[all-commits] [llvm/llvm-project] 553569: [mlir] Add gpu.allocate, gpu.deallocate ops with L...

[Christian Sigg via All-commits]

  Branch: refs/heads/temp-test-main
  Home:   https://github.com/llvm/llvm-project
  Commit: 5535696c386ba89b66c1b5a72a2aa98783571cc9
      https://github.com/llvm/llvm-project/commit/5535696c386ba89b66c1b5a72a2aa98783571cc9
  Author: Christian Sigg <csigg at google.com>
  Date:   2020-11-27 (Fri, 27 Nov 2020)

  Changed paths:
    M mlir/include/mlir/Dialect/GPU/GPUDialect.h
    M mlir/include/mlir/Dialect/GPU/GPUOps.td
    M mlir/lib/Conversion/GPUCommon/ConvertLaunchFuncToRuntimeCalls.cpp
    A mlir/test/Conversion/GPUCommon/lower-alloc-to-gpu-runtime-calls.mlir
    M mlir/test/Dialect/GPU/ops.mlir
    M mlir/tools/mlir-cuda-runner/cuda-runtime-wrappers.cpp
    M mlir/tools/mlir-rocm-runner/rocm-runtime-wrappers.cpp

  Log Message:
  -----------
  [mlir] Add gpu.allocate, gpu.deallocate ops with LLVM lowering to runtime function calls.

The ops are very similar to the std variants, but support async GPU execution.

gpu.alloc does not currently support an alignment attribute, and the new ops do not have
canonicalizers/folders like their std siblings do.

Reviewed By: herhut

Differential Revision: https://reviews.llvm.org/D91698