b9843

ggml-org/llama.cppb9843Jun 30, 2026by github-actions[bot]same day as b9842

What changed since b9842

low risk

b9843 reverts the split-compute synchronization optimization introduced earlier, restoring prior scheduler behavior with no API changes.

Improvements

  • Binary matrix is unchanged from b9842, so the same platform/backend artifacts remain available.
  • KleidiAI on macOS Apple Silicon stays disabled, matching the previous release.

Bug fixes

  • Reverts the 'less synchronizations during split compute' scheduler change to preempt potential issues it may have caused.

Upgrade notes

This is a drop-in swap with no config or API changes. If you benefited from b9842's reduced synchronization during split/multi-device compute, expect a small performance step back in those workloads. Otherwise no action is needed beyond re-pulling the matching binaries.

Analyzed Nov 1, 2026

Upgrading?

Release notes summary

This release reverts a previous optimization regarding synchronizations during split compute to address potential issues. It maintains the same comprehensive platform support and binary distribution as the previous release.

Key Highlights

  • Reverts 'less synchronizations during split compute' optimization
  • Includes standard binaries for macOS, Linux, Android, Windows, and openEuler
  • KleidiAI support remains disabled

Release Notes (excerpt)

<details open> Revert "sched : reintroduce less synchronizations during split compute (#20793)" (#25138) </details> **macOS/iOS:** - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED - macOS Intel (x64) - iOS XCFramework **Linux:** - Ubuntu x64 (CPU) - Ubuntu arm64 (CPU) - Ubuntu s390x (CPU) - Ubuntu x64 (Vulkan) - Ubuntu arm64 (Vulkan) - Ubuntu x64 (ROCm 7.2) - Ubuntu x64 (OpenVINO) - Ubuntu x64 (SYCL FP32) - Ubuntu x64 (SYCL FP16) **Android:**…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories