d7c94e0ca6
* Better support for AMD multi-GPU This resolves a number of problems related to AMD multi-GPU setups on linux. The numeric IDs used by rocm are not the same as the numeric IDs exposed in sysfs although the ordering is consistent. We have to count up from the first valid gfx (major/minor/patch with non-zero values) we find starting at zero. There are 3 different env vars for selecting GPUs, and only ROCR_VISIBLE_DEVICES supports UUID based identification, so we should favor that one, and try to use UUIDs if detected to avoid potential ordering bugs with numeric IDs * ROCR_VISIBLE_DEVICES only works on linux Use the numeric ID only HIP_VISIBLE_DEVICES on windows |
||
---|---|---|
.. | ||
amd_common.go | ||
amd_hip_windows.go | ||
amd_linux.go | ||
amd_windows.go | ||
cpu_common.go | ||
cuda_common.go | ||
gpu.go | ||
gpu_darwin.go | ||
gpu_info.h | ||
gpu_info_cudart.c | ||
gpu_info_cudart.h | ||
gpu_info_darwin.h | ||
gpu_info_darwin.m | ||
gpu_info_nvcuda.c | ||
gpu_info_nvcuda.h | ||
gpu_info_nvml.c | ||
gpu_info_nvml.h | ||
gpu_info_oneapi.c | ||
gpu_info_oneapi.h | ||
gpu_linux.go | ||
gpu_oneapi.go | ||
gpu_test.go | ||
gpu_windows.go | ||
gpu_windows_test.go | ||
types.go |