NVIDIA / MatX

An efficient C++17 GPU numerical computing library with Python-like syntax

Home Page:https://nvidia.github.io/MatX

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

[BUG] `cmake_sample_project` cannot build when setting `MATX_EN_CUTLASS=ON`

lucifer1004 opened this issue · comments

Describe the Bug
See the title.

To Reproduce
Build cmake_sample_project with MATX_EN_CUTLASS=ON:

cmake -S examples/cmake_sample_project -B build -DMATX_EN_CUTLASS=ON
cmake --build build

The compiler would complain about missing cutlass headers:

  38 | #include "cutlass/gemm/device/gemm.h"
      |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
compilation terminated.
gmake[2]: *** [CMakeFiles/sample_matx.dir/build.make:77: CMakeFiles/sample_matx.dir/src/single.cu.o] Error 1
gmake[1]: *** [CMakeFiles/Makefile2:696: CMakeFiles/sample_matx.dir/all] Error 2
gmake: *** [Makefile:91: all] Error 2

There seems to be some issue with MatX's CMakeLists.txt, which does not handle the inclusion of cutlass headers correctly.

Expected Behavior
Build should work.

Hi @lucifer1004, CUTLASS hasn't been tested since 3.x was released due to the long compile times. Is there a reason you're trying to use CUTLASS? Currently we use cuBLASLt as the GEMM backend.

Thank you for the quick reply. I am just trying this project and testing each compile option. I do not necessarily need to use CUTLASS.

Thanks. We will remove it as an option since it's not used currently. If you see any functionality missing please let us know. Also let me know if it's okay to close this

I just found it was the same case with MATX_EN_PYBIND11=ON.

Error message:

build/_deps/pybind11-src/include/pybind11/cast.h(446): error: pointer to incomplete class type "_frame" is not allowed
                  "  " + handle(frame->f_code->co_filename).cast<std::string>() +

This appears to be a different issue. What compiler are you using?

OK, then I'll close this issue and open a new one for that.