diff options
| author | Kawrakow <48489457+ikawrakow@users.noreply.github.com> | 2024-07-27 07:55:01 +0200 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2024-07-27 07:55:01 +0200 |
| commit | 154e0d75fccf1784fe9ff6fd76a630b66563da3d (patch) | |
| tree | 81ce6dbb5b1900c1aa78a879f0593c694cab9d27 /ggml/src/ggml-cann/kernels/ascendc_kernels.h | |
| parent | 0684c3e9c70d49323b4fc517128cbe222cab7f96 (diff) | |
Merge mainline llama.cpp (#3)
* Merging mainline - WIP
* Merging mainline - WIP
AVX2 and CUDA appear to work.
CUDA performance seems slightly (~1-2%) lower as it is so often
the case with llama.cpp/ggml after some "improvements" have been made.
* Merging mainline - fix Metal
* Remove check
---------
Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
Diffstat (limited to 'ggml/src/ggml-cann/kernels/ascendc_kernels.h')
| -rw-r--r-- | ggml/src/ggml-cann/kernels/ascendc_kernels.h | 17 |
1 files changed, 17 insertions, 0 deletions
diff --git a/ggml/src/ggml-cann/kernels/ascendc_kernels.h b/ggml/src/ggml-cann/kernels/ascendc_kernels.h new file mode 100644 index 00000000..bf891475 --- /dev/null +++ b/ggml/src/ggml-cann/kernels/ascendc_kernels.h @@ -0,0 +1,17 @@ +#ifndef ASCENDC_KERNELS_H +#define ASCENDC_KERNELS_H + +#include "aclrtlaunch_ascendc_get_row_f32.h" +#include "aclrtlaunch_ascendc_get_row_f16.h" +#include "aclrtlaunch_ascendc_get_row_q8_0.h" +#include "aclrtlaunch_ascendc_get_row_q4_0.h" + +#include "aclrtlaunch_ascendc_quantize_f32_q8_0.h" +#include "aclrtlaunch_ascendc_quantize_f16_q8_0.h" + +#include "aclrtlaunch_ascendc_dup_by_rows_fp16.h" +#include "aclrtlaunch_ascendc_dup_by_rows_fp32.h" +#include "aclrtlaunch_ascendc_dup_by_rows_fp32_to_fp16.h" +#include "aclrtlaunch_ascendc_dup_by_rows_fp16_to_fp32.h" + +#endif // ASCENDC_KERNELS_H |