diff options
Diffstat (limited to 'llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp')
| -rw-r--r-- | llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp | 24 |
1 files changed, 24 insertions, 0 deletions
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp index 45cacc73d63..a193dde1c39 100644 --- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp +++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp @@ -690,6 +690,30 @@ Value *InstCombiner::SimplifyDemandedUseBits(Value *V, APInt DemandedMask, // TODO: Could compute known zero/one bits based on the input. break; } + case Intrinsic::fshr: + case Intrinsic::fshl: { + const APInt *SA; + if (!match(I->getOperand(2), m_APInt(SA))) + break; + + // Normalize to funnel shift left. APInt shifts of BitWidth are well- + // defined, so no need to special-case zero shifts here. + uint64_t ShiftAmt = SA->urem(BitWidth); + if (II->getIntrinsicID() == Intrinsic::fshr) + ShiftAmt = BitWidth - ShiftAmt; + + APInt DemandedMaskLHS(DemandedMask.lshr(ShiftAmt)); + APInt DemandedMaskRHS(DemandedMask.shl(BitWidth - ShiftAmt)); + if (SimplifyDemandedBits(I, 0, DemandedMaskLHS, LHSKnown, Depth + 1) || + SimplifyDemandedBits(I, 1, DemandedMaskRHS, RHSKnown, Depth + 1)) + return I; + + Known.Zero = LHSKnown.Zero.shl(ShiftAmt) | + RHSKnown.Zero.lshr(BitWidth - ShiftAmt); + Known.One = LHSKnown.One.shl(ShiftAmt) | + RHSKnown.One.lshr(BitWidth - ShiftAmt); + break; + } case Intrinsic::x86_mmx_pmovmskb: case Intrinsic::x86_sse_movmsk_ps: case Intrinsic::x86_sse2_movmsk_pd: |
