DGEMM performance is data-dependent

TL;DR

This paper demonstrates that DGEMM performance varies significantly with matrix content, influenced by CPU bit flips causing energy overhead, challenging the assumption that performance depends solely on matrix sizes.

Contribution

It reveals the impact of matrix content and CPU bit flips on DGEMM performance, highlighting factors beyond asymptotic complexity.

Findings

Performance varies with matrix content

CPU bit flips cause energy overhead

Content-dependent performance impacts optimization

Abstract

The DGEMM function is a widely used implementation of the matrix product. While the asymptotic complexity of the algorithm only depends on the sizes of the matrices, we show that the performance is significantly impacted by the matrices content. Our experiments show that this may be due to bit flips in the CPU causing an energy consumption overhead.