[DAGCombiner] teach how to simplify xor/and/or nodes according to the following rules:

 1)  (AND (shuf (A, C, Mask), shuf (B, C, Mask)) -> shuf (AND (A, B), C, Mask)
 2)  (OR  (shuf (A, C, Mask), shuf (B, C, Mask)) -> shuf (OR  (A, B), C, Mask)
 3)  (XOR (shuf (A, C, Mask), shuf (B, C, Mask)) -> shuf (XOR (A, B), V_0, Mask)

 4)  (AND (shuf (C, A, Mask), shuf (C, B, Mask)) -> shuf (C, AND (A, B), Mask)
 5)  (OR  (shuf (C, A, Mask), shuf (C, B, Mask)) -> shuf (C, OR  (A, B), Mask)
 6)  (XOR (shuf (C, A, Mask), shuf (C, B, Mask)) -> shuf (V_0, XOR (A, B), Mask)

llvm-svn: 204160

2 files changed, 255 insertions, 0 deletions

diff --git a/llvm/test/CodeGen/X86/combine-or.ll b/llvm/test/CodeGen/X86/combine-or.ll
index 60b6d756165..c1ce53334ec 100644
--- a/llvm/test/CodeGen/X86/combine-or.ll
+++ b/llvm/test/CodeGen/X86/combine-or.ll
@@ -251,6 +251,7 @@ define <2 x i64> @test20(<2 x i64> %a, <2 x i64> %b) {
 ; CHECK-LABEL: test20
 ; CHECK-NOT: xorps
 ; CHECK: orps
+; CHECK-NEXT: movq
 ; CHECK-NEXT: ret
 
 
@@ -262,6 +263,7 @@ define <2 x i64> @test21(<2 x i64> %a, <2 x i64> %b) {
 }
 ; CHECK-LABEL: test21
 ; CHECK: por
+; CHECK-NEXT: pslldq
 ; CHECK-NEXT: ret
 
