b10019

ggml-org/llama.cppb10019Jul 15, 2026by github-actions[bot]same day as b10018

What changed since b10018

low risk

Adds FP16 kernel-type support for SYCL depthwise convolution (conv2d_dw), a single-commit increment over b10018.

New features

  • SYCL backend now supports the FP16 kernel type for the conv2d_dw depthwise convolution operation.

Improvements

  • Improved precision handling for depthwise convolution on SYCL devices via FP16 kernels.

Upgrade notes

This is a narrow SYCL-only increment; no config, model, or API changes are introduced. Users on non-SYCL backends see no functional difference and can upgrade without action. SYCL users running depthwise convolution workloads may benefit from FP16 precision.

Analyzed Oct 30, 2026

Upgrading?

Release notes summary

Adds support for FP16 kernel types in SYCL for depthwise convolution operations, enhancing precision handling.

Key Highlights

  • SYCL support for FP16 kernel type in `conv2d_dw`
  • Standard release binaries for all supported platforms

Release Notes (excerpt)

<details open> sycl : support kernel type fp16 for conv2d_dw (#25653) </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:**…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories