Pair tage 2 taken by Yakkhini · Pull Request #830 · OpenXiangShan/GEM5

Yakkhini · 2026-04-15T02:30:55Z

Summary by CodeRabbit

New Features
- Added PairTAGE two-block branch predictor with configurable phase/second-block behavior and optional integration into the decoupled BPU.
- Added side-effect-free BTB lookup paths to support safe, non-mutating prediction queries.
Improvements
- Added prediction-metadata refresh hooks across BTB/TAGE predictors and extended fetch-target metadata for pair-aware predictions.
- Improved const-correctness and expanded statistics for pair/second-block workflows.

coderabbitai · 2026-04-15T02:31:03Z

Note

Reviews paused

It looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the reviews.auto_review.auto_pause_after_reviewed_commits setting.

Use the following commands to manage reviews:

@coderabbitai resume to resume automatic reviews.
@coderabbitai review to trigger a single review.

Use the checkboxes below for quick actions:

▶️ Resume reviews
🔍 Trigger review

📝 Walkthrough

Walkthrough

Adds a new two-block PairTAGE BTB/TAGE predictor and integrates it into DecoupledBPUWithBTB; introduces TimedBaseBTBPredictor::refreshPredictionMeta, side-effect-free BTB lookup APIs and const-correct helpers across BTB predictors, extends FetchTarget with PairTAGE fields, and adds FTQ const accessors.

Changes

Cohort / File(s)	Summary
PairTAGE implementation `src/cpu/pred/btb/pairtage.hh`, `src/cpu/pred/btb/pairtage.cc`	New TimedBaseBTBPredictor subclass implementing two-block entries, folded-history meta (TageMeta), phase gating, first/second-block prediction assembly, refreshPredictionMeta, training/update, allocation, history recoveries, and statistics.
Predictor wiring & build `src/cpu/pred/BranchPredictor.py`, `src/cpu/pred/SConscript`	Adds `PairTAGE` SimObject/Params and exposes `DecoupledBPUWithBTB.pairtage`; registers pairtage source file in build.
Decoupled BPU integration & stats `src/cpu/pred/btb/decoupled_bpred.cc`, `src/cpu/pred/btb/decoupled_bpred.hh`, `src/cpu/pred/btb/decoupled_bpred_stats.cc`	Integrates PairTAGE into tick/enqueue/squash flows, per-thread pairPhase/state, createFetchTarget overloads, second-block teacher/student flow, history update API changes, and extensive PairTAGE-specific stats.
No-side-effect lookup & meta refresh (BTB family) `src/cpu/pred/btb/abtb.{cc,hh}`, `mbtb.{cc,hh}`, `btb_tage.{cc,hh}`, `btb_ittage.{cc,hh}`, `btb_mgsc.{cc,hh}`, `btb_ubtb.{cc,hh}`, `microtage.{cc,hh}`, `ras.{cc,hh}`	Adds const `lookupNoSideEffect`/`processEntriesNoSideEffect` variants and `refreshPredictionMeta` implementations to recompute FullBTBPrediction metadata without mutating predictor state or stats; tightens const-correctness of many helper methods.
Common BTB types & FetchTarget `src/cpu/pred/btb/common.hh`	Adds `PairPhase` enum; expands `FetchTarget` with `pairPhase`, `pairtageUsed`, `pairtageSecondBlock`; increases `predMetas` capacity (8→12); initializes FullBTBPrediction defaults.
Timed base predictor hook `src/cpu/pred/btb/timed_base_pred.hh`	Adds virtual `refreshPredictionMeta(Addr, const boost::dynamic_bitset<>&, FullBTBPrediction&)` hook (default no-op).
FTQ const accessors `src/cpu/pred/btb/ftq.hh`	Adds const overloads for `front`, `back`, and `get` to return `const FetchTarget&`.

Sequence Diagram(s)

sequenceDiagram
    participant Fetch as Fetch Unit
    participant DBPU as DecoupledBPUWithBTB
    participant PT as PairTAGE
    participant FTQ as FetchTargetQueue
    participant Hist as History Tracker

    Fetch->>DBPU: request prediction(startPC, history)
    DBPU->>PT: putPCHistory(startPC, history)
    PT->>PT: TAGE lookup -> build first-block FullBTBPrediction
    PT-->>DBPU: first-block FullBTBPrediction (pairtageUsed, pairPhase)
    DBPU->>FTQ: enqueue first FetchTarget (pairPhase)

    alt PairTAGE enabled && phase allowed && second-block candidate
        DBPU->>PT: refreshPredictionMeta(secondStartPC, history, pred)
        PT->>PT: lookupNoSideEffect -> second-block prediction/meta
        PT-->>DBPU: second-block info (pairtageSecondBlock)
        DBPU->>FTQ: conditionally enqueue second FetchTarget
    end

    Note over DBPU,PT: On resolution
    DBPU->>PT: trainFromActualPred(firstFetchTarget, optionalSecondPred)
    PT->>Hist: specUpdateHist / recoverHist / doUpdateHist

sequenceDiagram
    participant Caller as DecoupledBPU
    participant Predictor as BTB Predictor
    participant NoSE as NoSideEffect Lookup
    participant Meta as FullBTBPrediction

    Caller->>Predictor: refreshPredictionMeta(startAddr, history, pred)
    Predictor->>NoSE: lookupNoSideEffect(...) / processEntriesNoSideEffect(...)
    NoSE->>NoSE: tag match, select entries (no MRU/LRU/stats changes)
    NoSE-->>Meta: populate hit_entries / per-PC preds
    Meta-->>Caller: refreshed prediction metadata snapshot

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~70 minutes

Possibly related PRs

cpu-o3: update tage index #627 — BTBTAGE const-correctness and prediction/refresh API changes overlap with TAGE interface and meta-refresh work.
cpu-o3:Add S3 prediction update AheadBTB #629 — AheadBTB modifications touch no-side-effect lookup and ahead-path behavior introduced here.
cpu-o3: remove simple functions in fetch #725 — DecoupledBPUWithBTB changes overlap with fetch/FTQ/second-block integration and state bookkeeping.

Suggested reviewers

jensen-yan
CJ362ff

"🐇
I hop on tags both even and odd,
Two blocks foretold, each stitch a nod.
I peek without a single side effect,
I learn when true and gently correct.
A rabbit's hop: predict, teach, connect."

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (1 warning, 1 inconclusive)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 18.39% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.
Title check	❓ Inconclusive	The title 'Pair tage 2 taken' is vague and does not clearly convey the main change. It appears to reference PairTAGE but lacks specificity about what 'taken' means in this context, making it unclear to reviewers scanning commit history.	Consider a more descriptive title that clearly indicates the main functionality being added or modified, such as 'Add PairTAGE predictor with two-block prediction' or 'Implement PairTAGE for dual-block branch prediction.'

✅ Passed checks (3 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
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 unit tests (beta)

Create PR with unit tests
Commit unit tests in branch pair-tage-2-taken

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

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

github-actions · 2026-04-15T03:55:08Z

🚀 Coremark Smoke Test Results

Branch	IPC	Change
Base (`xs-dev`)	`2.3051`	-
This PR	`2.3051`	➡️ `0.0000` (`0.00%`)

✅ Difftest smoke test passed!

github-actions · 2026-04-16T10:23:02Z

🚀 Coremark Smoke Test Results

Branch	IPC	Change
Base (`xs-dev`)	`2.3051`	-
This PR	`2.3051`	➡️ `0.0000` (`0.00%`)

✅ Difftest smoke test passed!

github-actions · 2026-04-16T11:52:47Z

🚀 Performance test triggered: gcc15-spec06-0.8c

github-actions · 2026-04-16T16:06:51Z

🚀 Performance test triggered: gcc15-spec06-0.8c

github-actions · 2026-04-16T16:17:43Z

🚀 Performance test triggered: gcc15-spec06-0.8c

github-actions · 2026-04-16T16:18:10Z

🚀 Coremark Smoke Test Results

Branch	IPC	Change
Base (`xs-dev`)	`2.3051`	-
This PR	`1.9659`	📉 `-0.3392` (`-14.72%`)

✅ Difftest smoke test passed!

github-actions · 2026-04-21T09:35:54Z

🚀 Coremark Smoke Test Results

Branch	IPC	Change
Base (`xs-dev`)	`2.3051`	-
This PR	`2.3051`	➡️ `0.0000` (`0.00%`)

✅ Difftest smoke test passed!

github-actions · 2026-04-22T14:39:08Z

🚀 Coremark Smoke Test Results

Branch	IPC	Change
Base (`xs-dev`)	`2.3051`	-
This PR	`2.1343`	📉 `-0.1708` (`-7.41%`)

✅ Difftest smoke test passed!

github-actions · 2026-04-22T14:51:45Z

🚀 Performance test triggered: gcc15-spec06-0.8c

github-actions · 2026-04-22T15:39:49Z

🚀 Coremark Smoke Test Results

Branch	IPC	Change
Base (`xs-dev`)	`2.3051`	-
This PR	`2.2801`	📉 `-0.0250` (`-1.08%`)

