Digital Design & Comp. Arch: L17: Branch Prediction (Spring 2026)

Digital Design & Comp. Arch: L17: Branch Prediction (Spring 2026)

by Onur Mutlu

Formal & Physical Sciences Computing & Cybersecurity UYComputer scienceUYFComputer architecture and logic design
🎙 Onur Mutlu 👥 64K 📅 April 24, 2026 ⏱ 107 min 👁 1K 📄 lecture 🧭 2026-08-15
Available in: English (current) Français

Keywords

branch predictionBTBdirection predictorglobal historystatic prediction

Summary

This lecture, part of the Digital Design and Computer Architecture course at ETH Zürich, focuses on branch prediction techniques in modern processors. Professor Onur Mutlu begins by motivating the importance of branch prediction, showing how mispredictions degrade performance significantly in deeply pipelined, wide-issue processors. He reviews the basics of branch prediction, including the branch target buffer (BTB) and direction predictors. The lecture then covers static prediction schemes such as always not-taken, always taken, backward-taken/forward-not-taken, and profile-based hints. It transitions to dynamic predictors, emphasizing the use of global branch history to improve accuracy. The discussion includes the motivation for hybrid predictors and mentions advanced techniques like perceptron predictors and geometric history length predictors used in real processors. The lecture concludes with a brief overview of the Alpha 21264’s tournament predictor as an example of a hybrid approach. Throughout, Mutlu highlights the trade-offs between prediction accuracy, hardware complexity, and cycle time, and stresses the importance of keeping the pipeline full.

158 words

Critical Evaluation

Value of the Information & Strength of the Argument

The lecture provides a comprehensive and well-structured overview of branch prediction, from basic static schemes to advanced dynamic predictors. The value lies in its clear explanation of the problem and the incremental development of solutions, supported by quantitative examples (e.g., IPC degradation with misprediction rates). The argumentation is solid, grounded in established research and real processor implementations. Mutlu effectively motivates each technique by explaining its rationale and limitations, and he connects the concepts to practical design considerations such as cycle time and hardware budget. The inclusion of advanced topics like perceptron predictors and hybrid designs demonstrates the depth of the content.

Scientific Rigor, Source Quality, Title Accuracy

The lecture is scientifically rigorous, based on well-known research papers and textbook material. Mutlu references seminal works such as the Alpha 21264 branch predictor and mentions the ‘Branch Prediction for Free’ paper by Ball and Larus. The sources cited in the description are relevant and authoritative, including papers on RowHammer and memory-centric computing, though they are not directly related to branch prediction. The title accurately reflects the content, and the lecture is well-organized with clear slides and examples. The technical depth is appropriate for an advanced undergraduate or graduate course, and the presentation is precise without oversimplification.

213 words

Title / Content Match

The title accurately reflects the content, which is a lecture on branch prediction techniques in computer architecture.

Quality & Reliability

9/10

Lecture by a renowned professor in computer architecture, based on established research and textbook material. The content is technically accurate and well-structured, with references to seminal papers and real processor implementations.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

Concurring Sources

  • A Survey of Techniques for Dynamic Branch Prediction — Provides background on dynamic branch prediction techniques.
  • The Alpha 21264 Microprocessor Architecture — Describes the hybrid branch predictor used in the Alpha 21264.

External References

Contribution & Novelties

The lecture provides a comprehensive and up-to-date overview of branch prediction techniques, from classic static schemes to advanced dynamic predictors. It emphasizes the importance of branch prediction in modern processors and discusses the trade-offs between accuracy, hardware complexity, and cycle time. The inclusion of advanced topics like perceptron predictors and hybrid designs, as well as references to recent research, adds value for students and practitioners.

Pour aller plus loin :

  • Branch Prediction (Wikipedia) — Overview of branch prediction techniques and history.
  • The Alpha 21264 Microprocessor Architecture — Detailed description of the Alpha 21264’s tournament predictor.
  • Perceptron-Based Branch Prediction — Original paper on perceptron predictors.
  • TAGE Predictor — Paper on TAGE, a state-of-the-art hybrid predictor.

114 words

Radar Profile

The radar profile shows high scores across all dimensions, indicating a well-balanced and comprehensive lecture. The strong scores in information quantity and quality reflect the depth and accuracy of the content, while the technical level is appropriate for an advanced audience. The overall reliability is high, consistent with the lecturer's expertise and the use of established research.

Reliability 9/10

💬 No comments were provided for analysis.