b9814

ggml-org/llama.cppb9814Jun 26, 2026by github-actions[bot]same day as b9813

What changed since b9813

low risk

b9814 adds a Vulkan mul_mat_vecq optimization for AMD MI50 GPUs on top of b9813's Intel Xe1/coopmat1 support.

Improvements

  • Vulkan backend's quantized mat-vec (mul_mat_vecq) path is tuned specifically for AMD Radeon MI50 hardware.
  • Carries forward b9813's Intel Xe1 architecture enum and coopmat1 enablement on Xe-LPG Plus GPUs.

Upgrade notes

This is a single-commit increment over b9813, so no config or API changes are required. AMD MI50 users on the Vulkan backend should see faster quantized mat-vec throughput; other backends are unaffected. Intel GPU users retain the Xe1/coopmat1 behavior introduced in b9813.

Analyzed Nov 1, 2026

Upgrading?

Release notes summary

Optimizes the Vulkan backend's mul_mat_vecq operation for AMD Radeon MI50 GPUs, improving performance for quantized matrix multiplication operations.

Key Highlights

  • Optimized mul_mat_vecq for AMD Radeon MI50
  • Vulkan backend performance improvements

New Features

  • Vulkan optimization for MI50 GPUs

Release Notes (excerpt)

<details open> vulkan: opt mul_mat_vecq for mi50 (#22933) </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:** - Android arm64 (CPU) **Windows:** - Windows x64 (CPU)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories