Skip to content

Conversation

@lizhenneng
Copy link
Contributor

We can set CMAKE_CUDA_COMPILER to /usr/local/cuda/bin/nvcc, but it's is not compatible with other system, so we use cmake's variable CUDAToolkit_NVCC_EXECUTABLE.

@github-actions github-actions bot added Nvidia GPU Issues specific to Nvidia GPUs ggml changes relating to the ggml tensor library for machine learning labels May 14, 2025
@CISC
Copy link
Collaborator

CISC commented May 14, 2025

According to the docs:

CUDAToolkit_NVCC_EXECUTABLE
Note that this path may not be the same as CMAKE_CUDA_COMPILER

Which means you probably should not unconditionally override CMAKE_CUDA_COMPILER.

@lizhenneng lizhenneng closed this May 19, 2025
@lizhenneng lizhenneng deleted the cuda-fix branch May 19, 2025 01:24
@lizhenneng lizhenneng restored the cuda-fix branch May 19, 2025 01:24
@lizhenneng lizhenneng deleted the cuda-fix branch May 19, 2025 01:24
@lizhenneng lizhenneng restored the cuda-fix branch May 19, 2025 01:31
@lizhenneng lizhenneng reopened this May 19, 2025
@lizhenneng
Copy link
Contributor Author

According to the docs:

CUDAToolkit_NVCC_EXECUTABLE
Note that this path may not be the same as CMAKE_CUDA_COMPILER

Which means you probably should not unconditionally override CMAKE_CUDA_COMPILER.

I will revise my patch later。

lizhenneng pushed a commit to lizhenneng/llama.cpp that referenced this pull request May 19, 2025
@lizhenneng
Copy link
Contributor Author

According to the docs:

CUDAToolkit_NVCC_EXECUTABLE
Note that this path may not be the same as CMAKE_CUDA_COMPILER

Which means you probably should not unconditionally override CMAKE_CUDA_COMPILER.

@lizhenneng lizhenneng closed this May 21, 2025
@lizhenneng lizhenneng reopened this May 21, 2025
@lizhenneng
Copy link
Contributor Author

Will you merge this patch?

@slaren
Copy link
Member

slaren commented Jun 20, 2025

I don't think there is enough justification to merge this. You can set manually CMAKE_CUDA_COMPILER if you have an unusual system configuration with multiple nvcc compilers.

@CISC
Copy link
Collaborator

CISC commented Jun 20, 2025

I don't think there is enough justification to merge this. You can set manually CMAKE_CUDA_COMPILER if you have an unusual system configuration with multiple nvcc compilers.

And certainly not as-is, it still unconditionally overrides the variable.

@lizhenneng lizhenneng closed this Sep 24, 2025
@lizhenneng lizhenneng deleted the cuda-fix branch September 24, 2025 09:57
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ggml changes relating to the ggml tensor library for machine learning Nvidia GPU Issues specific to Nvidia GPUs

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants