mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2025-01-29 13:32:33 +00:00
[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) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@204160 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Andrea Di Biagio
parent
7e18b99f1e
commit
6077ca9abb
@ -2518,35 +2518,66 @@ SDValue DAGCombiner::SimplifyBinOpWithSameOpcodeHands(SDNode *N) {
|
|||||||
// The type-legalizer generates this pattern when loading illegal
|
// The type-legalizer generates this pattern when loading illegal
|
||||||
// vector types from memory. In many cases this allows additional shuffle
|
// vector types from memory. In many cases this allows additional shuffle
|
||||||
// optimizations.
|
// optimizations.
|
||||||
if (N0.getOpcode() == ISD::VECTOR_SHUFFLE && Level < AfterLegalizeDAG &&
|
// There are other cases where moving the shuffle after the xor/and/or
|
||||||
N0.getOperand(1).getOpcode() == ISD::UNDEF &&
|
// is profitable even if shuffles don't perform a swizzle.
|
||||||
N1.getOperand(1).getOpcode() == ISD::UNDEF) {
|
// If both shuffles use the same mask, and both shuffles have the same first
|
||||||
|
// or second operand, then it might still be profitable to move the shuffle
|
||||||
|
// after the xor/and/or operation.
|
||||||
|
if (N0.getOpcode() == ISD::VECTOR_SHUFFLE && Level < AfterLegalizeDAG) {
|
||||||
ShuffleVectorSDNode *SVN0 = cast<ShuffleVectorSDNode>(N0);
|
ShuffleVectorSDNode *SVN0 = cast<ShuffleVectorSDNode>(N0);
|
||||||
ShuffleVectorSDNode *SVN1 = cast<ShuffleVectorSDNode>(N1);
|
ShuffleVectorSDNode *SVN1 = cast<ShuffleVectorSDNode>(N1);
|
||||||
|
|
||||||
assert(N0.getOperand(0).getValueType() == N1.getOperand(1).getValueType() &&
|
assert(N0.getOperand(0).getValueType() == N1.getOperand(0).getValueType() &&
|
||||||
"Inputs to shuffles are not the same type");
|
"Inputs to shuffles are not the same type");
|
||||||
|
|
||||||
unsigned NumElts = VT.getVectorNumElements();
|
|
||||||
|
|
||||||
// Check that both shuffles use the same mask. The masks are known to be of
|
// Check that both shuffles use the same mask. The masks are known to be of
|
||||||
// the same length because the result vector type is the same.
|
// the same length because the result vector type is the same.
|
||||||
bool SameMask = true;
|
// Check also that shuffles have only one use to avoid introducing extra
|
||||||
for (unsigned i = 0; i != NumElts; ++i) {
|
// instructions.
|
||||||
int Idx0 = SVN0->getMaskElt(i);
|
if (SVN0->hasOneUse() && SVN1->hasOneUse() &&
|
||||||
int Idx1 = SVN1->getMaskElt(i);
|
SVN0->getMask().equals(SVN1->getMask())) {
|
||||||
if (Idx0 != Idx1) {
|
SDValue ShOp = N0->getOperand(1);
|
||||||
SameMask = false;
|
|
||||||
break;
|
|
||||||
}
|
|
||||||
}
|
|
||||||
|
|
||||||
if (SameMask) {
|
// Don't try to fold this node if it requires introducing a
|
||||||
SDValue Op = DAG.getNode(N->getOpcode(), SDLoc(N), VT,
|
// build vector of all zeros that might be illegal at this stage.
|
||||||
N0.getOperand(0), N1.getOperand(0));
|
if (N->getOpcode() == ISD::XOR && ShOp.getOpcode() != ISD::UNDEF) {
|
||||||
AddToWorkList(Op.getNode());
|
if (!LegalTypes)
|
||||||
return DAG.getVectorShuffle(VT, SDLoc(N), Op,
|
ShOp = DAG.getConstant(0, VT);
|
||||||
DAG.getUNDEF(VT), &SVN0->getMask()[0]);
|
else
|
||||||
|
ShOp = SDValue();
|
||||||
|
}
|
||||||
|
|
||||||
|
// (AND (shuf (A, C), shuf (B, C)) -> shuf (AND (A, B), C)
|
||||||
|
// (OR (shuf (A, C), shuf (B, C)) -> shuf (OR (A, B), C)
|
||||||
|
// (XOR (shuf (A, C), shuf (B, C)) -> shuf (XOR (A, B), V_0)
|
||||||
|
if (N0.getOperand(1) == N1.getOperand(1) && ShOp.getNode()) {
|
||||||
|
SDValue NewNode = DAG.getNode(N->getOpcode(), SDLoc(N), VT,
|
||||||
|
N0->getOperand(0), N1->getOperand(0));
|
||||||
|
AddToWorkList(NewNode.getNode());
|
||||||
|
return DAG.getVectorShuffle(VT, SDLoc(N), NewNode, ShOp,
|
||||||
|
&SVN0->getMask()[0]);
|
||||||
|
}
|
||||||
|
|
||||||
|
// Don't try to fold this node if it requires introducing a
|
||||||
|
// build vector of all zeros that might be illegal at this stage.
|
||||||
|
ShOp = N0->getOperand(0);
|
||||||
|
if (N->getOpcode() == ISD::XOR && ShOp.getOpcode() != ISD::UNDEF) {
|
||||||
|
if (!LegalTypes)
|
||||||
|
ShOp = DAG.getConstant(0, VT);
|
||||||
|
else
|
||||||
|
ShOp = SDValue();
|
||||||
|
}
|
||||||
|
|
||||||
|
// (AND (shuf (C, A), shuf (C, B)) -> shuf (C, AND (A, B))
|
||||||
|
// (OR (shuf (C, A), shuf (C, B)) -> shuf (C, OR (A, B))
|
||||||
|
// (XOR (shuf (C, A), shuf (C, B)) -> shuf (V_0, XOR (A, B))
|
||||||
|
if (N0->getOperand(0) == N1->getOperand(0) && ShOp.getNode()) {
|
||||||
|
SDValue NewNode = DAG.getNode(N->getOpcode(), SDLoc(N), VT,
|
||||||
|
N0->getOperand(1), N1->getOperand(1));
|
||||||
|
AddToWorkList(NewNode.getNode());
|
||||||
|
return DAG.getVectorShuffle(VT, SDLoc(N), ShOp, NewNode,
|
||||||
|
&SVN0->getMask()[0]);
|
||||||
|
}
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|||||||
@ -251,6 +251,7 @@ define <2 x i64> @test20(<2 x i64> %a, <2 x i64> %b) {
|
|||||||
; CHECK-LABEL: test20
|
; CHECK-LABEL: test20
|
||||||
; CHECK-NOT: xorps
|
; CHECK-NOT: xorps
|
||||||
; CHECK: orps
|
; CHECK: orps
|
||||||
|
; CHECK-NEXT: movq
|
||||||
; CHECK-NEXT: ret
|
; CHECK-NEXT: ret
|
||||||
|
|
||||||
|
|
||||||
@ -262,6 +263,7 @@ define <2 x i64> @test21(<2 x i64> %a, <2 x i64> %b) {
|
|||||||
}
|
}
|
||||||
; CHECK-LABEL: test21
|
; CHECK-LABEL: test21
|
||||||
; CHECK: por
|
; CHECK: por
|
||||||
|
; CHECK-NEXT: pslldq
|
||||||
; CHECK-NEXT: ret
|
; CHECK-NEXT: ret
|
||||||
|
|
||||||
|
|
||||||
|
|||||||
253
test/CodeGen/X86/combine-vec-shuffle.ll
Normal file
253
test/CodeGen/X86/combine-vec-shuffle.ll
Normal file
@ -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
|
||||||
|
|
||||||
Loading…
x
Reference in New Issue
Block a user