 
diff --git a/llvm/test/CodeGen/X86/combine-vec-shuffle.ll b/llvm/test/CodeGen/X86/combine-vec-shuffle.ll
new file mode 100644
index 00000000000..9e6ab892713
--- /dev/null
+++ b/llvm/test/CodeGen/X86/combine-vec-shuffle.ll
@@ -0,0 +1,253 @@
+; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mcpu=corei7 | FileCheck %s
+
+; Verify that the DAGCombiner correctly folds according to the following rules:
+
+; fold (AND (shuf (A, C), shuf (B, C)) -> shuf (AND (A, B), C)
+; fold (OR  (shuf (A, C), shuf (B, C)) -> shuf (OR  (A, B), C)
+; fold (XOR (shuf (A, C), shuf (B, C)) -> shuf (XOR (A, B), V_0)
+
+; fold (AND (shuf (C, A), shuf (C, B)) -> shuf (C, AND (A, B))
+; fold (OR  (shuf (C, A), shuf (C, B)) -> shuf (C, OR  (A, B))
+; fold (XOR (shuf (C, A), shuf (C, B)) -> shuf (V_0, XOR (A, B))
+
+
+
+define <4 x i32> @test1(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %a, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 1, i32 3>
+  %shuf2 = shufflevector <4 x i32> %b, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 1, i32 3>
+  %and = and <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %and
+}
+; CHECK-LABEL: test1
+; CHECK-NOT: pshufd
+; CHECK: pand
+; CHECK-NEXT: pshufd
+; CHECK-NEXT: ret
+
+
+define <4 x i32> @test2(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %a, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 1, i32 3>
+  %shuf2 = shufflevector <4 x i32> %b, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 1, i32 3>
+  %or = or <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %or
+}
+; CHECK-LABEL: test2
+; CHECK-NOT: pshufd
+; CHECK: por
+; CHECK-NEXT: pshufd
+; CHECK-NEXT: ret
+
+
+define <4 x i32> @test3(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %a, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 1, i32 3>
+  %shuf2 = shufflevector <4 x i32> %b, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 1, i32 3>
+  %xor = xor <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %xor
+}
+; CHECK-LABEL: test3
+; CHECK-NOT: pshufd
+; CHECK: pxor
+; CHECK-NEXT: pshufd
+; CHECK-NEXT: ret
+
+
+define <4 x i32> @test4(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %c, <4 x i32> %a, <4 x i32><i32 4, i32 6, i32 5, i32 7>
+  %shuf2 = shufflevector <4 x i32> %c, <4 x i32> %b, <4 x i32><i32 4, i32 6, i32 5, i32 7>
+  %and = and <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %and
+}
+; CHECK-LABEL: test4
+; CHECK-NOT: pshufd
+; CHECK: pand
+; CHECK-NEXT: pshufd
+; CHECK-NEXT: ret
+
+
+define <4 x i32> @test5(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %c, <4 x i32> %a, <4 x i32><i32 4, i32 6, i32 5, i32 7>
+  %shuf2 = shufflevector <4 x i32> %c, <4 x i32> %b, <4 x i32><i32 4, i32 6, i32 5, i32 7>
+  %or = or <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %or
+}
+; CHECK-LABEL: test5
+; CHECK-NOT: pshufd
+; CHECK: por
+; CHECK-NEXT: pshufd
+; CHECK-NEXT: ret
+
+
+define <4 x i32> @test6(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %c, <4 x i32> %a, <4 x i32><i32 4, i32 6, i32 5, i32 7>
+  %shuf2 = shufflevector <4 x i32> %c, <4 x i32> %b, <4 x i32><i32 4, i32 6, i32 5, i32 7>
+  %xor = xor <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %xor
+}
+; CHECK-LABEL: test6
+; CHECK-NOT: pshufd
+; CHECK: pxor
+; CHECK-NEXT: pshufd
+; CHECK-NEXT: ret
+
+
+; Verify that DAGCombiner moves the shuffle after the xor/and/or even if shuffles
+; are not performing a swizzle operations.
+
+define <4 x i32> @test1b(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %a, <4 x i32> %c, <4 x i32><i32 0, i32 5, i32 2, i32 7>
+  %shuf2 = shufflevector <4 x i32> %b, <4 x i32> %c, <4 x i32><i32 0, i32 5, i32 2, i32 7>
+  %and = and <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %and
+}
+; CHECK-LABEL: test1b
+; CHECK-NOT: blendps
+; CHECK: andps
+; CHECK-NEXT: blendps
+; CHECK-NEXT: ret
+
+
+define <4 x i32> @test2b(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %a, <4 x i32> %c, <4 x i32><i32 0, i32 5, i32 2, i32 7>
+  %shuf2 = shufflevector <4 x i32> %b, <4 x i32> %c, <4 x i32><i32 0, i32 5, i32 2, i32 7>
+  %or = or <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %or
+}
+; CHECK-LABEL: test2b
+; CHECK-NOT: blendps
+; CHECK: orps
+; CHECK-NEXT: blendps
