Qeravio
Canonical AI event

Hex-DMA adds direct-mapped cache for HVX FA mask handling

The new direct-mapped DMA cache improves HVX FA mask handling. It supports multiple platforms, including Ubuntu, Windows, and Android.

23 Sept 20261 verified claims1 sources1 observations
What happened

The official source reports this update: b11118: hex-dma: introduce direct-mapped DMA cache that is better suited for HVX FA mask handling.

Why it matters

This official update documents a development concerning b11118: hex-dma: introduce direct-mapped DMA cache that is better suited for HVX FA mask handling. 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
infrastructureLlama.cpp server now routes all model loads through a queue23 Sept 2026infrastructureLlama.cpp disables log file passing to children in server23 Sept 2026infrastructureLlama.cpp server now supports input_image in function_call_output.23 Sept 2026infrastructureLlama.cpp adds sanity checks, disables macOS Intel build23 Sept 2026infrastructureLlama.cpp updates reject MUL_MAT_ID for F32 src1 precision22 Sept 2026infrastructureggml updates IQ1_M to build prefix sums once per block.22 Sept 2026
Evidence trail

Sources behind the event

llama.cpp official releasesb11118: hex-dma: introduce direct-mapped DMA cache that is better suited for HVX FA mask handling
Editorial presentation

Read the public story

Open story