Commit Graph

10 Commits

Author SHA1 Message Date
Matt Arsenault a680199a96 Reapply "AMDGPU: Force inlining if LDS global address is used"
This reverts commit r336623

llvm-svn: 336675
2018-07-10 14:03:41 +00:00
Vlad Tsyrklevich 688e752207 Revert "AMDGPU: Force inlining if LDS global address is used"
This reverts commit r336587, it was causing test failures on the
sanitizer bots.

llvm-svn: 336623
2018-07-10 00:46:07 +00:00
Matt Arsenault 40cb6cab56 AMDGPU: Force inlining if LDS global address is used
These won't work for the forseeable future. These aren't allowed
from OpenCL, but IPO optimizations can make them appear.

Also directly set the attributes on functions, regardless
of the linkage rather than cloning functions like before.

llvm-svn: 336587
2018-07-09 19:22:22 +00:00
Davide Italiano ff829cea8b [PassManager] Run global optimizations after the inliner.
The inliner performs some kind of dead code elimination as it goes,
but there are cases that are not really caught by it. We might
at some point consider teaching the inliner about them, but it
is OK for now to run GlobalOpt + GlobalDCE in tandem as their
benefits generally outweight the cost, making the whole pipeline
faster.

This fixes PR34652.

Differential Revision: https://reviews.llvm.org/D38154

llvm-svn: 314997
2017-10-05 18:06:37 +00:00
Matt Arsenault 3333968771 Verifier: Check some amdgpu calling convention restrictions
llvm-svn: 299457
2017-04-04 18:43:11 +00:00
Stanislav Mekhanoshin 89653dfd2a [AMDGPU] Add GlobalOpt parameter to Always Inliner pass
If set to false it does not remove global aliases. With this parameter
set to false it should be safe to run the pass before link.

Differential Revision: https://reviews.llvm.org/D31489

llvm-svn: 299108
2017-03-30 20:16:02 +00:00
Stanislav Mekhanoshin 9053f22eeb [AMDGPU] Split -amdgpu-early-inline-all option
Previously it was covered by the internalization. It turns out we cannot
run internalizer in FE, it break separate compilation tests. Thus early
inliner gets its own option.

Differential Revision: https://reviews.llvm.org/D31429

llvm-svn: 298935
2017-03-28 18:23:24 +00:00
Konstantin Zhuravlyov 2534bc07f4 [AMDGPU] Run always inliner early in opt
Differential Revision: https://reviews.llvm.org/D31141

llvm-svn: 298281
2017-03-20 18:06:45 +00:00
Konstantin Zhuravlyov 8a67eb144f Revert "[AMDGPU] Run always inliner early in opt"
This reverts commit r297958, it breaks device-libs build.

llvm-svn: 298239
2017-03-20 09:26:08 +00:00
Stanislav Mekhanoshin f80507979d [AMDGPU] Run always inliner early in opt
We can mark functions to always inline early in the opt. Since we do not have
call support this early inlining creates opportunities for inter-procedural
optimizations which would not occur otherwise.

Differential Revision: https://reviews.llvm.org/D31016

llvm-svn: 297958
2017-03-16 16:11:46 +00:00