diff options
Diffstat (limited to 'llvm/lib')
| -rw-r--r-- | llvm/lib/IR/AutoUpgrade.cpp | 26 | ||||
| -rw-r--r-- | llvm/lib/Target/X86/X86IntrinsicsInfo.h | 12 |
2 files changed, 26 insertions, 12 deletions
diff --git a/llvm/lib/IR/AutoUpgrade.cpp b/llvm/lib/IR/AutoUpgrade.cpp index 79a8d18126d..96c34c0ad8c 100644 --- a/llvm/lib/IR/AutoUpgrade.cpp +++ b/llvm/lib/IR/AutoUpgrade.cpp @@ -256,6 +256,7 @@ static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) { Name.startswith("avx512.mask.pshuf.d.") || Name.startswith("avx512.mask.pshufl.w.") || Name.startswith("avx512.mask.pshufh.w.") || + Name.startswith("avx512.mask.shuf.p") || Name.startswith("avx512.mask.vpermil.p") || Name.startswith("avx512.mask.perm.df.") || Name.startswith("avx512.mask.perm.di.") || @@ -1149,6 +1150,31 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) { if (CI->getNumArgOperands() == 4) Rep = EmitX86Select(Builder, CI->getArgOperand(3), Rep, CI->getArgOperand(2)); + } else if (IsX86 && Name.startswith("avx512.mask.shuf.p")) { + Value *Op0 = CI->getArgOperand(0); + Value *Op1 = CI->getArgOperand(1); + unsigned Imm = cast<ConstantInt>(CI->getArgOperand(2))->getZExtValue(); + unsigned NumElts = CI->getType()->getVectorNumElements(); + + unsigned NumLaneElts = 128/CI->getType()->getScalarSizeInBits(); + unsigned HalfLaneElts = NumLaneElts / 2; + + SmallVector<uint32_t, 16> Idxs(NumElts); + for (unsigned i = 0; i != NumElts; ++i) { + // Base index is the starting element of the lane. + Idxs[i] = i - (i % NumLaneElts); + // If we are half way through the lane switch to the other source. + if ((i % NumLaneElts) >= HalfLaneElts) + Idxs[i] += NumElts; + // Now select the specific element. By adding HalfLaneElts bits from + // the immediate. Wrapping around the immediate every 8-bits. + Idxs[i] += (Imm >> ((i * HalfLaneElts) % 8)) & ((1 << HalfLaneElts) - 1); + } + + Rep = Builder.CreateShuffleVector(Op0, Op1, Idxs); + + Rep = EmitX86Select(Builder, CI->getArgOperand(4), Rep, + CI->getArgOperand(3)); } else if (IsX86 && (Name.startswith("avx512.mask.movddup") || Name.startswith("avx512.mask.movshdup") || Name.startswith("avx512.mask.movsldup"))) { diff --git a/llvm/lib/Target/X86/X86IntrinsicsInfo.h b/llvm/lib/Target/X86/X86IntrinsicsInfo.h index a883bc4c5d6..189fb1aec17 100644 --- a/llvm/lib/Target/X86/X86IntrinsicsInfo.h +++ b/llvm/lib/Target/X86/X86IntrinsicsInfo.h @@ -1352,18 +1352,6 @@ static const IntrinsicData IntrinsicsWithoutChain[] = { X86ISD::SHUF128, 0), X86_INTRINSIC_DATA(avx512_mask_shuf_i64x2_256, INTR_TYPE_3OP_IMM8_MASK, X86ISD::SHUF128, 0), - X86_INTRINSIC_DATA(avx512_mask_shuf_pd_128, INTR_TYPE_3OP_IMM8_MASK, - X86ISD::SHUFP, 0), - X86_INTRINSIC_DATA(avx512_mask_shuf_pd_256, INTR_TYPE_3OP_IMM8_MASK, - X86ISD::SHUFP, 0), - X86_INTRINSIC_DATA(avx512_mask_shuf_pd_512, INTR_TYPE_3OP_IMM8_MASK, - X86ISD::SHUFP, 0), - X86_INTRINSIC_DATA(avx512_mask_shuf_ps_128, INTR_TYPE_3OP_IMM8_MASK, - X86ISD::SHUFP, 0), - X86_INTRINSIC_DATA(avx512_mask_shuf_ps_256, INTR_TYPE_3OP_IMM8_MASK, - X86ISD::SHUFP, 0), - X86_INTRINSIC_DATA(avx512_mask_shuf_ps_512, INTR_TYPE_3OP_IMM8_MASK, - X86ISD::SHUFP, 0), X86_INTRINSIC_DATA(avx512_mask_sqrt_pd_128, INTR_TYPE_1OP_MASK, ISD::FSQRT, 0), X86_INTRINSIC_DATA(avx512_mask_sqrt_pd_256, INTR_TYPE_1OP_MASK, ISD::FSQRT, 0), X86_INTRINSIC_DATA(avx512_mask_sqrt_pd_512, INTR_TYPE_1OP_MASK_RM, ISD::FSQRT, |
