b9950

ggml-org/llama.cppb9950Jul 10, 2026by github-actions[bot]same day as b9949

What changed since b9949

low risk

b9950 is a small follow-up to b9949, adding llama-batch unit tests and fixing Windows builds with no functional API changes.

New features

  • Unit test coverage added for the llama-batch subsystem.

Improvements

  • Windows/win32 build compatibility fixed alongside the new batch tests.
  • Unused batch methods now assert as not implemented, and unreachable code was removed for cleaner internals.

Bug fixes

  • Resolved Windows build failures introduced or exposed by the batch test work.

Upgrade notes

This is a test-and-build-hygiene release with no config, API, or model-format changes, so upgrading from b9949 should be a drop-in replacement. Windows builders are the main beneficiaries; other platforms see no behavioral difference. The Adreno OpenCL cluster-parallel decode FlashAttention work from b9949 remains unchanged.

Analyzed Oct 24, 2026

Upgrading?

Release notes summary

Introduces unit tests for the `llama-batch` system and fixes Windows build issues.

Key Highlights

  • Added unit tests for `llama-batch`
  • Fixes for Windows builds

Release Notes (excerpt)

<details open> llama-batch: add unit test (#25471) * llama-batch: add unit test * fix win32 builds * add not implemented assertion in unused methods * remove unreachable code </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)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories