X86/combine-shl.ll
+++ b/llvm/test/CodeGen/X86/combine-shl.ll
@@ -98,11 +98,10 @@ define <4 x i32> @combine_vec_shl_known_zero1(<4 x i32> %x) {
 define <4 x i32> @combine_vec_shl_trunc_and(<4 x i32> %x, <4 x i64> %y) {
 ; SSE-LABEL: combine_vec_shl_trunc_and:
 ; SSE:       # BB#0:
-; SSE-NEXT:    pand {{.*}}(%rip), %xmm1
-; SSE-NEXT:    pand {{.*}}(%rip), %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; SSE-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5,6,7]
+; SSE-NEXT:    pand {{.*}}(%rip), %xmm1
 ; SSE-NEXT:    pslld $23, %xmm1
 ; SSE-NEXT:    paddd {{.*}}(%rip), %xmm1
 ; SSE-NEXT:    cvttps2dq %xmm1, %xmm1
@@ -111,9 +110,9 @@ define <4 x i32> @combine_vec_shl_trunc_and(<4 x i32> %x, <4 x i64> %y) {
 ;
 ; AVX-LABEL: combine_vec_shl_trunc_and:
 ; AVX:       # BB#0:
-; AVX-NEXT:    vpand {{.*}}(%rip), %ymm1, %ymm1
 ; AVX-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
 ; AVX-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
 ; AVX-NEXT:    vpsllvd %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq

diff --git a/llvm/test/CodeGen/X86/combine-sra.ll b/llvm/test/CodeGen/X86/combine-sra.ll
index 570eb7e..bfa5b9e 100644 (file)
--- a/llvm/test/CodeGen/X86/combine-sra.ll
+++ b/llvm/test/CodeGen/X86/combine-sra.ll
@@ -161,11 +161,10 @@ define <4 x i32> @combine_vec_ashr_ashr2(<4 x i32> %x) {
 define <4 x i32> @combine_vec_ashr_trunc_and(<4 x i32> %x, <4 x i64> %y) {
 ; SSE-LABEL: combine_vec_ashr_trunc_and:
 ; SSE:       # BB#0:
-; SSE-NEXT:    pand {{.*}}(%rip), %xmm1
-; SSE-NEXT:    pand {{.*}}(%rip), %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; SSE-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5,6,7]
+; SSE-NEXT:    pand {{.*}}(%rip), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
 ; SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
@@ -187,9 +186,9 @@ define <4 x i32> @combine_vec_ashr_trunc_and(<4 x i32> %x, <4 x i64> %y) {
 ;
 ; AVX-LABEL: combine_vec_ashr_trunc_and:
 ; AVX:       # BB#0:
-; AVX-NEXT:    vpand {{.*}}(%rip), %ymm1, %ymm1
 ; AVX-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
 ; AVX-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
 ; AVX-NEXT:    vpsravd %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq

diff --git a/llvm/test/CodeGen/X86/combine-srl.ll b/llvm/test/CodeGen/X86/combine-srl.ll
index 8dfdb91..2bbe779 100644 (file)
--- a/llvm/test/CodeGen/X86/combine-srl.ll
+++ b/llvm/test/CodeGen/X86/combine-srl.ll
@@ -507,11 +507,10 @@ declare <4 x i32> @llvm.ctlz.v4i32(<4 x i32>, i1)
 define <4 x i32> @combine_vec_lshr_trunc_and(<4 x i32> %x, <4 x i64> %y) {
 ; SSE-LABEL: combine_vec_lshr_trunc_and:
 ; SSE:       # BB#0:
-; SSE-NEXT:    pand {{.*}}(%rip), %xmm1
-; SSE-NEXT:    pand {{.*}}(%rip), %xmm2
 ; SSE-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[0,1,0,2]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
 ; SSE-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm2[4,5,6,7]
+; SSE-NEXT:    pand {{.*}}(%rip), %xmm1
 ; SSE-NEXT:    movdqa %xmm1, %xmm2
 ; SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
 ; SSE-NEXT:    movdqa %xmm0, %xmm3
@@ -533,9 +532,9 @@ define <4 x i32> @combine_vec_lshr_trunc_and(<4 x i32> %x, <4 x i64> %y) {
 ;
 ; AVX-LABEL: combine_vec_lshr_trunc_and:
 ; AVX:       # BB#0:
-; AVX-NEXT:    vpand {{.*}}(%rip), %ymm1, %ymm1
 ; AVX-NEXT:    vpshufd {{.*#+}} ymm1 = ymm1[0,2,2,3,4,6,6,7]
 ; AVX-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
 ; AVX-NEXT:    vpsrlvd %xmm1, %xmm0, %xmm0
 ; AVX-NEXT:    vzeroupper
 ; AVX-NEXT:    retq