bpu,configs: Add optional BTBTAGE upper-bound mode for kmhv3 - #770
Merged
Conversation
Change-Id: I3a357ccab4ae078edb30bd4ffe5e01b5c6d9ac7f
Change-Id: I0e029d43a10487cc7090e15973db1fd84a731375
Change-Id: Ibdd30c6049dca43e022e5bd4e94a89e493397c33
Change-Id: If9f2e2a8099ac58c8f9ad5b86b00292eeebc48d4
Change-Id: I06630ca8db53f4fd944355fd03a166203c5dca11
|
Note Reviews pausedIt 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 Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
📝 WalkthroughWalkthroughAdds a new BTBTAGEUpperBound predictor (exact-history, path-hash option), registers it as a SimObject, wires a kmhv3 CLI flag to select it, updates BTBTAGE method signatures, adds tests, documentation, CI perf tooling, and build entries to include the new source. Changes
Sequence Diagram(s)sequenceDiagram
rect rgba(200,200,255,0.5)
participant FetchUnit
end
rect rgba(200,255,200,0.5)
participant BTB
participant BTB_TAGE
end
rect rgba(255,200,200,0.5)
participant BTBTAGE_UB
participant BasePredictor
end
FetchUnit->>BTB: request fetch target (startPC)
BTB->>BTB_TAGE: lookup branch entry
BTB_TAGE->>BTBTAGE_UB: query exact-history prediction (per-table index)
BTBTAGE_UB-->>BTB_TAGE: prediction + meta (main/alt)
BTB_TAGE->>BasePredictor: consult base predictor if UB miss
BasePredictor-->>BTB_TAGE: base prediction
BTB_TAGE-->>FetchUnit: final prediction(s) with metadata
FetchUnit->>BTBTAGE_UB: on resolution -> update()/allocate() (history snapshot)
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Possibly related PRs
Suggested labels
Suggested reviewers
Poem
🚥 Pre-merge checks | ✅ 2 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches
🧪 Generate unit tests (beta)
📝 Coding Plan
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. Comment |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b44d5b169b
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
🧹 Nitpick comments (1)
configs/example/kmhv3.py (1)
115-120: Consider removing redundant MGSC table configuration.These MGSC table enable/disable settings have no effect since
mgsc.enabled = Falseon line 111. You may want to remove them to reduce configuration clutter, or keep them if they serve as documentation for future reference.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@configs/example/kmhv3.py` around lines 115 - 120, The MGSC table enable/disable lines (cpu.branchPred.mgsc.enableBwTable, enableLTable, enableITable, enableGTable, enablePTable, enableBiasTable) are redundant because cpu.branchPred.mgsc.enabled is set to False; remove these per-table settings to reduce clutter, or alternatively leave a single commented note documenting the intended defaults if you want them for future reference—either delete the six cpu.branchPred.mgsc.enable* lines or replace them with a one-line comment referencing mgsc.enabled and the intended table defaults.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Nitpick comments:
In `@configs/example/kmhv3.py`:
- Around line 115-120: The MGSC table enable/disable lines
(cpu.branchPred.mgsc.enableBwTable, enableLTable, enableITable, enableGTable,
enablePTable, enableBiasTable) are redundant because cpu.branchPred.mgsc.enabled
is set to False; remove these per-table settings to reduce clutter, or
alternatively leave a single commented note documenting the intended defaults if
you want them for future reference—either delete the six
cpu.branchPred.mgsc.enable* lines or replace them with a one-line comment
referencing mgsc.enabled and the intended table defaults.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: e4592414-45b1-43ab-88f9-7e915039121a
📒 Files selected for processing (4)
configs/example/kmhv3.pysrc/cpu/pred/BranchPredictor.pysrc/cpu/pred/btb/btb_tage.ccsrc/cpu/pred/btb/btb_tage.hh
Change-Id: Ib8b67365d8196e4b2ec0044a8883e6983b88b2f6
🚀 Coremark Smoke Test Results
✅ Difftest smoke test passed! |
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (4)
docs/Gem5_Docs/frontend/tage_index_capacity_exploration_20260305.md (3)
107-117: Commit-chain section may go stale; link to immutable PR artifacts.The hardcoded short-hash list is useful now but brittle over time. Consider adding direct links to the PR commits (or a permalinked range) so readers can always resolve the exact history.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@docs/Gem5_Docs/frontend/tage_index_capacity_exploration_20260305.md` around lines 107 - 117, The commit list under the section "7. 当前分支提交链（CI触发）" is brittle; replace the hardcoded short-hash entries for branch `bigger-tage-align` with immutable links (full commit SHAs or permalink URLs) pointing to each commit or a single permalinked range/PR so readers can always resolve history; update the five listed items (including the one mentioning `tableSizes`) to use their respective commit URLs or the PR permalink and leave the descriptive text unchanged.
