b9715

ggml-org/llama.cppb9715Jun 19, 2026by github-actions[bot]same day as b9714

What changed since b9714

low risk

b9715 adds a CUDA backend for the COL2IM_1D op, extending the existing CPU implementation to GPU users.

New features

  • CUDA backend now implements GGML_OP_COL2IM_1D, mirroring the CPU operator added earlier.

Improvements

  • Index decomposition in the CUDA COL2IM_1D kernel uses fast_div_modulo for faster address computation.
  • supports_op checks were tightened to require matching types and a contiguous destination tensor.

Upgrade notes

This is a routine build-to-build bump with no config or API changes. CUDA users gain GPU acceleration for COL2IM_1D, so rebuild from source or pull the new CUDA binaries to benefit. Non-CUDA backends are unaffected.

Analyzed Oct 24, 2026

Upgrading?

Release notes summary

Implements the CUDA backend for the COL2IM_1D operation, matching the CPU implementation with optimized index decomposition.

Key Highlights

  • Added CUDA support for GGML_OP_COL2IM_1D
  • Optimized index decomposition using fast_div_modulo
  • Tightened type matching and support checks

Release Notes (excerpt)

<details open> Ggml/cuda col2im 1d (#24417) * cuda: add GGML_OP_COL2IM_1D, follow-up to the CPU op * cuda: col2im_1d use fast_div_modulo for the index decomposition * cuda: col2im_1d tighten supports_op, type match and contiguous dst </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)…

Published by ggml-org on GitHub.

Read the full release notes

Related Repositories