[PATCH] D54351: [AMDGPU] combine extractelement into several selects

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec created this revision.
rampitec added reviewers: arsenm, nhaehnle.
Herald added subscribers: t-tye, tpr, dstuttard, yaxunl, wdng, jvesely, kzhuravl.

An extractelement with non-constant index will be lowered either to
scratch or movrel loop in most cases. This patch converts such
instruction into a set of selects if vector size is not too big.


https://reviews.llvm.org/D54351

Files:
  lib/Target/AMDGPU/SIISelLowering.cpp
  test/CodeGen/AMDGPU/extract_vector_dynelt.ll
  test/CodeGen/AMDGPU/extract_vector_elt-f64.ll
  test/CodeGen/AMDGPU/extract_vector_elt-i64.ll
  test/CodeGen/AMDGPU/indirect-addressing-si-noopt.ll
  test/CodeGen/AMDGPU/indirect-addressing-si.ll
  test/CodeGen/AMDGPU/promote-alloca-vector-to-vector.ll
  test/CodeGen/AMDGPU/smrd.ll
  test/CodeGen/AMDGPU/trunc-bitcast-vector.ll
  test/CodeGen/AMDGPU/vector-extract-insert.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D54351.173434.patch
Type: text/x-patch
Size: 30015 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20181109/145141ff/attachment-0001.bin>