Skip to content

Replace Command with CommandSchedule#20

Merged
jonathanvdc merged 33 commits intomainfrom
parallel-rule-iteration-fusion
Nov 9, 2025
Merged

Replace Command with CommandSchedule#20
jonathanvdc merged 33 commits intomainfrom
parallel-rule-iteration-fusion

Conversation

@jonathanvdc
Copy link
Owner

@jonathanvdc jonathanvdc commented Nov 8, 2025 •
edited
Loading

This pull request replaces the Command class hierarchy with CommandSchedule in order to perform on-the-fly command batch formation, parallelizing a sequential step.

@github-actions
Copy link

github-actions bot commented Nov 8, 2025 •
edited
Loading

JMH comparison (baseline vs PR)

Benchmark Params Baseline PR Δ (PR/Base) Unit
IncrementalBenchmarks.incrementalPolynomial depth=6, mutableEGraph=false, size=500, threadCount=1 210.5 194.29 0.923× (-7.7%) ms/op
IncrementalBenchmarks.incrementalPolynomial depth=6, mutableEGraph=false, size=500, threadCount=2 284.833 310.638 1.091× (+9.1%) ms/op
IncrementalBenchmarks.incrementalPolynomial depth=6, mutableEGraph=true, size=500, threadCount=1 5.07017 7.91167 1.560× (+56.0%) ms/op
IncrementalBenchmarks.incrementalPolynomial depth=6, mutableEGraph=true, size=500, threadCount=2 26.6326 76.5164 2.873× (+187.3%) ms/op
IncrementalBenchmarks.oneByOnePolynomial depth=6, mutableEGraph=false, size=500, threadCount=1 369.453 335.723 0.909× (-9.1%) ms/op
IncrementalBenchmarks.oneByOnePolynomial depth=6, mutableEGraph=false, size=500, threadCount=2 501.172 512.031 1.022× (+2.2%) ms/op
IncrementalBenchmarks.oneByOnePolynomial depth=6, mutableEGraph=true, size=500, threadCount=1 167.732 144.979 0.864× (-13.6%) ms/op
IncrementalBenchmarks.oneByOnePolynomial depth=6, mutableEGraph=true, size=500, threadCount=2 277.907 301.479 1.085× (+8.5%) ms/op
LiarBenchmarks.findGemmInMm threadCount=1 222.272 231.529 1.042× (+4.2%) ms/op
LiarBenchmarks.findGemmInMm threadCount=2 188.495 179.49 0.952× (-4.8%) ms/op
LiarBenchmarks.findGemvInMv threadCount=1 63.5259 59.8171 0.942× (-5.8%) ms/op
LiarBenchmarks.findGemvInMv threadCount=2 57.2284 55.1964 0.964× (-3.6%) ms/op
MatmulBenchmarks.nmm mutableEGraph=false, size=20, threadCount=1 5.75782 4.34889 0.755× (-24.5%) ms/op
MatmulBenchmarks.nmm mutableEGraph=false, size=20, threadCount=2 5.56953 4.35167 0.781× (-21.9%) ms/op
MatmulBenchmarks.nmm mutableEGraph=false, size=40, threadCount=1 38.3029 30.1172 0.786× (-21.4%) ms/op
MatmulBenchmarks.nmm mutableEGraph=false, size=40, threadCount=2 29.2298 23.7103 0.811× (-18.9%) ms/op
MatmulBenchmarks.nmm mutableEGraph=false, size=80, threadCount=1 290.41 272.661 0.939× (-6.1%) ms/op
MatmulBenchmarks.nmm mutableEGraph=false, size=80, threadCount=2 208.461 166.347 0.798× (-20.2%) ms/op
MatmulBenchmarks.nmm mutableEGraph=true, size=20, threadCount=1 3.49014 2.71994 0.779× (-22.1%) ms/op
MatmulBenchmarks.nmm mutableEGraph=true, size=20, threadCount=2 3.3044 2.43997 0.738× (-26.2%) ms/op
MatmulBenchmarks.nmm mutableEGraph=true, size=40, threadCount=1 24.8563 20.4156 0.821× (-17.9%) ms/op
MatmulBenchmarks.nmm mutableEGraph=true, size=40, threadCount=2 21.4311 15.1704 0.708× (-29.2%) ms/op
MatmulBenchmarks.nmm mutableEGraph=true, size=80, threadCount=1 233.612 174.646 0.748× (-25.2%) ms/op
MatmulBenchmarks.nmm mutableEGraph=true, size=80, threadCount=2 161.787 112.02 0.692× (-30.8%) ms/op
PolyBenchmarks.polynomial mutableEGraph=false, size=5, threadCount=1 97.3256 88.7649 0.912× (-8.8%) ms/op
PolyBenchmarks.polynomial mutableEGraph=false, size=5, threadCount=2 88.1037 80.3334 0.912× (-8.8%) ms/op
PolyBenchmarks.polynomial mutableEGraph=false, size=6, threadCount=1 445.165 412.7 0.927× (-7.3%) ms/op
PolyBenchmarks.polynomial mutableEGraph=false, size=6, threadCount=2 383.238 362.433 0.946× (-5.4%) ms/op
PolyBenchmarks.polynomial mutableEGraph=true, size=5, threadCount=1 41.2605 34.4317 0.834× (-16.6%) ms/op
PolyBenchmarks.polynomial mutableEGraph=true, size=5, threadCount=2 34.145 28.8319 0.844× (-15.6%) ms/op
PolyBenchmarks.polynomial mutableEGraph=true, size=6, threadCount=1 160.312 143.57 0.896× (-10.4%) ms/op
PolyBenchmarks.polynomial mutableEGraph=true, size=6, threadCount=2 134.745 115.145 0.855× (-14.5%) ms/op
VectorBenchmarks.blinnPhong mutableEGraph=false, threadCount=1 36129.6 34652.7 0.959× (-4.1%) ms/op
VectorBenchmarks.blinnPhong mutableEGraph=false, threadCount=2 34545 33643.4 0.974× (-2.6%) ms/op
VectorBenchmarks.blinnPhong mutableEGraph=true, threadCount=1 5339.89 4640.81 0.869× (-13.1%) ms/op
VectorBenchmarks.blinnPhong mutableEGraph=true, threadCount=2 4296.27 3364.05 0.783× (-21.7%) ms/op
VectorBenchmarks.gramSchmidt mutableEGraph=false, threadCount=1 542.232 464.929 0.857× (-14.3%) ms/op
VectorBenchmarks.gramSchmidt mutableEGraph=false, threadCount=2 445.549 391.76 0.879× (-12.1%) ms/op
VectorBenchmarks.gramSchmidt mutableEGraph=true, threadCount=1 208.404 168.738 0.810× (-19.0%) ms/op
VectorBenchmarks.gramSchmidt mutableEGraph=true, threadCount=2 167.778 138.516 0.826× (-17.4%) ms/op
VectorBenchmarks.reflection mutableEGraph=false, threadCount=1 321.673 306.639 0.953× (-4.7%) ms/op
VectorBenchmarks.reflection mutableEGraph=false, threadCount=2 266.93 243.992 0.914× (-8.6%) ms/op
VectorBenchmarks.reflection mutableEGraph=true, threadCount=1 136.91 123.608 0.903× (-9.7%) ms/op
VectorBenchmarks.reflection mutableEGraph=true, threadCount=2 104.531 84.4015 0.807× (-19.3%) ms/op
VectorBenchmarks.vectorNormalization mutableEGraph=false, threadCount=1 4.24104 3.83662 0.905× (-9.5%) ms/op
VectorBenchmarks.vectorNormalization mutableEGraph=false, threadCount=2 4.7103 4.34356 0.922× (-7.8%) ms/op
VectorBenchmarks.vectorNormalization mutableEGraph=true, threadCount=1 1.95671 1.64295 0.840× (-16.0%) ms/op
VectorBenchmarks.vectorNormalization mutableEGraph=true, threadCount=2 2.53502 2.30435 0.909× (-9.1%) ms/op
Geomean threadCount=1, mutableEGraph=false — — 0.905× (-9.5%)
Geomean threadCount=1, mutableEGraph=true — — 0.884× (-11.6%)
Geomean threadCount=2, mutableEGraph=false — — 0.917× (-8.3%)
Geomean threadCount=2, mutableEGraph=true — — 0.917× (-8.3%)

