]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 017761d)
vulkan: perf_logger improvements (#17672)
authorJeff Bolz <redacted>
Sat, 6 Dec 2025 17:46:46 +0000 (11:46 -0600)
committerGitHub <redacted>
Sat, 6 Dec 2025 17:46:46 +0000 (18:46 +0100)
commitdb97837385edfbc772230debbd49e5efae843a71
tree10be34c330432f23e3ce4b89aec14197c0838234tree
parent017761daf5f543e3f064d956784a9c9421dc0b66commit | diff
vulkan: perf_logger improvements (#17672)

* vulkan: perf_logger improvements

- Move perf_logger from device to ctx.
- Add an env var to control the frequency we dump the stats. If you set a very
large value, it just dumps when the ctx is destroyed.
- Add a fusion info string to the tracking, only log one item per fused op.
- Fix MUL_MAT_ID flops calculation.

* fix vector sizes
ggml/src/ggml-vulkan/ggml-vulkan.cpp diff | blob | history
Packaging of ggml-org/llama.cpp