+; CHECK-NEXT: ret
+
+
+define <4 x i32> @test3b(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %a, <4 x i32> %c, <4 x i32><i32 0, i32 5, i32 2, i32 7>
+  %shuf2 = shufflevector <4 x i32> %b, <4 x i32> %c, <4 x i32><i32 0, i32 5, i32 2, i32 7>
+  %xor = xor <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %xor
+}
+; CHECK-LABEL: test3b
+; CHECK-NOT: blendps
+; CHECK: xorps
+; CHECK-NEXT: xorps
+; CHECK-NEXT: blendps
+; CHECK-NEXT: ret
+
+
+define <4 x i32> @test4b(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %c, <4 x i32> %a, <4 x i32><i32 0, i32 5, i32 2, i32 7>
+  %shuf2 = shufflevector <4 x i32> %c, <4 x i32> %b, <4 x i32><i32 0, i32 5, i32 2, i32 7>
+  %and = and <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %and
+}
+; CHECK-LABEL: test4b
+; CHECK-NOT: blendps
+; CHECK: andps
+; CHECK-NEXT: blendps
+; CHECK: ret
+
+
+define <4 x i32> @test5b(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %c, <4 x i32> %a, <4 x i32><i32 0, i32 5, i32 2, i32 7>
+  %shuf2 = shufflevector <4 x i32> %c, <4 x i32> %b, <4 x i32><i32 0, i32 5, i32 2, i32 7>
+  %or = or <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %or
+}
+; CHECK-LABEL: test5b
+; CHECK-NOT: blendps
+; CHECK: orps
+; CHECK-NEXT: blendps
+; CHECK: ret
+
+
+define <4 x i32> @test6b(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %c, <4 x i32> %a, <4 x i32><i32 0, i32 5, i32 2, i32 7>
+  %shuf2 = shufflevector <4 x i32> %c, <4 x i32> %b, <4 x i32><i32 0, i32 5, i32 2, i32 7>
+  %xor = xor <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %xor
+}
+; CHECK-LABEL: test6b
+; CHECK-NOT: blendps
+; CHECK: xorps
+; CHECK-NEXT: xorps
+; CHECK-NEXT: blendps
+; CHECK: ret
+
+define <4 x i32> @test1c(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %a, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 5, i32 7>
+  %shuf2 = shufflevector <4 x i32> %b, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 5, i32 7>
+  %and = and <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %and
+}
+; CHECK-LABEL: test1c
+; CHECK-NOT: shufps
+; CHECK: andps
+; CHECK-NEXT: shufps
+; CHECK-NEXT: ret
+
+
+define <4 x i32> @test2c(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %a, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 5, i32 7>
+  %shuf2 = shufflevector <4 x i32> %b, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 5, i32 7>
+  %or = or <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %or
+}
+; CHECK-LABEL: test2c
+; CHECK-NOT: shufps
+; CHECK: orps
+; CHECK-NEXT: shufps
+; CHECK-NEXT: ret
+
+
+define <4 x i32> @test3c(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %a, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 5, i32 7>
+  %shuf2 = shufflevector <4 x i32> %b, <4 x i32> %c, <4 x i32><i32 0, i32 2, i32 5, i32 7>
+  %xor = xor <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %xor
+}
+; CHECK-LABEL: test3c
+; CHECK-NOT: shufps
+; CHECK: xorps
+; CHECK-NEXT: xorps
+; CHECK-NEXT: shufps
+; CHECK-NEXT: ret
+
+
+define <4 x i32> @test4c(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %c, <4 x i32> %a, <4 x i32><i32 0, i32 2, i32 5, i32 7>
+  %shuf2 = shufflevector <4 x i32> %c, <4 x i32> %b, <4 x i32><i32 0, i32 2, i32 5, i32 7>
+  %and = and <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %and
+}
+; CHECK-LABEL: test4c
+; CHECK-NOT: shufps
+; CHECK: andps
+; CHECK-NEXT: shufps
+; CHECK: ret
+
+
+define <4 x i32> @test5c(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %c, <4 x i32> %a, <4 x i32><i32 0, i32 2, i32 5, i32 7>
+  %shuf2 = shufflevector <4 x i32> %c, <4 x i32> %b, <4 x i32><i32 0, i32 2, i32 5, i32 7>
+  %or = or <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %or
+}
+; CHECK-LABEL: test5c
+; CHECK-NOT: shufps
+; CHECK: orps
+; CHECK-NEXT: shufps
+; CHECK: ret
+
+
+define <4 x i32> @test6c(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) {
+  %shuf1 = shufflevector <4 x i32> %c, <4 x i32> %a, <4 x i32><i32 0, i32 2, i32 5, i32 7>
+  %shuf2 = shufflevector <4 x i32> %c, <4 x i32> %b, <4 x i32><i32 0, i32 2, i32 5, i32 7>
+  %xor = xor <4 x i32> %shuf1, %shuf2
+  ret <4 x i32> %xor
+}
+; CHECK-LABEL: test6c
+; CHECK-NOT: shufps
+; CHECK: xorps
+; CHECK-NEXT: xorps
+; CHECK-NEXT: shufps
+; CHECK: ret
+