Qeravio
Canonical AI event

Vulkan adds IQ3_S MMQ matmul kernels, changes block_a_to_shmem to 2-byte loads.

Vulkan update includes new IQ3_S MMQ matmul kernels for optimized performance.

18 Sept 20261 verified claims1 sources2 observations
What happened

The official source reports this update: b11035: vulkan: add IQ3_S MMQ matmul kernels.

Why it matters

This official update documents a development concerning b11035: vulkan: add IQ3_S MMQ matmul kernels. Its practical significance depends on the scope and evidence stated by the source.

What to watch next

Read the official source update and verify its stated scope, evidence and timing before acting on it.

Connected knowledge

Entities affected by this event

Llama Cppsubject
Continue this topic
infrastructureVulkan raises mul_mat_id row-id limit to 1024 experts18 Sept 2026infrastructureLlama.cpp build fixed for CUDA-only configurations.18 Sept 2026infrastructureLlama.cpp disables macOS Apple Silicon with KleidiAI support18 Sept 2026infrastructureNemotron MTP support extended to multiple platforms18 Sept 2026infrastructureOpenVINO updated to version 2026.4 with compiler warning fixes.18 Sept 2026infrastructureLlama.cpp project refactors Vulkan code into separate files.18 Sept 2026
Evidence trail

Sources behind the event

llama.cpp official releasesb11035: vulkan: add IQ3_S MMQ matmul kernelsllama.cpp official releasesb11035: vulkan: add IQ3_S MMQ matmul kernels
Editorial presentation

Read the public story

Open story