85-94: Prefer a compact table with relative deltas.For quick comparison, present
off/mix2/mix4/mix8as a table and include%deltas vsofffor both IPC and mispred. This will make trend strength and tradeoffs much clearer.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@docs/Gem5_Docs/frontend/tage_index_capacity_exploration_20260305.md` around lines 85 - 94, Replace the current bullet lists for gobmk and sjeng with a compact table for each benchmark (or a single combined table) that lists the metrics IPC and mispred as rows and columns for off, mix2, mix4, mix8, plus additional columns showing the percent delta vs off for each mix (e.g., "mix2 Δ%" = (mix2 - off)/off * 100). Keep the original numeric values (IPC: 2.544804, 2.580220, 2.583373, 2.580409; mispred: 44249, 42602, 42686, 42887 for gobmk; and IPC: 2.116851, 2.120000, 2.129908, 2.124130; mispred: 47856, 47515, 47429, 47691 for sjeng), format them into columns Off / Mix2 / Mix4 / Mix8 and append percent-change columns for each mix relative to Off so readers can quickly see trend strength and tradeoffs.
22-30: Add full reproducibility metadata for the reported numbers.The setup/results sections still miss key replay inputs (exact command line, checkpoint ID/path, random seed policy, binary/input identifiers). Please add them so others can reproduce the 5M-window IPC/mispred numbers exactly.
Also applies to: 83-94
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@docs/Gem5_Docs/frontend/tage_index_capacity_exploration_20260305.md` around lines 22 - 30, The docs are missing full reproducibility metadata for the reported 5M-window IPC/mispred numbers; update the setup/results sections to include the exact gem5 command line used (including --maxinsts=5000000 --warmup-insts-no-switch=0 and any other flags), the exact checkpoint identifier or path and how it was created, the random seed policy (seed value or how seeds were chosen), and precise binary/input identifiers for each reported slice (e.g., gobmk_nngs_18098, sjeng_84999) plus any config toggles used (system.cpu[0].branchPred.mgsc.enabled and system.cpu[0].branchPred.tage.enableBankConflict); apply the same additions to the other occurrence mentioned (lines 83-94) so anyone can reproduce the IPC/mispred values exactly.src/cpu/pred/btb/btb_tage.cc (1)
87-90: Consider normalizingindexMixTablesagainstnumPredictors.If configured larger than
numPredictors, behavior silently becomes “mix all tables”. Clamping (or warning) would make config intent clearer.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/cpu/pred/btb/btb_tage.cc` around lines 87 - 90, Clamp or validate the indexMixTables value against numPredictors to prevent silent “mix all tables” behavior: in the constructor (where indexMixTables is initialized) or immediately after configuration, check if indexMixTables > numPredictors and either set indexMixTables = numPredictors (clamp) or log a warning and adjust accordingly; reference the members indexMixTables and numPredictors and update any code that reads indexMixTables so it uses the validated/clamped value.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@src/cpu/pred/btb/btb_tage.cc`:
- Around line 217-224: The bank-conflict bookkeeping must use the same PC used
for index calculation (the derived indexPc/baseIndexPc) instead of always using
startPC; update the bank-tracking logic that runs alongside the predictor loop
(when enableBankConflict is true) to compute bank/index IDs from indexPc (or
baseIndexPc) the same way getTageIndex uses it so conflict/defer decisions match
the selected indexed PC; locate where bank bookkeeping references startPC and
replace it to derive bank indices from indexPc (or reuse baseIndexPc) consistent
with buildIndexPC/useBranchPcForIndex and the loop over i, and ensure predMeta
paths (indexFoldedHist) are still honored.
---
Nitpick comments:
In `@docs/Gem5_Docs/frontend/tage_index_capacity_exploration_20260305.md`:
- Around line 107-117: The commit list under the section "7. 当前分支提交链（CI触发）" is
brittle; replace the hardcoded short-hash entries for branch `bigger-tage-align`
with immutable links (full commit SHAs or permalink URLs) pointing to each
commit or a single permalinked range/PR so readers can always resolve history;
update the five listed items (including the one mentioning `tableSizes`) to use
their respective commit URLs or the PR permalink and leave the descriptive text
unchanged.
- Around line 85-94: Replace the current bullet lists for gobmk and sjeng with a
compact table for each benchmark (or a single combined table) that lists the
metrics IPC and mispred as rows and columns for off, mix2, mix4, mix8, plus
additional columns showing the percent delta vs off for each mix (e.g., "mix2
Δ%" = (mix2 - off)/off * 100). Keep the original numeric values (IPC: 2.544804,
2.580220, 2.583373, 2.580409; mispred: 44249, 42602, 42686, 42887 for gobmk; and
IPC: 2.116851, 2.120000, 2.129908, 2.124130; mispred: 47856, 47515, 47429, 47691
for sjeng), format them into columns Off / Mix2 / Mix4 / Mix8 and append
percent-change columns for each mix relative to Off so readers can quickly see
trend strength and tradeoffs.
- Around line 22-30: The docs are missing full reproducibility metadata for the
reported 5M-window IPC/mispred numbers; update the setup/results sections to
include the exact gem5 command line used (including --maxinsts=5000000
--warmup-insts-no-switch=0 and any other flags), the exact checkpoint identifier
or path and how it was created, the random seed policy (seed value or how seeds
were chosen), and precise binary/input identifiers for each reported slice
(e.g., gobmk_nngs_18098, sjeng_84999) plus any config toggles used
(system.cpu[0].branchPred.mgsc.enabled and
system.cpu[0].branchPred.tage.enableBankConflict); apply the same additions to
the other occurrence mentioned (lines 83-94) so anyone can reproduce the
IPC/mispred values exactly.
In `@src/cpu/pred/btb/btb_tage.cc`:
- Around line 87-90: Clamp or validate the indexMixTables value against
numPredictors to prevent silent “mix all tables” behavior: in the constructor
(where indexMixTables is initialized) or immediately after configuration, check
if indexMixTables > numPredictors and either set indexMixTables = numPredictors
(clamp) or log a warning and adjust accordingly; reference the members
indexMixTables and numPredictors and update any code that reads indexMixTables
so it uses the validated/clamped value.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 2636f8a7-bf63-429d-946f-21ef7bd864f3
📒 Files selected for processing (5)
configs/example/kmhv3.pydocs/Gem5_Docs/frontend/tage_index_capacity_exploration_20260305.mdsrc/cpu/pred/BranchPredictor.pysrc/cpu/pred/btb/btb_tage.ccsrc/cpu/pred/btb/btb_tage.hh
🚧 Files skipped from review as they are similar to previous changes (1)
- configs/example/kmhv3.py
🚀 Coremark Smoke Test Results
✅ Difftest smoke test passed! |
|
[Generated by GEM5 Performance Robot] Align BTB PerformanceOverall Score
[Generated by GEM5 Performance Robot] Align BTB PerformanceOverall Score
|
Change-Id: If39f7384f40450cae4c7dbb285bc8ea71ef0486b
Change-Id: I8ccbe4c2e97c9e87274dabc61f41b4d4398b79d0
Change-Id: I6d73773fcaa021a7d2a0e7c45eb5f9d61aeddda5
|
[Generated by GEM5 Performance Robot] Align BTB PerformanceOverall Score
|
Add a path-hash-backed exact-history mode to BTBTAGEUpperBound. This keeps the UB implementation aligned with the current RTL-friendly PHR semantics and exposes a kmhv3 flag for direct comparison against UB-S. Change-Id: I9bb7832a5770833562a48dcf7116d55ef8db8b15
Make kmhv3 default to the UB-P path-hash mode when BTBTAGEUpperBound is enabled, and keep an explicit flag to fall back to outcome-history mode for comparisons. Change-Id: I98cade975baa98cbddfde28bda281cc97b384ec6
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (2)
src/cpu/pred/btb/test/btb_tage.test.cc (1)
1018-1059: Avoid pinning these history tests to table 0.Both tests are validating history/meta behavior, but they currently fail if allocation lands anywhere other than table 0. That couples them to the current allocation policy instead of the behavior under test.
