llama.cpp
4d0598e1 - HIP: add GGML_CUDA_CC_IS_* for amd familys as increasing cc archtectures for amd gpus are not supersets of eatch other (#11601)

Commit

250 days ago

HIP: add GGML_CUDA_CC_IS_* for amd familys as increasing cc archtectures for amd gpus are not supersets of eatch other (#11601) This fixes a bug where RDNA1 gpus other than gfx1010 where not handled correctly

References

#11601 - HIP: add GGML_CUDA_CC_IS_* macros for testing for amd architeure familys

Author

IMbackK

Parents

90f9b88a

llama.cpp 4d0598e1 - HIP: add GGML_CUDA_CC_IS_* for amd familys as increasing cc archtectures for amd gpus are not supersets of eatch other (#11601)

llama.cpp
4d0598e1 - HIP: add GGML_CUDA_CC_IS_* for amd familys as increasing cc archtectures for amd gpus are not supersets of eatch other (#11601)