[PATCH] D46010: [AArch64] Improve cost of vector division by constant

[Adhemerval Zanella via Phabricator via llvm-commits]

zatrazz created this revision.
zatrazz added reviewers: fhahn, rengolin, javed.absar, huntergr, SjoerdMeijer, t.p.northover, echristo, evandro.
Herald added a subscriber: kristof.beyls.

With custom lowering for vector MULLH{S,U} [1], it is now profitable to
vectorize a divide by constant loop for the custom types (v16i8, v8i16,
and v4i32).  The cost if based on TargetLowering::Build{S,U}DIV which
uses a multiply by constant plus adjustment to express a divide by
constant.

Both {u,s}mull{2} are expressed as Instruction::Mul and shifts by
Instruction::AShr.

[1] With custom lowering for vector MULLH{S,U}, it is now profitable to
vectorize a divide by constant loop for the custom types (v16i8, v8i16,
and v4i32).  The cost if based on TargetLowering::Build{S,U}DIV which
uses a multiply by constant plus adjustment to express a divide by
constant.

Both {u,s}mull{2} are expressed as Instruction::Mul and shifts by
Instruction::AShr.


Repository:
  rL LLVM

https://reviews.llvm.org/D46010

Files:
  include/llvm/CodeGen/BasicTTIImpl.h
  lib/Target/AArch64/AArch64TargetTransformInfo.cpp
  test/Analysis/CostModel/AArch64/div_cte.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D46010.143730.patch
Type: text/x-patch
Size: 5240 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180424/ad330cbf/attachment.bin>