This revision was automatically updated to reflect the committed changes.
Closed by commit rL298767: [AMDGPU] Switch address space mapping by triple
environment amdgiz (authored by yaxunl).
Changed prior to commit:
https://reviews.llvm.org/D31210?vs=92997=93031#toc
Repository:
rL LLVM
yaxunl added inline comments.
Comment at: lib/Basic/Targets.cpp:2015
-static const unsigned AMDGPUAddrSpaceMap[] = {
- 1,// opencl_global
- 3,// opencl_local
- 2,// opencl_constant
- 4,// opencl_generic
- 1,// cuda_device
- 2,// cuda_constant
- 3
t-tye accepted this revision.
t-tye added a comment.
This revision is now accepted and ready to land.
Just a couple of suggestions, otherwise:
LGTM
Comment at: lib/Basic/Targets.cpp:2015
-static const unsigned AMDGPUAddrSpaceMap[] = {
- 1,// opencl_global
- 3,//
yaxunl updated this revision to Diff 92997.
yaxunl retitled this revision from "[AMDGPU] Add new address space mapping" to
"[AMDGPU] Switch address space mapping by triple environment".
yaxunl added a comment.
Use triple environment name.
https://reviews.llvm.org/D31210
Files: