# BDCSVD implementation fully in Mojo

**URL:** <https://forum.modular.com/t/bdcsvd-implementation-fully-in-mojo/3504>\
**Category:** Community Showcase\
**Created:** [September 17, 2026, 5:34pm UTC](https://forum.modular.com/t/bdcsvd-implementation-fully-in-mojo/3504 "2026-09-17T17:34:57Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Doby](https://avatars.discourse-cdn.com/v4/letter/d/58f4c7/32.png) [@Doby](https://forum.modular.com/u/Doby)\
**Post date:** [September 17, 2026, 5:34pm UTC](https://forum.modular.com/t/bdcsvd-implementation-fully-in-mojo/3504/1 "2026-09-17T17:34:57Z")

</div>

I did a rewrite of Eigen’s BDCSVD - the “A Divide-And-Conquer Algorithm for the Bidiagonal SVD” research report written by Ming Gu and Stanley C.Eisenstat - C++ implementation in Mojo: [Mojmelo/mojmelo/utils/SVD at main · yetalit/Mojmelo · GitHub](https://github.com/yetalit/Mojmelo/tree/main/mojmelo/utils/SVD)  
I’m not an expert in benchmarking high-performance C++ code, but based on the results I got from the test code compiled with `g++ -O3 -march=native`, the performance appears to be close to the Mojo implementation. Those who are interested can investigate the performance differences further.  
This was done as a part of my Mojmelo project to implement ML algorithms from scratch: [GitHub - yetalit/Mojmelo: Machine Learning algorithms in pure Mojo 🔥 · GitHub](https://github.com/yetalit/Mojmelo)