t-head/hggc-samples — explained in plain English
Analysis updated 2026-05-18
Learn the HGGC Runtime API by reading and running working sample code.
Check what capabilities and bandwidth a Zhenwu PPU device supports.
See working examples of advanced features like Multi-Device Cooperative Groups and Unified Virtual Memory.
Run a full test sweep across all built samples to sanity check a toolkit installation.
| t-head/hggc-samples | alex-shayo/bakkes-mod-install | localai-org/moss-transcribe.cpp | |
|---|---|---|---|
| Stars | 29 | 29 | 29 |
| Language | C++ | C++ | C++ |
| Setup difficulty | hard | easy | moderate |
| Complexity | 3/5 | 2/5 | 4/5 |
| Audience | developer | general | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires the T-Head SAIL toolkit and compatible Zhenwu PPU hardware to actually run the samples.
HGGC-Samples is a collection of example programs built for people developing on T-Head's HGGC hardware platform, using the T-Head SAIL toolkit version 2.1. If you work with Zhenwu PPU chips (a kind of processor from T-Head), these samples show you how to use the HGGC Runtime API and related features through working code rather than abstract documentation. The samples are grouped by topic, starting with basic introductory examples for beginners, then moving into utility samples for checking device capabilities and measuring bandwidth. Further groups cover common algorithms and problem-solving techniques, general HGGC features, HGGC platform libraries, and finally performance optimization examples. This structure lets a newcomer start simple and work up to more advanced material at their own pace. To use the repository, you clone it with git or download it as a zip file, then build it with CMake (version 3.20 or later) on Linux, using standard commands like mkdir build, cmake, and make. A Python 3 script called run_tests.py is included so you can run all the built samples at once as a quick sanity check, though the README is clear that this is not a full validation suite and does not test edge cases or measure performance formally. Some samples need extra third-party tools to build or run, such as MPI for communication between processes or OpenMP for multiprocess programming, both of which are commonly available through a Linux package manager. Other samples demonstrate specific HGGC features like Multi-Block Cooperative Groups, Multi-Device Cooperative Groups, a runtime compilation library called HGRTC, Unified Virtual Memory for sharing memory between CPU and PPU, and 16-bit floating point support. The project is maintained by T-Head, a semiconductor company, and is released under the Apache License 2.0.
A collection of example C++ programs showing developers how to use T-Head's HGGC hardware platform and SAIL toolkit.
Mainly C++. The stack also includes C++, CMake, Python.
Use, modify, and distribute freely, including for commercial purposes, as long as you keep the copyright and license notice.
Setup difficulty is rated hard, with roughly 1h+ to a first successful run.
Mainly developer.
This repo across BitVibe Labs
double-check against the repo, no cap.