# Matrix Multiplication 10x+ Speedup

**URL:** <https://discourse.julialang.org/t/matrix-multiplication-10x-speedup/69452>\
**Category:** Performance\
**Tags:** announcement, machine-learning\
**Created:** [October 8, 2021, 8:39pm UTC](https://discourse.julialang.org/t/matrix-multiplication-10x-speedup/69452 "2021-10-08T20:39:14Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![laut](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/laut/32/15651_2.png) [@laut](https://discourse.julialang.org/u/laut)\
**Post date:** [October 8, 2021, 8:39pm UTC](https://discourse.julialang.org/t/matrix-multiplication-10x-speedup/69452/1 "2021-10-08T20:39:14Z")

</div>

Researchers at MIT’s [Computer Science & Artificial Intelligence Lab](https://www.csail.mit.edu/) (CSAIL) have open-sourced [Multiply-ADDitioN-lESS](https://arxiv.org/abs/2106.10860) (MADDNESS), an algorithm that speeds up machine learning using approximate matrix multiplication (AMM). MADDNESS requires zero multiply-add operations and runs 10x faster than other approximate methods and 100x faster than exact multiplication.

> **[Multiplying Matrices Without Multiplying](https://arxiv.org/abs/2106.10860)**
>
> Multiplying matrices is among the most fundamental and compute-intensive operations in machine learning. Consequently, there has been significant work on efficiently approximating matrix multiplies. We introduce a learning-based algorithm for this...

> **[GitHub - dblalock/bolt: 10x faster matrix and vector operations](https://github.com/dblalock/bolt)**
>
> 10x faster matrix and vector operations. Contribute to dblalock/bolt development by creating an account on GitHub.

---

<div class="post-metadata">

**Author:** ![ToucheSir](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/touchesir/32/14411_2.png) [@ToucheSir](https://discourse.julialang.org/u/ToucheSir)\
**Post date:** [October 8, 2021, 9:22pm UTC](https://discourse.julialang.org/t/matrix-multiplication-10x-speedup/69452/2 "2021-10-08T21:22:12Z")

</div>

The first author was gracious enough to answer questions about and detail the (many, many) caveats of their approach on [HN](https://news.ycombinator.com/item?id=28375096). AIUI, this is not going to be a drop-in replacement for BLAS or anything anytime soon.