> ## Content Index
> Fetch the complete content index at: https://orbiorbi.com/news/llms.txt
> Use this file to discover other available public pages before exploring further.

# llama.cpp b11721 fuses a CUDA HC_POST operation chain
- URL: https://orbiorbi.com/news/en/hot/llama-cpp-b11721-cuda-hc-post-fusion-en/
- Published: 2026-10-18T18:08:15.000Z
- Updated: 2026-10-18T23:00:40.000Z
- Description: According to ggml, llama.cpp b11721 fuses qwen4exp's SCALE, SIGMOID, SCALE and hc_post chain into one CUDA kernel instead of four. The change ports the Vulkan HC_POST_GATE fusion approach.
- Author: Orbis News
- Tags: #orbis-hot, #en

According to ggml, llama.cpp b11721 fuses qwen4exp's SCALE, SIGMOID, SCALE and hc\_post chain into one CUDA kernel instead of four. The change ports the Vulkan HC\_POST\_GATE fusion approach.

## Sources and originals

- \[1\] ggml · 2026-10-19 · [Read original](https://github.com/ggml-org/llama.cpp/releases/tag/b11721?ref=orbiorbi.com)  
The notes describe fusing qwen4exp's HC\_POST operation chain into one CUDA kernel.

{"version":2,"hot":{"event\_id":"llama-cpp-b11721-cuda-hc-post-fusion","group":"open-source","count":1,"heat":20,"verified":"confirmed","firstSeen":"2026-10-18T18:08:15.000Z","lastSeen":"2026-10-18T23:00:30.415Z","sources":\[{"n":"ggml","publisher":"ggml","u":"https://github.com/ggml-org/llama.cpp/releases/tag/b11721","say":"The notes describe fusing qwen4exp's HC\_POST operation chain into one CUDA kernel.","tier":"T0","at":"2026-10-18T18:08:15.000Z"}\],"signals":\[\],"facts":\[\]}}