News for hackers. Stuff that matters.

A comment on How to Optimize a CUDA Matmul Kernel for CuBLAS-Like Performance: A Worklog. Parent comment

Re:How to Optimize a CUDA Matmul Kernel for CuBLAS-Like Perf... by the_svd_doctor on Friday July 26, 2024 @11:57PM (#41084345)
FWIW Triton can actually be used pretty easily from C++ straight using MLIR. XLA does that. Maybe not exactly a frontend but not that far.
[ Parent | on HN ]
Re:How to Optimize a CUDA Matmul Kernel for CuBLAS-Like Perf... by almostgotcaught on Saturday July 27, 2024 @12:16AM (#41084425)
> FWIW Triton

i have PRs in Triton - i'm well familiar with the fact that triton is an MLIR project.

> C++ straight using MLIR

that's like saying llvm ir is usable through C++ ... or hell that's like saying NVPTX is usable through C++. it's not just not a frontend it's the exact opposite: it's emitting IR using IR builders i.e. writing a compiler.

[ Parent | on HN ]