LLM inference in C/C++
C++ ggml
cuda : skip UMA override for HIP builds (#27083)
Mario Limonciello60eeeb6082c1
diff --git a/ggml/src/ggml-cuda/ggml-cuda.cu b/ggml/src/ggml-cuda/ggml-cuda.cuindex 598f3228c4..f2e381ee00 100644--- a/ggml/src/ggml-cuda/ggml-cuda.cu+++ b/ggml/src/ggml-cuda/ggml-cuda.cu@@ -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) systemcudaDeviceProp 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 themconst int share_count = ggml_cuda_physical_device_share_count(ctx->device);