[PATCH] D24746: AMDGPU/R600: Don't use REGISTER_{LOAD, STORE} ISD nodes

[Jan Vesely via llvm-commits]

jvesely created this revision.
jvesely added a subscriber: llvm-commits.
jvesely set the repository for this revision to rL LLVM.
Herald added a reviewer: tstellarAMD.
Herald added subscribers: tony-tye, yaxunl, nhaehnle, wdng, kzhuravl, arsenm.

This helps eliminate few redundant MOVs from private AS loads/stores
I think we'd need to be smarter about MMO for truncating store/extending loads to eliminate all of them.

This also makes it much easier to switch to MEM_{RD,WR}_SCRATCH once mesa bits are done.

Repository:
  rL LLVM

https://reviews.llvm.org/D24746

Files:
  lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
  lib/Target/AMDGPU/R600ISelLowering.cpp
  lib/Target/AMDGPU/R600Instructions.td
  test/CodeGen/AMDGPU/load-constant-i16.ll
  test/CodeGen/AMDGPU/load-global-i16.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D24746.71879.patch
Type: text/x-patch
Size: 16835 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160919/899760cb/attachment.bin>