[X86][SSE] Fold extract_subvector(extend(x)) -> extend_vector_inreg(x)

This adds any extend support - folding to zero_extend_vector_inreg (PMOVZX) for legality

Minor improvement for PR39709

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@359608 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 1ed59bd7535b420f80561ea5c1bb3d22d5ff8f91..eedf024f6802b2558cf82c2d75b07ccc79cfab61 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -42765,15 +42765,17 @@ static SDValue combineExtractSubvector(SDNode *N, SelectionDAG &DAG,
         return DAG.getNode(X86ISD::VFPEXT, SDLoc(N), VT, InVec.getOperand(0));
       }
     }
-    if ((InOpcode == ISD::ZERO_EXTEND || InOpcode == ISD::SIGN_EXTEND) &&
+    if ((InOpcode == ISD::ANY_EXTEND || InOpcode == ISD::ZERO_EXTEND ||
+         InOpcode == ISD::SIGN_EXTEND) &&
         VT.is128BitVector() &&
         InVec.getOperand(0).getSimpleValueType().is128BitVector()) {
-      unsigned ExtOp =
-          InOpcode == ISD::ZERO_EXTEND ? ISD::ZERO_EXTEND_VECTOR_INREG
-                                       : ISD::SIGN_EXTEND_VECTOR_INREG;
+      unsigned ExtOp = InOpcode == ISD::SIGN_EXTEND
+                           ? ISD::SIGN_EXTEND_VECTOR_INREG
+                           : ISD::ZERO_EXTEND_VECTOR_INREG;
       return DAG.getNode(ExtOp, SDLoc(N), VT, InVec.getOperand(0));
     }
-    if ((InOpcode == ISD::ZERO_EXTEND_VECTOR_INREG ||
+    if ((InOpcode == ISD::ANY_EXTEND_VECTOR_INREG ||
+         InOpcode == ISD::ZERO_EXTEND_VECTOR_INREG ||
          InOpcode == ISD::SIGN_EXTEND_VECTOR_INREG) &&
         VT.is128BitVector() &&
         InVec.getOperand(0).getSimpleValueType().is128BitVector()) {

diff --git a/test/CodeGen/X86/vector-reduce-mul-widen.ll b/test/CodeGen/X86/vector-reduce-mul-widen.ll
index 413336f06b5b27ed76e432773666954a15319caa..3cb46460acf64e41b4238d35445b077b3b209fc2 100644 (file)
--- a/test/CodeGen/X86/vector-reduce-mul-widen.ll
+++ b/test/CodeGen/X86/vector-reduce-mul-widen.ll
@@ -1835,28 +1835,27 @@ define i8 @test_v16i8(<16 x i8> %a0) {
 ; AVX2-LABEL: test_v16i8:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
+; AVX2-NEXT:    vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX2-NEXT:    vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
 ; AVX2-NEXT:    vpmullw %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
 ; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpackuswb %xmm0, %xmm0, %xmm2
 ; AVX2-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,2,3]
-; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero
+; AVX2-NEXT:    vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
 ; AVX2-NEXT:    vpmullw %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpackuswb %xmm0, %xmm0, %xmm2
 ; AVX2-NEXT:    vpsrld $16, %xmm2, %xmm2
-; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero
+; AVX2-NEXT:    vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
 ; AVX2-NEXT:    vpmullw %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpackuswb %xmm0, %xmm0, %xmm1
 ; AVX2-NEXT:    vpsrlw $8, %xmm1, %xmm1
-; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
+; AVX2-NEXT:    vpmovzxbw {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero
 ; AVX2-NEXT:    vpmullw %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpextrb $0, %xmm0, %eax
 ; AVX2-NEXT:    # kill: def $al killed $al killed $eax
-; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512BW-LABEL: test_v16i8:

diff --git a/test/CodeGen/X86/vector-reduce-mul.ll b/test/CodeGen/X86/vector-reduce-mul.ll
index f5ce68807cebabcfc2b3f3791093c9d3171caef1..fd067a6e0fe7ddddc863cdc8ed9f26d4d8353c7a 100644 (file)
--- a/test/CodeGen/X86/vector-reduce-mul.ll
+++ b/test/CodeGen/X86/vector-reduce-mul.ll
@@ -1792,29 +1792,22 @@ define i8 @test_v16i8(<16 x i8> %a0) {
 ;
 ; AVX2-LABEL: test_v16i8:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
-; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
-; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
+; AVX2-NEXT:    vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
 ; AVX2-NEXT:    vpmullw %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [255,255,255,255,255,255,255,255]
 ; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpackuswb %xmm0, %xmm0, %xmm2
-; AVX2-NEXT:    vpshufd {{.*#+}} xmm2 = xmm2[1,1,2,3]
-; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[2,3,2,3]
 ; AVX2-NEXT:    vpmullw %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpackuswb %xmm0, %xmm0, %xmm2
-; AVX2-NEXT:    vpsrld $16, %xmm2, %xmm2
-; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero,xmm2[8],zero,xmm2[9],zero,xmm2[10],zero,xmm2[11],zero,xmm2[12],zero,xmm2[13],zero,xmm2[14],zero,xmm2[15],zero
+; AVX2-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[1,1,2,3]
 ; AVX2-NEXT:    vpmullw %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpackuswb %xmm0, %xmm0, %xmm1
-; AVX2-NEXT:    vpsrlw $8, %xmm1, %xmm1
-; AVX2-NEXT:    vpmovzxbw {{.*#+}} ymm1 = xmm1[0],zero,xmm1[1],zero,xmm1[2],zero,xmm1[3],zero,xmm1[4],zero,xmm1[5],zero,xmm1[6],zero,xmm1[7],zero,xmm1[8],zero,xmm1[9],zero,xmm1[10],zero,xmm1[11],zero,xmm1[12],zero,xmm1[13],zero,xmm1[14],zero,xmm1[15],zero
+; AVX2-NEXT:    vpshuflw {{.*#+}} xmm1 = xmm0[1,1,2,3,4,5,6,7]
 ; AVX2-NEXT:    vpmullw %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vpextrb $0, %xmm0, %eax
 ; AVX2-NEXT:    # kill: def $al killed $al killed $eax
-; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
 ; AVX512BW-LABEL: test_v16i8: