Move MBTB and BTBTAGE to S0 for Evaluation

[Yakkhini]

Summary by CodeRabbit

• Chores
  • Adjusted default configuration parameters for branch predictor components to align with baseline simulation settings.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: e10eaffa-1cf9-4102-86e8-e30c593743ae

📥 Commits

Reviewing files that changed from the base of the PR and between 648e246 and f626d9a.

📒 Files selected for processing (1)

• src/cpu/pred/BranchPredictor.py

---

📝 Walkthrough

Walkthrough

In src/cpu/pred/BranchPredictor.py, the numDelay default parameter for both the MBTB and BTBTAGE branch predictor configurations is changed from 2 to 0.

Changes

numDelay Default Reset for MBTB and BTBTAGE

Layer / File(s)	Summary
MBTB and BTBTAGE numDelay defaults src/cpu/pred/BranchPredictor.py	MBTB.numDelay (line 990) and BTBTAGE.numDelay (line 1068) default values are each changed from 2 to 0.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~2 minutes

Poem

🐇 Two little delays, set to naught,
No waiting around as once was thought.
MBTB and BTBTAGE both agree,
Zero is the number, quick and free!
hop hop — the branch predictor flies with glee!

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'Move MBTB and BTBTAGE to S0 for Evaluation' accurately describes the main change: moving predictor configurations to S0 by changing numDelay parameters from 2 to 0.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch mbtb-tage-s0-eval

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[github-actions]

🚀 Coremark Smoke Test Results

Branch	IPC	Change
Base (xs-dev)	2.3155	-
This PR	2.3517	📈 +0.0362 (+1.56%)

✅ Difftest smoke test passed!

[XiangShanRobot]

[Generated by GEM5 Performance Robot]
commit: f626d9a
workflow: gem5 Ideal BTB Performance Test

Ideal BTB Performance

Overall Score

	PR	Master	Diff(%)
Score	21.85	21.26	+2.79 🟢

[Yakkhini]

[Generated by GEM5 Performance Robot] commit: f626d9a workflow: gem5 Ideal BTB Performance Test

Ideal BTB Performance

Overall Score

PR Master Diff(%)
Score 21.85 21.26 +2.79 🟢

It shows that on current implementation there is about 0.6 score per GHz increase when take MainBTB and BTBTAGE as S0 predictor. Maybe it's a abnormal phenomenon.

[Yakkhini]

Baseline: https://github.com/OpenXiangShan/GEM5/actions/runs/27937852410