b9915

ggml-org/llama.cppb9915Jul 8, 2026by github-actions[bot]same day as b9914

What changed since b9914

low risk

b9915 adds CPU f16->f16 support for GGML_OP_SET_ROWS and consolidates the set_rows implementation.

New features

  • CPU backend now supports the f16->f16 variant of GGML_OP_SET_ROWS.

Improvements

  • f32 and f16 set_rows forward paths are unified into a single shared implementation function.
  • Added missing type checks for the f16 GGML_OP_SET_ROWS path.
  • Standard assert() calls replaced with GGML_ASSERT() for consistent assertion handling.

Upgrade notes

This is a routine incremental build with no config or API changes; existing binaries and model files remain compatible. Users relying on f16 set_rows operations on CPU should see the new path exercised automatically. No migration steps are required.

Analyzed Oct 31, 2026

Upgrading?

Release notes summary

Adds support for f16->f16 GGML_OP_SET_ROWS on the CPU, including type checks and merging implementation functions.

Key Highlights

  • Added support for CPU f16->f16 GGML_OP_SET_ROWS
  • Merged ggml_compute_forward_set_rows_f32() and f16() into a single implementation
  • Replaced standard assert() with GGML_ASSERT()

Release Notes (excerpt)

<details open> ggml : add support for CPU f16->f16 GGML_OP_SET_ROWS (#25344) * ggml : add support for CPU f16->f16 GGML_OP_SET_ROWS * ggml : add missing type checks in f16 GGML_OP_SET_ROWS * ggml : merge ggml_compute_forward_set_rows_f32() and ggml_compute_forward_set_rows_f16() into ggml_compute_forward_set_rows_impl() * chore : replace assert() with GGML_ASSERT() --------- Co-authored-by: Stanisław Szymczyk <sszymczy@gmail.com> </details> **macOS/iOS:** - macOS Apple Silicon (arm64)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories