Conversation
- Updated specUpdateIHist and recoverIHist methods to remove unused history parameters, passing an empty bitset instead. - Removed references to IMLI history in DecoupledBPUWithBTB and FetchStream structures, as it is no longer utilized. - Added notes in documentation to clarify that IMLI only uses counters, not history bits.
- Added conditional compilation for unit testing, allowing for a separate constructor and initialization logic. - Introduced debug flags and test-specific structures to facilitate testing. - Refactored the initStorage method to improve clarity and organization of storage initialization. - Created a new test file for BTBMGSC to validate its functionality with unit tests. - Updated SConscript to include the new test file in the build process. Change-Id: Ib1ab026117ef87ebbf7097a8d32710843d32ed5b
📝 WalkthroughWalkthroughRemoves IMLI (I-) history tracking and switches IMLI to counter-only, updates predictor interfaces and call sites (specUpdateIHist/recoverIHist), centralizes MGSC table allocation via initStorage(), and adds UNIT_TEST-scoped test scaffolding and a test-only constructor/namespace. Changes
Sequence Diagram(s)sequenceDiagram
participant Fetch as Fetch (frontend)
participant DBPU as DecoupledBPU
participant Stream as FetchStream
participant Predictor as BTBMGSC / TimedBase
Note over Fetch,DBPU: Prior flow (with I-history)
Fetch->>DBPU: request fetch
DBPU->>Stream: create FetchStream (includes ihistory)
DBPU->>Predictor: specUpdateIHist(ihistory, finalPred)
Predictor-->>DBPU: prediction (may update history)
DBPU->>DBPU: on squash/recover call recoverIHist(ihistory,...)
Note over Fetch,DBPU: New flow (IMLI counter-only)
Fetch->>DBPU: request fetch
DBPU->>Stream: create FetchStream (no ihistory)
DBPU->>Predictor: specUpdateIHist(finalPred)
Predictor-->>DBPU: prediction (IMLI counters updated without history)
DBPU->>DBPU: on squash/recover call recoverIHist(stream,...)
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Possibly related PRs
Suggested reviewers
Poem
🚥 Pre-merge checks | ✅ 1 | ❌ 2❌ Failed checks (1 warning, 1 inconclusive)
✅ Passed checks (1 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing touches
📜 Recent review detailsConfiguration used: defaults Review profile: CHILL Plan: Pro 📒 Files selected for processing (2)
🧰 Additional context used🧠 Learnings (1)📚 Learning: 2026-01-12T02:49:12.886ZApplied to files:
🧬 Code graph analysis (2)src/cpu/pred/btb/test/btb_mgsc.test.cc (1)
src/cpu/pred/btb/btb_mgsc.cc (4)
🪛 Cppcheck (2.19.0)src/cpu/pred/btb/test/btb_mgsc.test.cc[error] 337-337: syntax error (syntaxError) src/cpu/pred/btb/btb_mgsc.cc[error] 119-119: Reference to temporary returned. (returnTempReference) [error] 132-132: Reference to temporary returned. (returnTempReference) [error] 125-125: Shifting 64-bit value by 64 bits is undefined behaviour (shiftTooManyBits) [error] 138-138: Shifting 64-bit value by 64 bits is undefined behaviour (shiftTooManyBits) ⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)
🔇 Additional comments (13)
✏️ Tip: You can disable this entire section by setting 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 |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 0
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
src/cpu/pred/btb/btb_mgsc.cc (1)
1266-1273: Fix indentation error and stray closing brace.Lines 1271-1272 have excessive indentation (12 spaces instead of 4), and line 1273 has an orphan
}that shouldn't exist. This appears to be a merge or copy-paste error. ThedoUpdateHistcall should be at the same level as the preceding for loop.🐛 Proposed fix
for (unsigned int k = 0; k < numEntriesFirstLocalHistories; ++k) { for (int i = 0; i < lTableNum; i++) { indexLFoldedHist[k][i].recover(predMeta->indexLFoldedHist[k][i]); } } - doUpdateHist(history[getPcIndex(entry.startPC, log2(numEntriesFirstLocalHistories))], shamt, cond_taken, - indexLFoldedHist[getPcIndex(entry.startPC, log2(numEntriesFirstLocalHistories))]); - } + doUpdateHist(history[getPcIndex(entry.startPC, log2(numEntriesFirstLocalHistories))], shamt, cond_taken, + indexLFoldedHist[getPcIndex(entry.startPC, log2(numEntriesFirstLocalHistories))]); +}
🧹 Nitpick comments (1)
src/cpu/pred/btb/test/btb_mgsc.test.cc (1)
18-35: Good basic smoke test for BTBMGSC construction.The test verifies that
BTBMGSCcan be constructed and handles empty input correctly. The assertions confirm metadata creation and empty prediction state.Consider expanding the test suite with additional scenarios in follow-up PRs:
- Non-empty history inputs
- Multiple prediction cycles
- History recovery paths
📜 Review details
Configuration used: defaults
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (9)
src/cpu/pred/btb/btb_mgsc.ccsrc/cpu/pred/btb/btb_mgsc.hhsrc/cpu/pred/btb/decoupled_bpred.ccsrc/cpu/pred/btb/decoupled_bpred.hhsrc/cpu/pred/btb/stream_struct.hhsrc/cpu/pred/btb/test/SConscriptsrc/cpu/pred/btb/test/btb_mgsc.test.ccsrc/cpu/pred/btb/timed_base_pred.ccsrc/cpu/pred/btb/timed_base_pred.hh
💤 Files with no reviewable changes (2)
- src/cpu/pred/btb/stream_struct.hh
- src/cpu/pred/btb/decoupled_bpred.hh
🧰 Additional context used
🧠 Learnings (1)
📚 Learning: 2026-01-12T02:49:12.886Z
Learnt from: jensen-yan
Repo: OpenXiangShan/GEM5 PR: 686
File: src/cpu/pred/btb/btb_mgsc.cc:226-230
Timestamp: 2026-01-12T02:49:12.886Z
Learning: In src/cpu/pred/btb/btb_mgsc.cc, the calculateScaledPercsum function returns the unscaled percsum (weight scaling disabled) to align with RTL implementation. The RTL currently does not have a weight table, so setting weight to 1 (equivalent to disabling the weight table) maintains hardware/software consistency with minimal performance impact.
Applied to files:
src/cpu/pred/btb/test/btb_mgsc.test.ccsrc/cpu/pred/btb/btb_mgsc.hhsrc/cpu/pred/btb/btb_mgsc.cc
🧬 Code graph analysis (5)
src/cpu/pred/btb/test/btb_mgsc.test.cc (2)
src/cpu/pred/btb/btb_mgsc.hh (11)
history(162-162)history(163-163)history(164-164)history(166-166)history(170-171)history(172-173)history(174-175)history(178-179)history(500-500)pred(165-165)pred(286-287)src/cpu/pred/btb/btb_ubtb.hh (1)
meta(144-147)
src/cpu/pred/btb/timed_base_pred.hh (1)
src/cpu/pred/btb/btb_mgsc.hh (6)
pred(165-165)pred(286-287)history(162-162)entry(176-177)entry(182-182)entry(284-285)
src/cpu/pred/btb/timed_base_pred.cc (2)
src/cpu/pred/btb/timed_base_pred.hh (2)
TimedBaseBTBPredictor(48-48)TimedBaseBTBPredictor(53-53)src/cpu/pred/BranchPredictor.py (1)
TimedBaseBTBPredictor(962-972)
src/cpu/pred/btb/btb_mgsc.hh (2)
src/cpu/pred/btb/timed_base_pred.hh (11)
pred(69-69)pred(69-69)entry(74-74)entry(74-74)entry(76-76)entry(76-76)entry(80-80)history(66-66)history(66-66)history(67-67)history(67-67)src/cpu/pred/btb/btb_ittage.hh (1)
stream(115-115)
src/cpu/pred/btb/btb_mgsc.cc (3)
src/cpu/pred/btb/folded_hist.hh (2)
ImliFoldedHist(156-159)PathFoldedHist(167-167)src/cpu/pred/btb/timed_base_pred.hh (4)
pred(69-69)pred(69-69)entry(74-74)entry(74-74)src/cpu/pred/btb/stream_struct.hh (3)
shamt(606-631)shamt(633-656)entry(512-534)
🪛 Cppcheck (2.19.0)
src/cpu/pred/btb/btb_mgsc.cc
[error] 119-119: Reference to temporary returned.
(returnTempReference)
[error] 132-132: Reference to temporary returned.
(returnTempReference)
[error] 125-125: Shifting 64-bit value by 64 bits is undefined behaviour
(shiftTooManyBits)
[error] 138-138: Shifting 64-bit value by 64 bits is undefined behaviour
(shiftTooManyBits)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)
- GitHub Check: Quick Build, Unit Tests & Smoke Test
🔇 Additional comments (17)
src/cpu/pred/btb/timed_base_pred.hh (1)
69-69: LGTM! Signature changes align with IMLI refactoring.The removal of history parameters from
specUpdateIHistandrecoverIHistis consistent with the PR objective to remove IMLI history tracking. The simplified signatures match the overrides inbtb_mgsc.hh(lines 164 and 175-176 in relevant snippets).Also applies to: 74-74
src/cpu/pred/btb/decoupled_bpred.cc (2)
1218-1218: LGTM! Call site correctly updated for new signature.The
specUpdateIHistcall now passes onlyfinalPred, matching the updated signature inTimedBaseBTBPredictor. The surrounding code maintains updates for other history types (specUpdateHist,specUpdatePHist,specUpdateBwHist,specUpdateLHist), showing the I-history removal is isolated.
1314-1314: LGTM! Recovery call site properly aligned with signature change.The
recoverIHistcall now passes(stream, real_bw_shamt, real_bw_taken), consistent with the updated base class signature. The parallel recovery calls for other history types remain unchanged, indicating a clean refactoring scope.src/cpu/pred/btb/timed_base_pred.cc (1)
15-21: LGTM! Test constructor provides sensible defaults.The
UNIT_TESTconstructor properly initializes all private members with reasonable test values:
numDelay(0)removes artificial delays for faster test executionenabled(true)ensures the predictor is active during tests- Other values provide working defaults for unit test scenarios
src/cpu/pred/btb/test/SConscript (2)
31-36: LGTM! Test target configuration is correct.The
mgsc.testtarget includes the appropriate dependencies:
btb_mgsc.ccfor the MGSC implementationfolded_hist.ccfor folded history utilities used by MGSCtimed_base_pred.ccfor the base predictor classThis matches the pattern used by similar tests like
tage.test.
76-83: Test registration complete.The
mgsc.testis properly added to theUNITTESTSaggregate.Note:
abtb.test(defined on lines 16-21) appears to be missing from theUNITTESTSlist. If this is unintentional, consider adding it in a follow-up.src/cpu/pred/btb/btb_mgsc.cc (6)
42-113: LGTM:initStorage()centralizes table allocation logic.The refactoring consolidates initialization code into a single method, reducing duplication between UNIT_TEST and production constructors. The
pow2lambda properly guards against shifting by 63+ bits withassert(width < 63).
115-162: LGTM: UNIT_TEST constructor with appropriate small configurations.The test constructor uses small, deterministic values for fast unit tests. The comment at lines 127-128 correctly explains the
ImliFoldedHistconstraint requiringfoldedLen >= histLen, andiHistLen({1})satisfies this withfoldedLen = iTableIdxWidth - log2(numCtrsPerLine) = 4 - 3 = 1.
1107-1115: LGTM: IMLI history update simplified to counter-only behavior.The change aligns with the documented IMLI semantics (counters only, no history bits). Using an empty
dummybitset is appropriate sinceImliFoldedHist::update()doesn't use the history parameter for its counter logic.
1231-1243: LGTM: IMLI recovery aligned with speculative update pattern.The
recoverIHistsignature change matchesspecUpdateIHist, and the dummy bitset approach is consistent. The comment at line 1240 clearly documents the intent.
672-678: LGTM: Lambda signature usesautofor parameter deduction.Using
auto &correct, auto &wrongallows the lambda to work with bothstatistics::Scalar(production) anduint64_t(UNIT_TEST) types, maintaining flexibility across build configurations.
1416-1418: LGTM: Test namespace properly closed.The
testnamespace wrapper is correctly closed only whenUNIT_TESTis defined, maintaining consistency with the opening at line 38-40.src/cpu/pred/btb/btb_mgsc.hh (5)
34-37: LGTM: Conditional namespace wrapper for test isolation.The
testnamespace is properly wrapped with#ifdef UNIT_TESTguards, and correctly opened (lines 35-37) and closed (lines 543-545). This isolates the test BTBMGSC from the production version, preventing symbol conflicts during testing.Also applies to: 541-545
165-177: LGTM: Updated method signatures match base class.Both
specUpdateIHist(line 165) andrecoverIHist(lines 176-177) now omit the history parameter, aligning with the base class signatures intimed_base_pred.hh(lines 68 and 73). Theoverridespecifier ensures compile-time verification of the virtual function override.
418-422: LGTM:Scalartypedef enables dual-mode statistics.The typedef pattern (
uint64_tfor UNIT_TEST,statistics::Scalarotherwise) allowsMgscStatsfields to be simple counters in tests while using gem5's statistics framework in production. The brace initialization{}for each field ensures zero-initialization in UNIT_TEST mode.Also applies to: 425-490
184-186: LGTM:commitBranchexcluded in UNIT_TEST builds.This method depends on
DynInstPtrwhich requires full gem5 infrastructure. Excluding it under UNIT_TEST avoids unnecessary dependencies for focused unit testing.
504-504: LGTM:initStorage()declared as private helper.The new private method declaration supports the refactored initialization pattern in the implementation file.
✏️ Tip: You can disable this entire section by setting review_details to false in your review settings.
🚀 Coremark Smoke Test Results
✅ Difftest smoke test passed! |
- Increased index width for various tables in BTBMGSC to improve capacity. - Added a TestAccess struct for unit testing, providing access to private members of BTBMGSC. - Implemented new utility functions for handling BTB entries and indices in unit tests. - Enhanced unit tests to validate the behavior of BTBMGSC under different conditions. Change-Id: I4b76a70b1270d7fb96db816779b66dfaf192c370
- Introduced a new test case to validate GTable's learning of alternating patterns. Change-Id: I2f55b712b92d119a6b4682d7a191ed54fb6c7607
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Fix all issues with AI agents
In `@src/cpu/pred/btb/btb_mgsc.cc`:
- Around line 1276-1278: In recoverLHist, the doUpdateHist(...) call and the
following closing brace are mis-indented and there is an extra closing brace
causing a compile error; locate the
doUpdateHist(history[getPcIndex(entry.startPC,
log2(numEntriesFirstLocalHistories))], shamt, cond_taken,
indexLFoldedHist[getPcIndex(entry.startPC,
log2(numEntriesFirstLocalHistories))]) line and the stray closing brace after
it, remove the superfluous brace and fix indentation so the doUpdateHist call
aligns with the surrounding block scope in recoverLHist.
🧹 Nitpick comments (1)
src/cpu/pred/btb/test/btb_mgsc.test.cc (1)
34-54: Consider adding bounds checking in lineLaneForHistIndex/lineLaneForBiasIndex.These helpers assume table index 0 exists. While this is valid for the current test configuration (all tables have at least 1 entry), adding an assertion would make failures easier to debug.
♻️ Optional: Add defensive assertion
std::pair<unsigned, unsigned> lineLaneForHistIndex(BTBMGSC &mgsc, Addr start_pc, Addr branch_pc, unsigned table_idx_width) { const unsigned num_bits = BTBMGSC::TestAccess::numCtrsPerLineBits(mgsc); + assert(table_idx_width > num_bits && "table_idx_width must be larger than numCtrsPerLineBits"); const unsigned idx_bits = table_idx_width - num_bits; Addr line_idx = BTBMGSC::TestAccess::getHistIndex(mgsc, start_pc, idx_bits, /*foldedHist=*/0); auto [idx1, idx2] = BTBMGSC::TestAccess::posHash(mgsc, branch_pc, line_idx); return {idx1, idx2}; }
📜 Review details
Configuration used: defaults
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (3)
src/cpu/pred/btb/btb_mgsc.ccsrc/cpu/pred/btb/btb_mgsc.hhsrc/cpu/pred/btb/test/btb_mgsc.test.cc
🧰 Additional context used
🧠 Learnings (1)
📚 Learning: 2026-01-12T02:49:12.886Z
Learnt from: jensen-yan
Repo: OpenXiangShan/GEM5 PR: 686
File: src/cpu/pred/btb/btb_mgsc.cc:226-230
Timestamp: 2026-01-12T02:49:12.886Z
Learning: In src/cpu/pred/btb/btb_mgsc.cc, the calculateScaledPercsum function returns the unscaled percsum (weight scaling disabled) to align with RTL implementation. The RTL currently does not have a weight table, so setting weight to 1 (equivalent to disabling the weight table) maintains hardware/software consistency with minimal performance impact.
Applied to files:
src/cpu/pred/btb/btb_mgsc.ccsrc/cpu/pred/btb/test/btb_mgsc.test.ccsrc/cpu/pred/btb/btb_mgsc.hh
🧬 Code graph analysis (3)
src/cpu/pred/btb/btb_mgsc.cc (3)
src/cpu/pred/btb/folded_hist.hh (2)
ImliFoldedHist(156-159)PathFoldedHist(167-167)src/cpu/pred/btb/timed_base_pred.hh (4)
pred(69-69)pred(69-69)entry(74-74)entry(74-74)src/cpu/pred/btb/stream_struct.hh (3)
shamt(606-631)shamt(633-656)entry(512-534)
src/cpu/pred/btb/test/btb_mgsc.test.cc (4)
src/cpu/pred/btb/btb_mgsc.hh (36)
pc(195-195)pc(240-240)pc(245-245)pc(252-252)pc(255-255)pc(258-258)pc(258-258)pc(272-275)pc(272-272)mgsc(506-506)mgsc(506-506)mgsc(507-507)mgsc(507-507)mgsc(509-509)mgsc(509-509)mgsc(510-510)mgsc(510-510)mgsc(511-511)mgsc(511-511)mgsc(512-512)mgsc(512-512)mgsc(513-513)mgsc(513-513)mgsc(514-514)mgsc(514-514)history(162-162)history(163-163)history(164-164)history(166-166)history(170-171)history(172-173)history(174-175)history(178-179)history(500-500)pred(165-165)pred(286-287)src/cpu/pred/btb/btb_mgsc.cc (4)
getHistIndex(907-918)getHistIndex(908-908)getBiasIndex(920-930)getBiasIndex(921-921)src/cpu/pred/btb/folded_hist.hh (1)
dynamic_bitset(105-105)src/cpu/pred/btb/btb_ubtb.hh (1)
meta(144-147)
src/cpu/pred/btb/btb_mgsc.hh (1)
src/cpu/pred/btb/timed_base_pred.hh (7)
pred(69-69)pred(69-69)entry(74-74)entry(74-74)entry(76-76)entry(76-76)entry(80-80)
🪛 Cppcheck (2.19.0)
src/cpu/pred/btb/btb_mgsc.cc
[error] 119-119: Reference to temporary returned.
(returnTempReference)
[error] 132-132: Reference to temporary returned.
(returnTempReference)
[error] 125-125: Shifting 64-bit value by 64 bits is undefined behaviour
(shiftTooManyBits)
[error] 138-138: Shifting 64-bit value by 64 bits is undefined behaviour
(shiftTooManyBits)
src/cpu/pred/btb/test/btb_mgsc.test.cc
[error] 263-263: syntax error
(syntaxError)
🔇 Additional comments (18)
src/cpu/pred/btb/btb_mgsc.hh (6)
34-37: LGTM on UNIT_TEST namespace wrapper.The conditional namespace wrapper pattern isolates the test build properly, preventing symbol conflicts with production code.
42-47: LGTM on conditional constructor declaration.The UNIT_TEST guard appropriately provides a default constructor for testing while maintaining the production constructor that requires parameters.
165-177: Interface changes align with base class declarations.The
specUpdateIHistandrecoverIHistsignatures now match the virtual declarations intimed_base_pred.hh. The removal of the history parameter is documented in the implementation as IMLI using counter-only behavior.
418-422: LGTM on Scalar typedef for statistics.Using
uint64_tfor UNIT_TEST mode avoids dependency on the gem5 statistics framework while maintaining type compatibility for counter operations.
504-555: TestAccess helper is well-designed for unit testing.The friend-like accessor pattern using static methods is a clean way to expose private members for testing without modifying the class interface. The accessors cover all necessary tables, flags, and index calculation methods.
596-600: LGTM on namespace closing.The conditional namespace wrapper is properly closed to match the opening guard.
src/cpu/pred/btb/btb_mgsc.cc (5)
42-113: Good refactor: initStorage centralizes table allocation.Extracting storage initialization into
initStorage()eliminates duplication between the UNIT_TEST and production constructors. The local lambdaspow2andallocPredTablekeep the allocation logic readable.
1111-1120: LGTM on IMLI counter-only history update.The change to pass a dummy empty bitset to
doUpdateHistfor IMLI is well-documented. The comment explains thatImliFoldedHist::updatedoesn't use the history bits, only the counter behavior.
1236-1248: LGTM on IMLI counter-only recovery.Consistent with
specUpdateIHist, the recovery path now uses a dummy bitset for IMLI, maintaining the counter-only behavior.
677-683: Lambda parameter type is appropriate.The
auto &correct, auto &wrongparameters inrecordPercsumwork correctly since they accept bothuint64_t(UNIT_TEST) andstatistics::Scalar(production) types through template deduction.
115-165: Assertions at line 82 correctly protect against shift UB; static analysis warnings are false positives.The code is safe:
- Line 82 asserts
iHistLen[i] < 63, preventing shift UB before any shift operations inpow2()- Brace-initialized vectors (
bwHistLen({4}),lHistLen({4}), etc.) are not references to temporaries—static analyzer false positives- UNIT_TEST config with
iHistLen({1})satisfies the constraint:foldedLen = 5 - log2(8) = 2 >= histLen = 1calculateScaledPercsum()correctly returns unscaled percsum (weight scaling disabled) to align with RTL, as intendedsrc/cpu/pred/btb/test/btb_mgsc.test.cc (7)
21-32: LGTM on makeCondBTBEntry helper.The helper properly constructs a minimal conditional BTB entry for testing with reasonable defaults.
98-122: LGTM on history shift helpers.
histShiftInandpHistShiftIncorrectly implement the history update semantics matching the production code's behavior for global and path histories.
124-259: MgscHarness provides comprehensive test scaffolding.The harness correctly implements:
- Prediction via
putPCHistory- Speculative history updates
- Misprediction recovery with history rollback
- Training update via
update()The
step()method mirrors the actual prediction/recovery/update flow inDecoupledBPUWithBTB.
263-280: Good basic construction test.This test validates that:
- Default construction works under UNIT_TEST
putPCHistorycreates valid metadata- Empty BTB entries result in empty
condTakensThe Cppcheck "syntax error" is a false positive -
TESTis a valid GoogleTest macro.
282-318: GateHighConfUsesSCWhenStrong test is well-designed.The test verifies the high-confidence gating threshold: when
abs(total_sum) > total_thres/2, SC should be used. The counter setup producestotal_sum = 18which exceeds35/2 = 17, triggering SC usage.
398-460: UpdateOnlyOnWrongOrLowMargin test validates training conditions.This test correctly verifies that:
- Correct predictions with high confidence don't trigger training
- Wrong predictions do trigger training (counter decrement)
This matches the update condition
if (sc_pred_taken != actual_taken || abs(total_sum) < total_thres)in the implementation.
462-510: GTableLearnsAlternatingPattern is a valuable learning test.This test validates that:
- The G-table can learn distinct contexts from GHR
- After warmup, accuracy exceeds 80% on an alternating pattern
- Multiple GHR contexts are actually observed (
seen_g_indices.size() >= 2)The 80% accuracy threshold is reasonable for a pattern that a simple predictor can only achieve ~50% on.
✏️ Tip: You can disable this entire section by setting review_details to false in your review settings.
🚀 Coremark Smoke Test Results
✅ Difftest smoke test passed! |
- Added new test cases to validate the learning of alternating patterns for different tables (BwTable, ITable, BiasTable). Change-Id: I2273e645f276fac9089581631001be0f4b37b5ca
🚀 Coremark Smoke Test Results
✅ Difftest smoke test passed! |
- Added new test cases to validate the learning of local and path tables under different conditions.
🚀 Coremark Smoke Test Results
✅ Difftest smoke test passed! |
2 participants
add mgsc unit test
Summary by CodeRabbit
Refactor
Tests
✏️ Tip: You can customize this high-level summary in your review settings.