[AArch64][SVE] Teach cost model masked gathers/scatters are cheap
authorMatthew Devereau <matthew.devereau@arm.com>
Thu, 26 Aug 2021 10:08:03 +0000 (11:08 +0100)
committerMatthew Devereau <matthew.devereau@arm.com>
Thu, 26 Aug 2021 10:17:47 +0000 (11:17 +0100)
commit9b830c798ebf2695772e8eaaeed3cd24fa66b68b
tree32b13f8365b9f2b3d992dc0df163409362dedd4b
parentbd7ece4e063e7afd08cbaa311878c09aadf5ec21
[AArch64][SVE] Teach cost model masked gathers/scatters are cheap

Tell the cost model to use the scalable calculation for non-neon fixed vector.
This results in a cheaper cost for fixed-length SVE masked gathers/scatters
allowing the vectorizor to emit them more frequently.
llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
llvm/test/Analysis/CostModel/AArch64/mem-op-cost-model.ll