PulseAugur
EN
LIVE 11:48:18

llama.cpp b12433 release fixes AMD MUL_MAT_ID aborts

The llama.cpp project has released version b12433, addressing a specific issue with the MUL_MAT_ID operation on AMD devices. This fix ensures that the operation correctly handles matrices with an odd number of columns when using formats like bfloat16, preventing aborts that occurred with certain LoRA configurations. The update, which utilized GitHub Copilot for assistance, also includes broader compatibility and testing across various operating systems and hardware architectures. AI

IMPACT Improves stability and performance for users running llama.cpp on AMD hardware.

RANK_REASON Software release addressing a specific bug in an open-source project.

Read on llama.cpp — Releases →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

llama.cpp b12433 release fixes AMD MUL_MAT_ID aborts

How we ranked this

Signal score
20 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
Software release addressing a specific bug in an open-source project.
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
infra, other
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. llama.cpp — Releases TIER_1 English(EN) · liminfei-amd ·

    b12433: cuda : fix MUL_MAT_ID abort with odd src0 columns on AMD (#32385)

    <p>On AMD devices MUL_MAT_ID with an F32, F16 or BF16 src0 and at most 8 tokens always used MMVF.<br /> MMVF reads src0 in pairs of values and asserts an even ne[0], so a rank 1 LoRA on MoE expert weights (lora_b has ne[0] == 1) aborted.<br /> Use MMVF there only for an even ne[0…