[llvm] 30211c4 - [X86] combineANDXORWithAllOnesIntoANDNP - add BROADCAST handling

[Simon Pilgrim via llvm-commits]

Author: Simon Pilgrim
Date: 2020-04-30T16:24:17+01:00
New Revision: 30211c47837ec7bd332e6da159647855e23a6aef

URL: https://github.com/llvm/llvm-project/commit/30211c47837ec7bd332e6da159647855e23a6aef
DIFF: https://github.com/llvm/llvm-project/commit/30211c47837ec7bd332e6da159647855e23a6aef.diff

LOG: [X86] combineANDXORWithAllOnesIntoANDNP - add BROADCAST handling

Fold BROADCAST(NOT(Y)) -> BROADCAST(Y) as part of finding a NOT inversion.

Added: 
    

Modified: 
    llvm/lib/Target/X86/X86ISelLowering.cpp
    llvm/test/CodeGen/X86/combine-bitselect.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 71515d960cd7..78015544c232 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -41249,10 +41249,27 @@ static SDValue combineANDXORWithAllOnesIntoANDNP(SDNode *N, SelectionDAG &DAG) {
   SDValue N0 = N->getOperand(0);
   SDValue N1 = N->getOperand(1);
 
-  if (SDValue Not = IsNOT(N0, DAG)) {
+  auto GetNot = [&VT, &DAG](SDValue V) {
+    // Basic X = NOT(Y) detection.
+    if (SDValue Not = IsNOT(V, DAG))
+      return Not;
+    // Fold BROADCAST(NOT(Y)) -> BROADCAST(Y).
+    if (V.getOpcode() == X86ISD::VBROADCAST) {
+      SDValue Src = V.getOperand(0);
+      EVT SrcVT = Src.getValueType();
+      if (!SrcVT.isVector())
+        return SDValue();
+      if (SDValue Not = IsNOT(Src, DAG))
+        return DAG.getNode(X86ISD::VBROADCAST, SDLoc(V), VT,
+                           DAG.getBitcast(SrcVT, Not));
+    }
+    return SDValue();
+  };
+
+  if (SDValue Not = GetNot(N0)) {
     X = Not;
     Y = N1;
-  } else if (SDValue Not = IsNOT(N1, DAG)) {
+  } else if (SDValue Not = GetNot(N1)) {
     X = Not;
     Y = N0;
   } else

diff  --git a/llvm/test/CodeGen/X86/combine-bitselect.ll b/llvm/test/CodeGen/X86/combine-bitselect.ll
index 5945376ec9da..1fc18357ef7c 100644
--- a/llvm/test/CodeGen/X86/combine-bitselect.ll
+++ b/llvm/test/CodeGen/X86/combine-bitselect.ll
@@ -318,24 +318,18 @@ define <4 x i64> @bitselect_v4i64_broadcast_rrr(<4 x i64> %a0, <4 x i64> %a1, i6
 ; AVX2-LABEL: bitselect_v4i64_broadcast_rrr:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vmovq %rdi, %xmm2
-; AVX2-NEXT:    vpbroadcastq %xmm2, %ymm3
-; AVX2-NEXT:    vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX2-NEXT:    vpxor %xmm4, %xmm2, %xmm2
 ; AVX2-NEXT:    vpbroadcastq %xmm2, %ymm2
-; AVX2-NEXT:    vpand %ymm3, %ymm0, %ymm0
-; AVX2-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpandn %ymm1, %ymm2, %ymm1
 ; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: bitselect_v4i64_broadcast_rrr:
 ; AVX512F:       # %bb.0:
 ; AVX512F-NEXT:    vmovq %rdi, %xmm2
-; AVX512F-NEXT:    vmovq %rdi, %xmm3
-; AVX512F-NEXT:    vpbroadcastq %xmm3, %ymm3
-; AVX512F-NEXT:    vpternlogq $15, %zmm2, %zmm2, %zmm2
 ; AVX512F-NEXT:    vpbroadcastq %xmm2, %ymm2
-; AVX512F-NEXT:    vpand %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpandn %ymm1, %ymm2, %ymm1
 ; AVX512F-NEXT:    vpor %ymm1, %ymm0, %ymm0
 ; AVX512F-NEXT:    retq
   %1 = insertelement <4 x i64> undef, i64 %a2, i32 0
@@ -394,27 +388,18 @@ define <4 x i64> @bitselect_v4i64_broadcast_rrm(<4 x i64> %a0, <4 x i64> %a1, i6
 ;
 ; AVX2-LABEL: bitselect_v4i64_broadcast_rrm:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vmovq {{.*#+}} xmm2 = mem[0],zero
-; AVX2-NEXT:    vpbroadcastq %xmm2, %ymm3
-; AVX2-NEXT:    vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX2-NEXT:    vpxor %xmm4, %xmm2, %xmm2
-; AVX2-NEXT:    vpbroadcastq %xmm2, %ymm2
-; AVX2-NEXT:    vpand %ymm3, %ymm0, %ymm0
-; AVX2-NEXT:    vpand %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpor %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vbroadcastsd (%rdi), %ymm2
+; AVX2-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vandnps %ymm1, %ymm2, %ymm1
+; AVX2-NEXT:    vorps %ymm1, %ymm0, %ymm0
 ; AVX2-NEXT:    retq
 ;
 ; AVX512F-LABEL: bitselect_v4i64_broadcast_rrm:
 ; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    movq (%rdi), %rax
-; AVX512F-NEXT:    vmovq %rax, %xmm2
-; AVX512F-NEXT:    vmovq %rax, %xmm3
-; AVX512F-NEXT:    vpbroadcastq %xmm3, %ymm3
-; AVX512F-NEXT:    vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT:    vpbroadcastq %xmm2, %ymm2
-; AVX512F-NEXT:    vpand %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
-; AVX512F-NEXT:    vpor %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vbroadcastsd (%rdi), %ymm2
+; AVX512F-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vandnps %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vorps %ymm1, %ymm0, %ymm0
 ; AVX512F-NEXT:    retq
   %a2 = load i64, i64* %p2
   %1 = insertelement <4 x i64> undef, i64 %a2, i32 0