[llvm] r302641 - [DAGCombiner] Add vector support to fold (shl/srl 0, x) -> 0

[Simon Pilgrim via llvm-commits]

Author: rksimon
Date: Wed May 10 07:34:27 2017
New Revision: 302641

URL: http://llvm.org/viewvc/llvm-project?rev=302641&view=rev
Log:
[DAGCombiner] Add vector support to fold (shl/srl 0, x) -> 0

Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
    llvm/trunk/test/CodeGen/X86/combine-shl.ll
    llvm/trunk/test/CodeGen/X86/combine-srl.ll

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp?rev=302641&r1=302640&r2=302641&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp Wed May 10 07:34:27 2017
@@ -5331,7 +5331,7 @@ SDValue DAGCombiner::visitSHL(SDNode *N)
   if (N0C && N1C && !N1C->isOpaque())
     return DAG.FoldConstantArithmetic(ISD::SHL, SDLoc(N), VT, N0C, N1C);
   // fold (shl 0, x) -> 0
-  if (isNullConstant(N0))
+  if (isNullConstantOrNullSplatConstant(N0))
     return N0;
   // fold (shl x, c >= size(x)) -> undef
   if (N1C && N1C->getAPIntValue().uge(OpSizeInBits))
@@ -5714,7 +5714,7 @@ SDValue DAGCombiner::visitSRL(SDNode *N)
   if (N0C && N1C && !N1C->isOpaque())
     return DAG.FoldConstantArithmetic(ISD::SRL, SDLoc(N), VT, N0C, N1C);
   // fold (srl 0, x) -> 0
-  if (isNullConstant(N0))
+  if (isNullConstantOrNullSplatConstant(N0))
     return N0;
   // fold (srl x, c >= size(x)) -> undef
   if (N1C && N1C->getAPIntValue().uge(OpSizeInBits))

Modified: llvm/trunk/test/CodeGen/X86/combine-shl.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/combine-shl.ll?rev=302641&r1=302640&r2=302641&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/combine-shl.ll (original)
+++ llvm/trunk/test/CodeGen/X86/combine-shl.ll Wed May 10 07:34:27 2017
@@ -11,8 +11,7 @@ define <4 x i32> @combine_vec_shl_zero(<
 ;
 ; AVX-LABEL: combine_vec_shl_zero:
 ; AVX:       # BB#0:
-; AVX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX-NEXT:    vpsllvd %xmm0, %xmm1, %xmm0
+; AVX-NEXT:    vxorps %xmm0, %xmm0, %xmm0
 ; AVX-NEXT:    retq
   %1 = shl <4 x i32> zeroinitializer, %x
   ret <4 x i32> %1

Modified: llvm/trunk/test/CodeGen/X86/combine-srl.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/combine-srl.ll?rev=302641&r1=302640&r2=302641&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/combine-srl.ll (original)
+++ llvm/trunk/test/CodeGen/X86/combine-srl.ll Wed May 10 07:34:27 2017
@@ -6,30 +6,12 @@
 define <4 x i32> @combine_vec_lshr_zero(<4 x i32> %x) {
 ; SSE-LABEL: combine_vec_lshr_zero:
 ; SSE:       # BB#0:
-; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    psrldq {{.*#+}} xmm2 = xmm2[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    pxor %xmm3, %xmm3
-; SSE-NEXT:    psrld %xmm2, %xmm3
-; SSE-NEXT:    movdqa %xmm0, %xmm2
-; SSE-NEXT:    psrlq $32, %xmm2
-; SSE-NEXT:    pxor %xmm4, %xmm4
-; SSE-NEXT:    psrld %xmm2, %xmm4
-; SSE-NEXT:    pblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm3[4,5,6,7]
-; SSE-NEXT:    pmovzxdq {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero
-; SSE-NEXT:    punpckhdq {{.*#+}} xmm0 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; SSE-NEXT:    pxor %xmm3, %xmm3
-; SSE-NEXT:    psrld %xmm0, %xmm3
-; SSE-NEXT:    psrld %xmm2, %xmm1
-; SSE-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm3[4,5,6,7]
-; SSE-NEXT:    pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,3],xmm1[4,5],xmm4[6,7]
-; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    xorps %xmm0, %xmm0
 ; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: combine_vec_lshr_zero:
 ; AVX:       # BB#0:
-; AVX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX-NEXT:    vpsrlvd %xmm0, %xmm1, %xmm0
+; AVX-NEXT:    vxorps %xmm0, %xmm0, %xmm0
 ; AVX-NEXT:    retq
   %1 = lshr <4 x i32> zeroinitializer, %x
   ret <4 x i32> %1