Dr. Daniel A. Jiménez, professor in the Department of Computer Science and Engineering at Texas A&M University, has received the High-Performance Computer Architecture (HPCA) Test of Time (ToT) Award for his 2001 paper titled “Dynamic Branch Prediction with Perceptrons.”
This prestigious award is given at most to one paper from the Institute of Electrical and Electronics Engineers (IEEE) International Symposium on HPCA, whose influence is still felt 18-22 years after its initial publication. This is only the second HPCA ToT Award to be given, and the first year that papers from 2001 were eligible for the award, heightening this honor.
Jiménez co-authored the winning paper with his former doctoral advisor from The University of Texas at Austin, Dr. Calvin Lin. Their interdisciplinary research between computer architecture and machine learning presented a new method for how microprocessors run computer programs. Of the 225 papers considered, theirs was selected for how it fundamentally changed research into branch prediction.
Prior to their research, branch predictors used ad hoc techniques that made inefficient use of processor resources. Jiménez used his background in neural networks to apply perceptrons to branch prediction. His method of applying neural learning directly into the hardware has significantly improved performance, achieving accuracy superior to previous state-of-the-art techniques.
According to Google Scholar, the paper is the most highly cited on branch
“It is rare for a hardware technique proposed in
Jiménez and Lin accepted their award and delivered a talk at the HPCA conference on Feb. 19 in Washington, D.C.