b9758

ggml-org/llama.cppb9758Jun 22, 2026by github-actions[bot]same day as b9757

What changed since b9757

low risk

b9758 adds BF16 support for binary broadcast and unary ops in the SYCL backend, plus older Intel compiler compatibility.

New features

  • SYCL backend now supports BFloat16 (BF16) for the bin_bcast operation.
  • SYCL backend now supports BF16 for unary operations.

Improvements

  • SYCL builds now work with Intel compilers older than 2026.0, widening the toolchain compatibility range.

Upgrade notes

This is a build-to-build increment with changes confined to the SYCL backend; CUDA, Metal, Vulkan, ROCm, and CPU paths are unaffected. SYCL users on BF16 models should see correct execution of broadcast and unary ops, and those on pre-2026.0 Intel compilers can now build without workarounds. No config or model migration is required.

Analyzed Oct 31, 2026

Upgrading?

Release notes summary

Added BFloat16 (BF16) support for binary broadcast and unary operations in the SYCL backend.

Key Highlights

  • Support BF16 on bin_bcast OP
  • Support BF16 on unary OPs
  • Support older Intel compilers

Release Notes (excerpt)

<details open> [SYCL] support bf16 on bin_bcast OP and unary OPs (#24838) * support bf16 on bin_bcast OP and unary OPs * support the older Intel compiler than 2026.0 </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