summaryrefslogtreecommitdiffstats
path: root/llvm/lib/Transforms
diff options
context:
space:
mode:
authorSanjay Patel <spatel@rotateright.com>2019-09-02 13:33:20 +0000
committerSanjay Patel <spatel@rotateright.com>2019-09-02 13:33:20 +0000
commit561c39994b95ec6c930850ff556b161d9d27af37 (patch)
tree0158962bf0a199a5aa8ba25f4f5cdea4c261089f /llvm/lib/Transforms
parenteafede2afeb45c826957d49c7895a04655f1f08d (diff)
downloadbcm5719-llvm-561c39994b95ec6c930850ff556b161d9d27af37.tar.gz
bcm5719-llvm-561c39994b95ec6c930850ff556b161d9d27af37.zip
[InstCombine] recognize bswap disguised as shufflevector
bitcast <N x i8> (shuf X, undef, <N, N-1,...0>) to i{N*8} --> bswap (bitcast X to i{N*8}) In PR43146: https://bugs.llvm.org/show_bug.cgi?id=43146 ...we have a more complicated case where SLP is making a mess of bswap. This patch won't do anything for that currently, but we need to improve bswap recognition in instcombine, SLP, and/or a standalone pass to avoid that problem. This is limited using the data-layout so we don't try to do this transform with actual vector types. The backend does not appear to have folds to convert in either direction, so we don't want to mess up something that is actually better lowered as a shuffle. On x86, we're trading something like this: vmovd %edi, %xmm0 vpshufb LCPI0_0(%rip), %xmm0, %xmm0 ## xmm0 = xmm0[3,2,1,0,u,u,u,u,u,u,u,u,u,u,u,u] vmovd %xmm0, %eax For: movl %edi, %eax bswapl %eax Differential Revision: https://reviews.llvm.org/D66965 llvm-svn: 370659
Diffstat (limited to 'llvm/lib/Transforms')
-rw-r--r--llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp16
1 files changed, 16 insertions, 0 deletions
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
index 4be44dac4ca..1a569d2638f 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
@@ -2416,6 +2416,22 @@ Instruction *InstCombiner::visitBitCast(BitCastInst &CI) {
return new ShuffleVectorInst(LHS, RHS, Shuf->getOperand(2));
}
}
+
+ // A bitcasted-to-scalar and byte-reversing shuffle is better recognized as
+ // a byte-swap:
+ // bitcast <N x i8> (shuf X, undef, <N, N-1,...0>) --> bswap (bitcast X)
+ // TODO: We should match the related pattern for bitreverse.
+ if (DestTy->isIntegerTy() &&
+ DL.isLegalInteger(DestTy->getScalarSizeInBits()) &&
+ SrcTy->getScalarSizeInBits() == 8 && NumShufElts % 2 == 0 &&
+ Shuf->hasOneUse() && Shuf->isReverse()) {
+ assert(ShufOp0->getType() == SrcTy && "Unexpected shuffle mask");
+ assert(isa<UndefValue>(ShufOp1) && "Unexpected shuffle op");
+ Function *Bswap =
+ Intrinsic::getDeclaration(CI.getModule(), Intrinsic::bswap, DestTy);
+ Value *ScalarX = Builder.CreateBitCast(ShufOp0, DestTy);
+ return IntrinsicInst::Create(Bswap, { ScalarX });
+ }
}
// Handle the A->B->A cast, and there is an intervening PHI node.
OpenPOWER on IntegriCloud