public_git
/
whisper.cpp
mirror of https://github.com/ggml-org/whisper.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

cuda : skip UMA override for HIP builds (llama/27083) 

AMD APUs report accurate memory via hipMemGetInfo. Using
MemAvailable over-promises on small-carveout systems.

fixes #18159
This commit is contained in:
Mario Limonciello 2026-08-17 11:35:36 -05:00 committed by Georgi Gerganov
parent 964bb1b8f5
commit 71759b7c9f
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/src/ggml-cuda/ggml-cuda.cu
View File

@ -4770,7 +4770,7 @@ static void ggml_backend_cuda_device_get_memory(ggml_backend_dev_t dev, size_t *
}
// ref: https://github.com/ggml-org/llama.cpp/pull/17368
#if defined(__linux__)
#if defined(__linux__) && !defined(GGML_USE_HIP)
// Check if this is a UMA (Unified Memory Architecture) system
cudaDeviceProp prop;
CUDA_CHECK(cudaGetDeviceProperties(&prop, ggml_cuda_get_physical_device(ctx->device)));
@ -4790,7 +4790,7 @@ static void ggml_backend_cuda_device_get_memory(ggml_backend_dev_t dev, size_t *
GGML_LOG_ERROR("%s: /proc/meminfo reading failed, using cudaMemGetInfo\n", __func__);
}
}
#endif // defined(__linux__)
#endif // defined(__linux__) && !defined(GGML_USE_HIP)
// virtual devices sharing one physical GPU share its memory pool; split it between them
const int share_count = ggml_cuda_physical_device_share_count(ctx->device);