Skip to content

UPSTREAM PR #17988: sync : ggml#547

Open
loci-dev wants to merge 3 commits intomainfrom
upstream-PR17988-branch_ggml-org-sync-ggml-25-12-13
Open

UPSTREAM PR #17988: sync : ggml#547
loci-dev wants to merge 3 commits intomainfrom
upstream-PR17988-branch_ggml-org-sync-ggml-25-12-13

Conversation

@loci-dev
Copy link
Copy Markdown

Mirrored from ggml-org/llama.cpp#17988

HerrCai0907 and others added 3 commits December 13, 2025 10:07
… (ggml/1394)

Some backend depends on CMAKE_RUNTIME_OUTPUT_DIRECTORY to create temporary file like metal backened.
Missing CMAKE_RUNTIME_OUTPUT_DIRECTORY will cause some cmake error like permission denied (try to copy file to root).
This PR wants to setup a default path for CMAKE_RUNTIME_OUTPUT_DIRECTORY when it does not exist.
@loci-review
Copy link
Copy Markdown

loci-review bot commented Dec 13, 2025

Explore the complete analysis inside the Version Insights

Performance Analysis Summary

Analysis Overview

Project: llama.cpp
Versions Compared:

  • Base: 67ec858b-3016-41ee-a793-53a0553c833c
  • Target: e9981450-b895-46a9-9c28-a47f5cb0372e

Summary

No performance changes detected between versions. All analyzed functions show identical throughput and response time metrics. The llama_decode function maintains 70 ns throughput and 733772 ns response time in both versions with 0% change. Power consumption across all binaries remains unchanged, with libggml-cpu.so showing a negligible 3 nJ reduction. No impact on tokens per second for inference workloads.

@loci-dev loci-dev force-pushed the main branch 24 times, most recently from f251487 to 752a7d5 Compare December 16, 2025 10:10
@loci-dev loci-dev force-pushed the main branch 30 times, most recently from 7ceec3c to c8dcfe6 Compare December 21, 2025 10:08
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants