From 8425f453ef77e3e1e14f804195802eae80261bef Mon Sep 17 00:00:00 2001 From: Ahmed Bougacha Date: Sat, 11 Feb 2017 01:53:04 +0000 Subject: [PATCH] [ARM] Make f16 interleaved accesses expensive. There are no vldN/vstN f16 variants, even with +fullfp16. We could use the i16 variants, but, in practice, even with +fullfp16, the f16 sequence leading to the i16 shuffle usually gets scalarized. We'd need to improve our support for f16 codegen before getting there. Teach the cost model to consider f16 interleaved operations as expensive. Otherwise, we are all but guaranteed to end up with a large block of scalarized vector code. llvm-svn: 294819 --- .../lib/Target/ARM/ARMTargetTransformInfo.cpp | 3 +- .../LoopVectorize/ARM/interleaved_cost.ll | 31 +++++++++++++++++++ 2 files changed, 33 insertions(+), 1 deletion(-) diff --git a/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp b/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp index 2b6b36bc3e68..90e4f3a7cc6f 100644 --- a/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp +++ b/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp @@ -533,7 +533,8 @@ int ARMTTIImpl::getInterleavedMemoryOpCost(unsigned Opcode, Type *VecTy, unsigned SubVecSize = DL.getTypeSizeInBits(SubVecTy); // vldN/vstN only support legal vector types of size 64 or 128 in bits. - if (NumElts % Factor == 0 && (SubVecSize == 64 || SubVecSize == 128)) + if (NumElts % Factor == 0 && (SubVecSize == 64 || SubVecSize == 128) && + !VecTy->getScalarType()->isHalfTy()) return Factor; } diff --git a/llvm/test/Transforms/LoopVectorize/ARM/interleaved_cost.ll b/llvm/test/Transforms/LoopVectorize/ARM/interleaved_cost.ll index 012a5fc16867..685c523d495a 100644 --- a/llvm/test/Transforms/LoopVectorize/ARM/interleaved_cost.ll +++ b/llvm/test/Transforms/LoopVectorize/ARM/interleaved_cost.ll @@ -99,3 +99,34 @@ for.body: for.end: ret void } + +%half.2 = type {half, half} +define void @half_factor_2(%half.2* %data, i64 %n) { +entry: + br label %for.body + +; VF_4-LABEL: Checking a loop in "half_factor_2" +; VF_4: Found an estimated cost of 40 for VF 4 For instruction: %tmp2 = load half, half* %tmp0, align 2 +; VF_4-NEXT: Found an estimated cost of 0 for VF 4 For instruction: %tmp3 = load half, half* %tmp1, align 2 +; VF_4-NEXT: Found an estimated cost of 0 for VF 4 For instruction: store half 0xH0000, half* %tmp0, align 2 +; VF_4-NEXT: Found an estimated cost of 40 for VF 4 For instruction: store half 0xH0000, half* %tmp1, align 2 +; VF_8-LABEL: Checking a loop in "half_factor_2" +; VF_8: Found an estimated cost of 80 for VF 8 For instruction: %tmp2 = load half, half* %tmp0, align 2 +; VF_8-NEXT: Found an estimated cost of 0 for VF 8 For instruction: %tmp3 = load half, half* %tmp1, align 2 +; VF_8-NEXT: Found an estimated cost of 0 for VF 8 For instruction: store half 0xH0000, half* %tmp0, align 2 +; VF_8-NEXT: Found an estimated cost of 80 for VF 8 For instruction: store half 0xH0000, half* %tmp1, align 2 +for.body: + %i = phi i64 [ 0, %entry ], [ %i.next, %for.body ] + %tmp0 = getelementptr inbounds %half.2, %half.2* %data, i64 %i, i32 0 + %tmp1 = getelementptr inbounds %half.2, %half.2* %data, i64 %i, i32 1 + %tmp2 = load half, half* %tmp0, align 2 + %tmp3 = load half, half* %tmp1, align 2 + store half 0., half* %tmp0, align 2 + store half 0., half* %tmp1, align 2 + %i.next = add nuw nsw i64 %i, 1 + %cond = icmp slt i64 %i.next, %n + br i1 %cond, label %for.body, label %for.end + +for.end: + ret void +}