# llama.cpp moves a delta-net index calculation to the host

> Release b12074 changes how llama.cpp handles delta-net convolution state, supplying a rollback-slot index from the host instead of calculating it on the device.

Oossa · 2026-11-03 · https://oossa.com/en/llama-cpp-moves-a-delta-net-index-calculation-to-the-host

The ggml-org team released llama.cpp version b12074 with changes to delta-net convolution state. The update makes the rollback slot index an I32 input filled on the host, rather than building it on the device with arange and argsort operations.

The release notes say the index is constant for each graph shape and that the change removes the need for the metadata backend to mirror a split for arange. They do not report a measured speed or memory improvement.

## The facts

- Release: llama.cpp b12074, published November 3, 2026.
- The rollback slot index is now filled on the host as an I32 input.

## Why it matters

People running llama.cpp models that use delta-net may encounter a changed internal calculation, but the release notes do not establish whether it improves performance in practice.

## Sources & references

1. [ggml-org/llama.cpp b12074](https://github.com/ggml-org/llama.cpp/releases/tag/b12074) – llama.cpp, 2026-11-03

Last updated: 2026-11-03