✅ Difftest smoke test passed!

github-actions · 2026-04-22T15:41:18Z

🚀 Performance test triggered: gcc15-spec06-0.8c

github-actions · 2026-04-23T04:28:21Z

🚀 Coremark Smoke Test Results

Branch	IPC	Change
Base (`xs-dev`)	`2.3051`	-
This PR	`2.2875`	📉 `-0.0176` (`-0.76%`)

✅ Difftest smoke test passed!

github-actions · 2026-04-23T04:30:51Z

🚀 Performance test triggered: gcc15-spec06-0.8c

Change-Id: Ib124277cb25cd90e4a2d75ed089c18fbe1e56db2

github-actions · 2026-05-07T08:33:32Z

🚀 Coremark Smoke Test Results

Branch	IPC	Change
Base (`xs-dev`)	`2.3130`	-
This PR	`2.3029`	📉 `-0.0101` (`-0.44%`)

✅ Difftest smoke test passed!

github-actions · 2026-05-07T08:36:37Z

🚀 Performance test triggered: gcc15-spec06-0.8c

XiangShanRobot · 2026-05-07T10:56:28Z

[Generated by GEM5 Performance Robot]
commit: c8c0c3e
workflow: On-Demand SPEC Test (Tier 1.5)

Ideal BTB Performance

Overall Score

	PR	Master	Diff(%)
Score	21.59	21.54	+0.23 🟢

Change-Id: I0aa0122de316356e918b7002710c3a44921ede35

github-actions · 2026-05-07T13:05:13Z

🚀 Coremark Smoke Test Results

Branch	IPC	Change
Base (`xs-dev`)	`2.3130`	-
This PR	`2.3368`	📈 `+0.0238` (`+1.03%`)

✅ Difftest smoke test passed!

github-actions · 2026-05-07T13:11:59Z

🚀 Performance test triggered: gcc15-spec06-0.8c

XiangShanRobot · 2026-05-07T15:31:54Z

[Generated by GEM5 Performance Robot]
commit: 4126e7b
workflow: On-Demand SPEC Test (Tier 1.5)

Ideal BTB Performance

Overall Score

	PR	Master	Diff(%)
Score	22.13	21.54	+2.74 🟢

[Generated by GEM5 Performance Robot]
commit: 4126e7b
workflow: On-Demand SPEC Test (Tier 1.5)

Ideal BTB Performance

Overall Score

	PR	Previous Commit	Diff(%)
Score	22.13	21.59	+2.50 🟢

github-actions · 2026-05-14T10:59:04Z

🚀 Coremark Smoke Test Results

Branch	IPC	Change
Base (`xs-dev`)	`2.3130`	-
This PR	`2.3368`	📈 `+0.0238` (`+1.03%`)

✅ Difftest smoke test passed!

github-actions · 2026-05-15T02:49:40Z

🚀 Coremark Smoke Test Results

Branch	IPC	Change
Base (`xs-dev`)	`2.3130`	-
This PR	`2.3342`	📈 `+0.0212` (`+0.92%`)

✅ Difftest smoke test passed!

github-actions · 2026-05-15T02:55:20Z

🚀 Performance test triggered: gcc15-spec06-0.8c

Change-Id: I9cb025867777ca70a66d2048282cfabf76b4bb99

github-actions · 2026-05-15T08:59:07Z

🚀 Performance test triggered: gcc15-spec06-0.8c

github-actions · 2026-05-15T09:07:39Z

🚀 Coremark Smoke Test Results

Branch	IPC	Change
Base (`xs-dev`)	`2.3130`	-
This PR	`2.3391`	📈 `+0.0261` (`+1.13%`)

✅ Difftest smoke test passed!

Change-Id: Id119399248485e79a7b757200d4f52a01e628877

github-actions · 2026-05-26T07:39:00Z

🚀 Coremark Smoke Test Results

Branch	IPC	Change
Base (`xs-dev`)	`2.3130`	-
This PR	`2.3391`	📈 `+0.0261` (`+1.13%`)

✅ Difftest smoke test passed!

Change-Id: I7c545a3c6f2525e932a0c4f9d063d2b6932d3058

github-actions · 2026-05-26T08:02:12Z

🚀 Coremark Smoke Test Results

Branch	IPC	Change
Base (`xs-dev`)	`2.3130`	-
This PR	`2.3511`	📈 `+0.0381` (`+1.65%`)

✅ Difftest smoke test passed!

github-actions · 2026-05-26T08:07:01Z

🚀 Performance test triggered: gcc15-spec06-0.8c

Yakkhini added the perf label Apr 16, 2026

Yakkhini added perf and removed perf labels Apr 16, 2026

Yakkhini force-pushed the pair-tage-2-taken branch from 105484a to becfce1 Compare April 21, 2026 08:24

Yakkhini force-pushed the pair-tage-2-taken branch from 6b5465f to 59f9782 Compare April 22, 2026 14:29

Yakkhini added perf and removed perf labels Apr 22, 2026

Yakkhini marked this pull request as ready for review April 23, 2026 04:30

Yakkhini added perf and removed perf labels Apr 23, 2026

cpu: Preserve PairTAGE second-block context

c8c0c3e

Change-Id: Ib124277cb25cd90e4a2d75ed089c18fbe1e56db2

Yakkhini added perf and removed perf labels May 7, 2026

cpu: Age PairTAGE provider identity

4126e7b

Change-Id: I0aa0122de316356e918b7002710c3a44921ede35

Yakkhini added perf and removed perf labels May 7, 2026

Yakkhini added perf and removed perf labels May 15, 2026

cpu-o3: Port ideal 2-fetch

3e56761

Change-Id: I9cb025867777ca70a66d2048282cfabf76b4bb99

Yakkhini force-pushed the pair-tage-2-taken branch from 21b37aa to 3e56761 Compare May 15, 2026 08:58

Yakkhini added perf and removed perf labels May 15, 2026

cpu: Fix PairTAGE multi-way provider selection

bdb68e5

Change-Id: Id119399248485e79a7b757200d4f52a01e628877

cpu: Set PairTAGE default ways to two

591a12e

Change-Id: I7c545a3c6f2525e932a0c4f9d063d2b6932d3058

Yakkhini added perf and removed perf labels May 26, 2026

Conversation

Yakkhini commented Apr 15, 2026 • edited by coderabbitai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary by CodeRabbit

Uh oh!

coderabbitai Bot commented Apr 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Reviews paused

Walkthrough

Changes

Sequence Diagram(s)

Estimated code review effort

Possibly related PRs

Suggested reviewers

❌ Failed checks (1 warning, 1 inconclusive)

Uh oh!

github-actions Bot commented Apr 15, 2026

🚀 Coremark Smoke Test Results

Uh oh!

github-actions Bot commented Apr 16, 2026

🚀 Coremark Smoke Test Results

Uh oh!

github-actions Bot commented Apr 16, 2026

Uh oh!

github-actions Bot commented Apr 16, 2026

Uh oh!

github-actions Bot commented Apr 16, 2026

Uh oh!

github-actions Bot commented Apr 16, 2026

🚀 Coremark Smoke Test Results

Uh oh!

github-actions Bot commented Apr 21, 2026

🚀 Coremark Smoke Test Results

Uh oh!

github-actions Bot commented Apr 22, 2026

🚀 Coremark Smoke Test Results

Uh oh!

github-actions Bot commented Apr 22, 2026

Uh oh!

github-actions Bot commented Apr 22, 2026

🚀 Coremark Smoke Test Results

Uh oh!

github-actions Bot commented Apr 22, 2026

Uh oh!

github-actions Bot commented Apr 23, 2026

🚀 Coremark Smoke Test Results

Uh oh!

github-actions Bot commented Apr 23, 2026

Uh oh!

github-actions Bot commented May 7, 2026

🚀 Coremark Smoke Test Results

Uh oh!

github-actions Bot commented May 7, 2026

Uh oh!

XiangShanRobot commented May 7, 2026

Ideal BTB Performance

Overall Score

Uh oh!

github-actions Bot commented May 7, 2026

🚀 Coremark Smoke Test Results

Uh oh!

github-actions Bot commented May 7, 2026

Uh oh!

XiangShanRobot commented May 7, 2026

Ideal BTB Performance

Overall Score

Ideal BTB Performance

Overall Score

Uh oh!

github-actions Bot commented May 14, 2026

🚀 Coremark Smoke Test Results

Uh oh!

github-actions Bot commented May 15, 2026

🚀 Coremark Smoke Test Results

Uh oh!

github-actions Bot commented May 15, 2026

Uh oh!

github-actions Bot commented May 15, 2026

Uh oh!

github-actions Bot commented May 15, 2026

Yakkhini commented Apr 15, 2026 •

edited by coderabbitai Bot

Loading

coderabbitai Bot commented Apr 15, 2026 •

edited

Loading