[all-commits] [llvm/llvm-project] 283b5e: [MLIR] Make gpu.launch implicitly capture uses of ...

[Stephan Herhut via All-commits]

  Branch: refs/heads/master
  Home:   https://github.com/llvm/llvm-project
  Commit: 283b5e733d1b388e9e912e83c8c24d543e389dc1
      https://github.com/llvm/llvm-project/commit/283b5e733d1b388e9e912e83c8c24d543e389dc1
  Author: Stephan Herhut <herhut at google.com>
  Date:   2020-02-03 (Mon, 03 Feb 2020)

  Changed paths:
    M mlir/include/mlir/Dialect/GPU/GPUOps.td
    M mlir/lib/Conversion/LoopsToGPU/LoopsToGPU.cpp
    M mlir/lib/Dialect/GPU/IR/GPUDialect.cpp
    M mlir/lib/Dialect/GPU/Transforms/KernelOutlining.cpp
    M mlir/test/Conversion/LoopsToGPU/imperfect_2D.mlir
    M mlir/test/Conversion/LoopsToGPU/linalg_to_gpu.mlir
    M mlir/test/Conversion/LoopsToGPU/step_one.mlir
    R mlir/test/Dialect/GPU/canonicalize.mlir
    M mlir/test/Dialect/GPU/invalid.mlir
    M mlir/test/Dialect/GPU/ops.mlir
    M mlir/test/Dialect/GPU/outlining.mlir
    M mlir/test/mlir-cuda-runner/all-reduce-op.mlir
    M mlir/test/mlir-cuda-runner/all-reduce-region.mlir
    M mlir/test/mlir-cuda-runner/gpu-to-cubin.mlir
    M mlir/test/mlir-cuda-runner/shuffle.mlir

  Log Message:
  -----------
  [MLIR] Make gpu.launch implicitly capture uses of values defined above.

Summary:
In the original design, gpu.launch required explicit capture of uses
and passing them as operands to the gpu.launch operation. This was
motivated by infrastructure restrictions rather than design. This
change lifts the requirement and removes the concept of kernel
arguments from gpu.launch. Instead, the kernel outlining
transformation now does the explicit capturing.

This is a breaking change for users of gpu.launch.

Differential Revision: https://reviews.llvm.org/D73769