b9992

ggml-org/llama.cppb9992Jul 13, 2026by github-actions[bot]same day as b9990

What changed since b9990

low risk

Refactored CUDA MMQ kernel configuration; focus shifted from Minimax2 EAGLE3 support.

Improvements

  • Refactored CUDA MMQ kernel configuration for better code organization
  • Improved kernel management structure
  • Fixed Blackwell GPU configuration

Bug fixes

  • Fixed Blackwell configuration in CUDA MMQ kernels

Upgrade notes

No configuration changes are required for users running other backends. The CUDA kernel refactor is an internal change that should not affect application behavior.

Analyzed Oct 17, 2026

Upgrading?

Release notes summary

The Multi-MatMul-Q (MMQ) kernel configuration for CUDA was refactored to improve code organization and maintainability. This includes fixing the Blackwell configuration and removing legacy code to streamline the kernel management.

Key Highlights

  • Refactored CUDA MMQ kernel configuration
  • Fixed Blackwell config
  • Removed legacy code
  • Improved kernel management structure

New Features

  • Refactored CUDA MMQ kernel configuration

Release Notes (excerpt)

<details open> CUDA: refactor MMQ kernel configuration (#24127) * CUDA: refactor MMQ kernel configuration * fix Blackwell config * remove legacy 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) - Ubuntu x64 (SYCL FP32)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories