[all-commits] [llvm/llvm-project] f69555: [mlir][sparse] Use the correct ABI on x86 and re-e...

[Benjamin Kramer via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: f695554a2a5550ae40da35af9ac22bfcca5db09a
      https://github.com/llvm/llvm-project/commit/f695554a2a5550ae40da35af9ac22bfcca5db09a
  Author: Benjamin Kramer <benny.kra at googlemail.com>
  Date:   2022-08-11 (Thu, 11 Aug 2022)

  Changed paths:
    M mlir/lib/ExecutionEngine/Float16bits.cpp
    M mlir/test/Integration/Dialect/SparseTensor/CPU/dense_output_bf16.mlir
    M mlir/test/Integration/Dialect/SparseTensor/CPU/sparse_sum_bf16.mlir

  Log Message:
  -----------
  [mlir][sparse] Use the correct ABI on x86 and re-enable tests

c7ec6e19d5446a448f888b33f66316cf2ec6ecae made LLVM adhere to the x86
psABI and pass bf16 in SSE registers instead of GPRs. This breaks the
custom versions of runtime functions we have for bf16 conversion. A
great fix for this would be to use __bf16 types instead which carry the
right ABI, but that type isn't widely available.

Instead just pretend it's a 32 bit float on the ABI boundary and
carefully cast it to the right type.

Fixes #57042