Skip to content

b5139

Choose a tag to compare

@github-actions github-actions released this 15 Apr 11:09
84778e9
CUDA/HIP: Share the same unified memory allocation logic. (#12934)

Replace compile-time `GGML_HIP_UMA` with environment variable `GGML_CUDA_ENABLE_UNIFIED_MEMORY`. This unifies the usage on NVIDIA and AMD GPUs, and allows a single binary to be shared between integrated and dedicated GPUs.