deepseek-ai/DeepEP-Ascend
A high-performance communication library for machine learning training and inference on Huawei Ascend NPUs.
GitHub 
6.8K
lines of main code
69 files
1.6K
lines of test code
10 files
0.5K
lines of other code
7 files
5.1K
hpp
1.6K
py
0.08K
cpp
Rule-of-thumb estimates

Rebuild value

to rebuild
rebuild value

Maintenance effort

per year to maintain
maintenance cost per year
Rules of thumb, not measurements: one man-year rebuilds about 10,000 lines of main code (design, testing and coordination included), and keeping a code base alive takes about 15% of its rebuild value per year (20 man-years of rebuild value need about 3 man-years of maintenance every year). Your changes are remembered in this browser for every report.