b9892

ggml-org/llama.cppb9892Jul 6, 2026by github-actions[bot]same day as b9891

What changed since b9891

low risk

b9892 adds PPC-specific default thread count tuning for Linux and AIX; otherwise a routine build bump from b9891.

New features

  • PowerPC builds on Linux and AIX now pick an optimal default thread count automatically instead of relying on a generic fallback.

Improvements

  • Thread-count defaults are now architecture-aware for PPC, which should improve out-of-the-box throughput on those systems.

Upgrade notes

This is a single-commit delta over b9891, so no config or API migration is required. PPC users on Linux or AIX get better defaults automatically; anyone who explicitly sets thread count via CLI or config is unaffected. Other platforms should see no behavioral difference.

Analyzed Oct 30, 2026

Upgrading?

Release notes summary

Binary release with extensive platform support including macOS, Linux, Android, and Windows. The primary technical change is setting optimal default thread count for PowerPC (PPC) on Linux and AIX systems.

Key Highlights

  • Optimal default thread count for PPC (Linux and AIX)
  • Extensive platform support across multiple OS and architectures
  • Support for various GPU backends including CUDA, Vulkan, ROCm, OpenVINO, SYCL, and HIP

Release Notes (excerpt)

<details open> common: Set optimal default thread count for ppc ( linux as well as AIX) (#25237) </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:**…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories