[PATCH] D98512: [LoopVectorize] Simplify scalar cost calculation in getInstructionCost

Mon Mar 22 05:29:35 PDT 2021

david-arm updated this revision to Diff 332257.
david-arm added a comment.

- Added an assert that if an instruction remains scalar after vectorisation, then both the instruction and all it's users will not be scalarised. If any users were to be scalarised then potentially that means multiple copies of the instruction will be generated. In the case of GEPs the cost is actually calculated as part of the load/store cost.

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D98512/new/

https://reviews.llvm.org/D98512

Files:
  llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
  llvm/test/Transforms/LoopVectorize/AArch64/no_vector_instructions.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D98512.332257.patch
Type: text/x-patch
Size: 5688 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210322/bf5cfcf3/attachment.bin>