[llvm-dev] AMDGPU and support for the new pass manager

[Arthur Eubanks via llvm-dev]

I've ported most of the IR passes and added them to AMDGPU's opt pipeline.
There are 2 issues remaining:
1) LegacyDivergenceAnalysis is used in LoopUnswitch to avoid unswitching
loops with divergent condition Values. I'm not sure what the state of the
LegacyDivergenceAnalysis vs DivergenceAnalysis is. (Also the new PM only
has SimpleLoopUnswitch instead of LoopUnswitch).
Transforms/LoopUnswitch/AMDGPU/divergent-unswitch.ll.
2) The AMDGPU backend has its own inliner
<https://github.com/llvm/llvm-project/blob/main/llvm/lib/Target/AMDGPU/AMDGPUInline.cpp>
(set
here
<https://github.com/llvm/llvm-project/blob/92be640bd7d4fbc8e032a0aa81381a0246efa0be/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp#L438>).
Any ideas for how to do custom inliner cost modeling in the new PM
pipelines? Allow targets to override an analysis pass that the inliner uses
to get an InlineAdvisor?
CodeGen/AMDGPU/amdgpu-inline.ll

On Fri, Dec 4, 2020 at 4:44 PM Arthur Eubanks <aeubanks at google.com> wrote:

> Some indicative failing tests are under the llvm/test/CodeGen/AMDGPU
> directory when opt's -enable-new-pm flag is set to true by default. They
> should also be repro'able by adding a corresponding NPM RUN line, e.g. "opt
> -S -O1 -mtriple=amdgcn-- ..." -> "opt -S -passes='default<O1>'
> -mtriple=amdgcn-- ...".
>
> I see the following AMDGPU-specific failures:
>   LLVM :: CodeGen/AMDGPU/amdgpu-inline.ll
>   LLVM :: CodeGen/AMDGPU/infer-addrpace-pipeline.ll
>   LLVM :: CodeGen/AMDGPU/internalize.ll
>   LLVM :: CodeGen/AMDGPU/llvm.amdgcn.wavefrontsize.ll
>   LLVM :: CodeGen/AMDGPU/opt-pipeline.ll
>   LLVM :: CodeGen/AMDGPU/propagate-attributes-clone.ll
>   LLVM :: CodeGen/AMDGPU/propagate-attributes-single-set.ll
>   LLVM :: CodeGen/AMDGPU/simplify-libcalls.ll
>   LLVM :: CodeGen/AMDGPU/sroa-before-unroll.ll
>   LLVM :: Transforms/LoopUnswitch/AMDGPU/divergent-unswitch.ll
>
> (Also I took a closer look, actually it looks like NVPTX also needs
> updating, but NVPTXTargetMachine only adds two passes)
>
> On Fri, Dec 4, 2020 at 3:48 PM Arthur Eubanks <aeubanks at google.com> wrote:
>
>> Currently AMDGPU is the last target which injects passes into the
>> pipeline but hasn't been updated to work with the new pass manager.
>>
>> Recently there was added support for the NPM's equivalent of
>> TargetMachine::adjustPassManager(). (Bug
>> <https://bugs.llvm.org/show_bug.cgi?id=47244>, Phab
>> <https://reviews.llvm.org/D88138>). BPF
>> <https://github.com/llvm/llvm-project/blob/99f79cbf31cc6ccdfa1aed253a64c5e8012f4ef7/llvm/lib/Target/BPF/BPFTargetMachine.cpp#L126>
>> and Hexagon have their NPM equivalents implemented already, but AMDGPU
>> has a lot of custom passes added by adjustPassManager
>> <https://github.com/llvm/llvm-project/blob/99f79cbf31cc6ccdfa1aed253a64c5e8012f4ef7/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp#L420> and
>> hasn't been updated to work with the NPM.
>>
>> Could the maintainers of AMDGPU port the necessary passes to the new pass
>> manager add them to AMDGPUTargetMachine::registerPassBuilderCallbacks()?
>> I'm happy to provide any guidance if necessary. Here's
>> <https://reviews.llvm.org/D91990> an example of porting a pass and
>> adding it to the pipeline.
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20210104/5bab0877/attachment.html>