♻️ Suggested test hardening
+ auto hasExactEntryInAnyTable = + [&](Addr pc, const boost::dynamic_bitset<> &hist) { + for (int t = 0; t < tage->numPredictors; ++t) { + if (tage->hasExactEntry(t, pc, hist)) { + return true; + } + } + return false; + }; + EXPECT_TRUE(tage->hasExactEntry(0, entry.pc, historyA)); EXPECT_FALSE(tage->hasExactEntry(0, entry.pc, historyB)); + EXPECT_TRUE(hasExactEntryInAnyTable(entry.pc, historyA)); + EXPECT_FALSE(hasExactEntryInAnyTable(entry.pc, historyB)); @@ - EXPECT_TRUE(tage->hasExactEntry(0, newEntry.pc, historyA)); + EXPECT_TRUE(hasExactEntryInAnyTable(newEntry.pc, historyA));🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@src/cpu/pred/btb/test/btb_tage.test.cc` around lines 1018 - 1059, The tests assume allocation always lands in table 0 by calling hasExactEntry(0, ...); instead make the assertion table-index agnostic by checking for the exact entry across all TAGE tables (e.g., loop i from 0 to tage->numTables()-1 or use an existing finder API) and assert that at least one table contains the entry; update both tests (the one after update/recoverHist and the new conditional-entry test that calls putPCHistory/update) to replace the fixed hasExactEntry(0, ...) checks with a scan over all tables (using tage->numTables() or the equivalent method) or a helper like findExactEntry(...) so the test validates history/meta behavior regardless of which table allocation chooses..codex/skills/ci-perf-analysis/scripts/ci_perf_info.py (1)
36-50: Iterate through jobs to find archive source instead of assuming first job.This code currently assumes the first job returned by GitHub is the archive-producing job. While this is correct for the current workflow structure (which has a single job), it's fragile. If the workflow later adds setup/post jobs, enables matrix fan-out, or reorders jobs, this code will break. Scanning all run jobs until finding one whose logs contain
ARCHIVE_REmakes the archive lookup resilient to workflow changes.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.codex/skills/ci-perf-analysis/scripts/ci_perf_info.py around lines 36 - 50, Change the brittle single-job assumption by returning all job ids and scanning each job's logs until ARCHIVE_RE is found: modify get_job_id to return a list of job ids (e.g., rename to get_job_ids and return [str(job["id"]) for job in data.get("jobs", [])]), then update get_archive_path to accept a run_id (or call get_job_ids) and iterate over those job ids, calling run_cmd(["gh","api",f"repos/{REPO}/actions/jobs/{job_id}/logs"]) for each and returning match.group(1) for the first job whose logs match ARCHIVE_RE; raise the existing SystemExit if no job contains the archive.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In @.codex/skills/ci-perf-analysis/SKILL.md:
- Around line 43-49: Fix the shell export example: replace the incorrect
assignment that expands the variable name (`export $GEM5_DATA_PROC_HOME=xxx`)
with a proper export/assignment for GEM5_DATA_PROC_HOME (e.g., use export
GEM5_DATA_PROC_HOME=xxx or assign then export GEM5_DATA_PROC_HOME) so the
environment variable GEM5_DATA_PROC_HOME is actually set as intended in
SKILL.md.
---
Nitpick comments:
In @.codex/skills/ci-perf-analysis/scripts/ci_perf_info.py:
- Around line 36-50: Change the brittle single-job assumption by returning all
job ids and scanning each job's logs until ARCHIVE_RE is found: modify
get_job_id to return a list of job ids (e.g., rename to get_job_ids and return
[str(job["id"]) for job in data.get("jobs", [])]), then update get_archive_path
to accept a run_id (or call get_job_ids) and iterate over those job ids, calling
run_cmd(["gh","api",f"repos/{REPO}/actions/jobs/{job_id}/logs"]) for each and
returning match.group(1) for the first job whose logs match ARCHIVE_RE; raise
the existing SystemExit if no job contains the archive.
In `@src/cpu/pred/btb/test/btb_tage.test.cc`:
- Around line 1018-1059: The tests assume allocation always lands in table 0 by
calling hasExactEntry(0, ...); instead make the assertion table-index agnostic
by checking for the exact entry across all TAGE tables (e.g., loop i from 0 to
tage->numTables()-1 or use an existing finder API) and assert that at least one
table contains the entry; update both tests (the one after update/recoverHist
and the new conditional-entry test that calls putPCHistory/update) to replace
the fixed hasExactEntry(0, ...) checks with a scan over all tables (using
tage->numTables() or the equivalent method) or a helper like findExactEntry(...)
so the test validates history/meta behavior regardless of which table allocation
chooses.
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 1496048c-95fc-4c2d-a1fe-a0dbd44fe341
📒 Files selected for processing (14)
.codex/skills/ci-perf-analysis/SKILL.md.codex/skills/ci-perf-analysis/scripts/ci_perf_info.pyconfigs/common/xiangshan.pyconfigs/example/kmhv3.pydocs/Gem5_Docs/frontend/tage_index_capacity_exploration_20260305.mddocs/Gem5_Docs/frontend/tage_upper_bound_design_20260310.mdsrc/cpu/pred/BranchPredictor.pysrc/cpu/pred/SConscriptsrc/cpu/pred/btb/btb_tage.ccsrc/cpu/pred/btb/btb_tage.hhsrc/cpu/pred/btb/btb_tage_ub.ccsrc/cpu/pred/btb/btb_tage_ub.hhsrc/cpu/pred/btb/test/SConscriptsrc/cpu/pred/btb/test/btb_tage.test.cc
🚧 Files skipped from review as they are similar to previous changes (2)
- src/cpu/pred/btb/btb_tage.cc
- src/cpu/pred/btb/btb_tage.hh
There was a problem hiding this comment.
🧹 Nitpick comments (1)
configs/example/kmhv3.py (1)
118-128: MGSC disabled; downstream table configs are now no-ops.With
mgsc.enabled = False, the MGSC component'sisEnabled()returns false andputPCHistory()returns early (persrc/cpu/pred/btb/btb_mgsc.cc:513), bypassing all prediction logic. This makes the table configuration on lines 122-127 (enableBwTable,enableLTable, etc.) effectively dead code.If this is intentional (preserving config for future re-enablement), consider adding a brief comment. Otherwise, these lines could be removed or wrapped in a conditional to improve clarity.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@configs/example/kmhv3.py` around lines 118 - 128, The MGSC config sets cpu.branchPred.mgsc.enabled = False which makes MGSC.isEnabled() false and causes putPCHistory() to return early, rendering downstream flags (cpu.branchPred.mgsc.enableBwTable, enableLTable, enableITable, enableGTable, enablePTable, enableBiasTable) no-ops; either remove these dead flag lines or wrap them in a conditional that only sets them when cpu.branchPred.mgsc.enabled is True, or add a concise inline comment explaining they are intentionally preserved for future re-enablement so readers know the redundancy is deliberate (reference symbols: cpu.branchPred.mgsc.enabled, MGSC.isEnabled(), putPCHistory(), and the enable* table flags).
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Nitpick comments:
In `@configs/example/kmhv3.py`:
- Around line 118-128: The MGSC config sets cpu.branchPred.mgsc.enabled = False
which makes MGSC.isEnabled() false and causes putPCHistory() to return early,
rendering downstream flags (cpu.branchPred.mgsc.enableBwTable, enableLTable,
enableITable, enableGTable, enablePTable, enableBiasTable) no-ops; either remove
these dead flag lines or wrap them in a conditional that only sets them when
cpu.branchPred.mgsc.enabled is True, or add a concise inline comment explaining
they are intentionally preserved for future re-enablement so readers know the
redundancy is deliberate (reference symbols: cpu.branchPred.mgsc.enabled,
MGSC.isEnabled(), putPCHistory(), and the enable* table flags).
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro
Run ID: 4974b1b2-2193-460d-8d72-183a6aa5b21f
📒 Files selected for processing (2)
configs/common/xiangshan.pyconfigs/example/kmhv3.py
|
[Generated by GEM5 Performance Robot] Align BTB PerformanceOverall Score
|
jensen-yan
changed the title
🚀 Coremark Smoke Test Results
✅ Difftest smoke test passed! |
|
[Generated by GEM5 Performance Robot] Align BTB PerformanceOverall Score
[Generated by GEM5 Performance Robot] Align BTB PerformanceOverall Score
|
CaoJiaming776
approved these changes
Mar 16, 2026
Yakkhini
approved these changes
Mar 16, 2026
This was referenced Mar 16, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
BTBTAGEUpperBoundas an optionalkmhv3predictor modekmhv3default predictor settings aligned with currentxs-dev--btb-tage-upper-boundswitchWhat changed
BTBTAGEUpperBoundSimObject and implementationkmhv3so it replaces the defaultBTBTAGEonly when the new flag is passedxs-devpredictor changes, including the standaloneMicroTAGEintegrationNotes
kmhv3behavior remains unchanged unless--btb-tage-upper-boundis specifiedValidation
python3 -m py_compile configs/common/xiangshan.py configs/example/kmhv3.py src/cpu/pred/BranchPredictor.pyscons build/RISCV/cpu/pred/btb/test/tage.test.debug --unit-test -j32./build/RISCV/cpu/pred/btb/test/tage.test.debug --gtest_filter=BTBTAGEUpperBoundTest.ExactContextLookup:BTBTAGEUpperBoundPathHashTest.PredictionUsesPathHashHistorySnapshot