[PATCH] D35974: AMDGPU: Add pass to replace out arguments

[Matt Arsenault via Phabricator via llvm-commits]

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

It is better to return arguments directly in registers
if we are making a call rather than introducing expensive
stack usage. In one of sample compile from one of
 Blender's many kernel variants, this fires on about
~20 different functions. Future improvements may be to
recognize simple cases where the pointer is indexing a small
array. This also fails when the store to the out argument
is in a separate block from the return, which happens in
a few of the Blender functions. This should also probably
be using MemorySSA which might help with that.

      

I'm not sure this is correct as a FunctionPass, but
MemoryDependenceAnalysis seems to not work with
a ModulePass.

      

I'm also not sure where it should run.I think it should
run  before DeadArgumentElimination, so maybe either
EP_CGSCCOptimizerLate or EP_ScalarOptimizerLate.


https://reviews.llvm.org/D35974

Files:
  lib/Target/AMDGPU/AMDGPU.h
  lib/Target/AMDGPU/AMDGPURewriteOutArguments.cpp
  lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
  lib/Target/AMDGPU/CMakeLists.txt
  test/CodeGen/AMDGPU/rewrite-out-arguments-address-space.ll
  test/CodeGen/AMDGPU/rewrite-out-arguments.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D35974.108573.patch
Type: text/x-patch
Size: 40832 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170728/7c4ab611/attachment.bin>