[PATCH] D141700: AMDGPU: Move enqueued block handling into clang

[Matt Arsenault via Phabricator via cfe-commits]

arsenm created this revision.
arsenm added reviewers: yaxunl, t-tye, b-sumner, rampitec, AMDGPU, Anastasia, JonChesterfield, jhuber6.
Herald added subscribers: kosarev, foad, kerbowa, hiraditya, tpr, dstuttard, jvesely, kzhuravl.
Herald added a project: All.
arsenm requested review of this revision.
Herald added a subscriber: wdng.
Herald added a project: LLVM.

The previous implementation wasn't maintaining a faithful IR
representation of how this really works. The value returned by
createEnqueuedBlockKernel wasn't actually used as a function, and
hacked up later to be a pointer to the runtime handle global
variable. In reality, the enqueued block is a struct where the first
field is a pointer to the kernel descriptor, not the kernel itself. We
were also relying on passing around a reference to a global using a
string attribute containing its name. It's better to base this on a
proper IR symbol reference during final emission.

      

This now avoids using a function attribute on kernels and avoids using
the additional "runtime-handle" attribute to populate the final
metadata. Instead, associate the runtime handle reference to the
kernel with the !associated global metadata. We can then get a final,
correctly mangled name at the end.

      

I couldn't figure out how to get rename-with-external-symbol behavior
using a combination of comdats and aliases, so leaves an IR pass to
 externalize the runtime handles for codegen. If anything breaks, it's
most likely this, so leave avoiding this for a later step. Use a
special section name to enable this behavior. This also means it's
possible to declare enqueuable kernels in source without going through
the dedicated block syntax or other dedicated compiler support.

      

We could move towards initializing the runtime handle in the
compiler/linker. I have a working patch where the linker sets up the
first field of the handle, avoiding the need to export the block
kernel symbol for the runtime. We would need new relocations to get
the private and group sizes, but that would avoid the runtime's
special case handling that requires the device_enqueue_symbol metadata
field.

      

Handle autoupgrade from the old kernel attribute. Not sure where I
could put the code shared with clang (maybe could rename
AMDGPUEmitPrintf to AMDGPUUtils).


https://reviews.llvm.org/D141700

Files:
  clang/lib/CodeGen/TargetInfo.cpp
  clang/test/CodeGenOpenCL/amdgpu-enqueue-kernel-linking.cl
  clang/test/CodeGenOpenCL/amdgpu-enqueue-kernel.cl
  llvm/docs/AMDGPUUsage.rst
  llvm/lib/IR/AutoUpgrade.cpp
  llvm/lib/IR/CMakeLists.txt
  llvm/lib/Target/AMDGPU/AMDGPU.h
  llvm/lib/Target/AMDGPU/AMDGPUExportKernelRuntimeHandles.cpp
  llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.cpp
  llvm/lib/Target/AMDGPU/AMDGPUHSAMetadataStreamer.h
  llvm/lib/Target/AMDGPU/AMDGPUOpenCLEnqueuedBlockLowering.cpp
  llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
  llvm/lib/Target/AMDGPU/CMakeLists.txt
  llvm/test/Bitcode/amdgpu-autoupgrade-enqueued-block.ll
  llvm/test/CodeGen/AMDGPU/amdgpu-export-kernel-runtime-handles.ll
  llvm/test/CodeGen/AMDGPU/enqueue-kernel.ll
  llvm/test/CodeGen/AMDGPU/hsa-metadata-from-llvm-ir-full-v3.ll
  llvm/test/CodeGen/AMDGPU/hsa-metadata-from-llvm-ir-full.ll
  llvm/test/CodeGen/AMDGPU/llc-pipeline.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D141700.489026.patch
Type: text/x-patch
Size: 97126 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20230113/e5c9dfd2/attachment-0001.bin>