[all-commits] [llvm/llvm-project] 77eafb: [mlir][nvvm] Generalize wmma ops to handle more ty...

Thomas via All-commits all-commits at lists.llvm.org
Mon Nov 1 10:29:21 PDT 2021


  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 77eafb8430d505d7e5f3dd84ea84fc0fc1f32647
      https://github.com/llvm/llvm-project/commit/77eafb8430d505d7e5f3dd84ea84fc0fc1f32647
  Author: thomasraoux <thomasraoux at google.com>
  Date:   2021-11-01 (Mon, 01 Nov 2021)

  Changed paths:
    M mlir/include/mlir/Dialect/LLVMIR/CMakeLists.txt
    M mlir/include/mlir/Dialect/LLVMIR/NVVMDialect.h
    M mlir/include/mlir/Dialect/LLVMIR/NVVMOps.td
    M mlir/include/mlir/Target/LLVMIR/ModuleTranslation.h
    M mlir/lib/Conversion/GPUToNVVM/WmmaOpsToNvvm.cpp
    M mlir/lib/Dialect/LLVMIR/IR/NVVMDialect.cpp
    M mlir/lib/Target/LLVMIR/Dialect/NVVM/NVVMToLLVMIRTranslation.cpp
    M mlir/lib/Target/LLVMIR/ModuleTranslation.cpp
    M mlir/test/Conversion/GPUToNVVM/wmma-ops-to-nvvm.mlir
    M mlir/test/Dialect/LLVMIR/invalid.mlir
    M mlir/test/Dialect/LLVMIR/nvvm.mlir
    M mlir/test/Target/LLVMIR/nvvmir.mlir
    M utils/bazel/llvm-project-overlay/mlir/BUILD.bazel

  Log Message:
  -----------
  [mlir][nvvm] Generalize wmma ops to handle more types and shapes

wmma intrinsics have a large number of combinations, ideally we want to be able
to target all the different variants. To avoid a combinatorial explosion in the
number of mlir op we use attributes to represent the different variation of
load/store/mma ops. We also can generate with tablegen helpers to know which
combinations are available. Using this we can avoid having too hardcode a path
for specific shapes and can support more types.
This patch also adds boiler plates for tf32 op support.

Differential Revision: https://reviews.llvm.org/D112689




More information about the All-commits mailing list