]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: d32e03f)
vulkan: perf_logger improvements (#15246)
authorJeff Bolz <redacted>
Thu, 14 Aug 2025 13:38:10 +0000 (08:38 -0500)
committerGitHub <redacted>
Thu, 14 Aug 2025 13:38:10 +0000 (08:38 -0500)
commit863d341eeb81db104902b14b6f9413daa515e957
tree57115c1c0b59b9b47c1e8ee49ceceeadce5f8a7ftree
parentd32e03f4495d3efa1c5126f53b449f1d429c5664commit | diff
vulkan: perf_logger improvements (#15246)

* vulkan: perf_logger improvements

- Account for batch dimension in flops calculation.
- Fix how "_VEC" is detected for mat_mul_id.
- Fix "n" dimension for mat_mul_id (in case of broadcasting).
- Include a->type in name.

* use <=mul_mat_vec_max_cols rather than ==1
ggml/src/ggml-vulkan/ggml-vulkan.cpp diff | blob | history
Packaging of ggml-org/llama.cpp