b10180

ggml-org/llama.cppb10180Jul 29, 2026by github-actions[bot]same day as b10179

What changed since b10179

low risk

SYCL backend gains a contiguous fast path and 32-bit index math for unary elementwise ops.

Improvements

  • SYCL unary elementwise operations now take a contiguous fast path when tensors are laid out contiguously.
  • Elementwise index math in the SYCL backend switched to 32-bit arithmetic, with fastdiv used for index computation.
  • Reduced indexing overhead for SYCL unary ops, improving throughput on Intel GPU workloads.

Upgrade notes

This is a single-commit incremental build with no API, config, or model-format changes; existing invocations and scripts carry over unchanged. Only SYCL users are affected, and they should see faster unary elementwise kernels with identical numerical behavior. No migration steps are required.

Analyzed Oct 31, 2026

Upgrading?

Release notes summary

Optimizes the SYCL backend by implementing a contiguous fast path and using 32-bit index math for unary elementwise operations.

Key Highlights

  • SYCL contiguous fast path implementation
  • 32-bit index math for unary elementwise ops
  • Performance optimization for SYCL

New Features

  • SYCL backend optimizations

Release Notes (excerpt)

<details open> sycl: contiguous fast path + 32-bit index math for unary elementwise ops (#25946) * sycl: contiguous fast path + 32-bit index math for unary elementwise ops * sycl: use fastdiv for elementwise index math </details> **Website:** - <https://llama.app> **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)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories