6th Championship Branch Prediction (CBP2025) 1st Place

Award winners

Assistant Professor KOIZUMI Toru
Mr.MAEKAWA Toshiki (Master Course, Program of Networks)
Mr.MIZUNO Masanari (Master Course, Program of Networks)
Professor TSUMURA Tomoaki

Related Website of the winner

Tsumura and Koizumi Lab.(Link to Japanese website)

Related Website

6th Championship Branch Prediction (CBP2025)New Window

Outline

At CBP2025, held in conjunction with ISCA2025, a top conference in the field of computer architecture, a team led by Assistant Professor KOIZUMI won first place.

CBP (Championship Branch Prediction) is an academic competition pitting algorithms for branch prediction, which has become increasingly important in recent high-performance CPUs. The branch predictor called TAGE, which won the previous CBP, and its derivatives have had a major industrial impact, including being incorporated into practical commercial processors. CBP is an important competition that will determine the future of branch predictors.

This time, they proposed a branch predictor that focuses on introducing a completely new method for capturing the correlation between register values and branch directions, and combines this with various improvements. Eighteen teams from around the world participated in CBP2025, and the team led by Assistant Professor KOIZUMI won by a large margin over the second place team.