ASTRAGAL
A
STR
assen-like A
lgorithm for G
eneral mA
trix multiplication with peeL
ing
This repository contains a (tentatively) high-performance, non-SIMD implementation of a Strassen-like recursive-partitioning-based algorithm for general matrix multiplication.
More in detail, the following approaches are exploited:
- Binary block-wise partitioning over even dimensions;
- Peeling as the even-ification strategy of choice.
The peeling implementation closely follows that described in Huss-Lederman et al., 1996.
This work has been originally produced in partial fulfilment of the requirements for the 2nd part of the course on Advanced Programming and Algorithmic Design held at the University of Trieste (Spring Semester, 2020) by prof. Alberto Casagrande.
Furthermore, such implementation is derived from original work by prof. Alberto Casagrande, released under the MIT License.