Oossa

llama.cpp v0.13 adds extended DMA support for faster binary ops

The open‑source llama.cpp library released a new version on Oct 13 2026 that expands DMA broadcasting for binary operations across multiple platforms.

NoteBy Published by Oossa: 1 min read

The ggml‑org team published llama.cpp release b11580 on Oct 13 2026. The update adds extended scalar DMA (direct memory access) support for binary operations, including new kernels for repeat and complex data. It also cleans up non‑DMA paths and adds support for non‑contiguous strides. The changes affect CPU, CUDA, Vulkan and Hexagon NPU builds on Linux, Windows, macOS and Android. Developers can now run more efficient tensor calculations on devices ranging from Apple Silicon to Snapdragon chips.

Why it matters

Developers can run llama.cpp models faster on a wider range of hardware, reducing inference time for end‑users.

Was this article useful?
Share

Read next

Oossa · Newsletter

The week in AI, explained

Every Monday: the stories worth knowing, in plain language. Free, no spam.