AMDGPU: Add a fast path for icmp.i1(src, false, NE)
Summary:
This allows moving the condition from the intrinsic to the standard ICmp
opcode, so that LLVM can do simplifications on it. The icmp.i1 intrinsic
is an identity for retrieving the SGPR mask.
And we can also get the mask from and i1, or i1, xor i1.
Reviewers: arsenm, nhaehnle
Subscribers: kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits
Differential Revision: https://reviews.llvm.org/D52060
llvm-svn: 351150
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 9e16737..0ba9216 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -5355,6 +5355,11 @@
Denominator, Numerator);
}
case Intrinsic::amdgcn_icmp: {
+ // There is a Pat that handles this variant, so return it as-is.
+ if (Op.getOperand(1).getValueType() == MVT::i1 &&
+ Op.getConstantOperandVal(2) == 0 &&
+ Op.getConstantOperandVal(3) == ICmpInst::Predicate::ICMP_NE)
+ return Op;
return lowerICMPIntrinsic(*this, Op.getNode(), DAG);
}
case Intrinsic::amdgcn_fcmp: {