Note: < 1.0× means faster on PR; > 1.0× means slower.

@jonathanvdc
Add CommandQueue.highestDependencyForNodes method to efficiently comp…
1131915 
…ute batch index from node dependencies
@jonathanvdc
Temporarily disable search-loop interchange optimization in SearchAnd…
f286344 
…Apply
@jonathanvdc
Optimize CommandQueue.optimizeAdds
c1e0661
@jonathanvdc
Eliminate legacy Command simplification logic
9e1bc7f
@jonathanvdc
Restore parallelization across rules
286df58
@jonathanvdc
Replace Command class hierarchy with CommandSchedule
60b84b3
@jonathanvdc
Fix additions method to handle empty batchZero case
14e1095
@jonathanvdc
Add unit tests for CommandScheduleBuilder functionality
e7043f2
@jonathanvdc
Add tests for addSimplifiedReal method in CommandScheduleBuilder
cd28047
@jonathanvdc
Add test for addSimplifiedReal method to ensure batch layering stabil…
1fec937 
…ity in complex trees
@jonathanvdc
Refactor addSimplifiedReal method documentation for clarity and consi…
ff50e1a 
…stency
@jonathanvdc
Rename test method for addSimplifiedReal to improve clarity
91aaa34
@jonathanvdc
Define custom IntRef for cross-version compatibility
fbe7d1d
@jonathanvdc
Replace IntRef instantiation with new keyword for consistency
2daf835
@jonathanvdc
Use new keyword for IdentityHashMap instantiation for consistency
448ff74
@jonathanvdc
Refactor import statements in EGraphWithMetadata for consistency
66bf25a
@jonathanvdc
Use Function interface for computeIfAbsent in ConcurrentCommandSchedu…
90eaf58 
…leBuilder
@jonathanvdc
Refactor import statements in EGraphWithRoot
bfcadf6
@jonathanvdc
Refactor import statements in EGraphWithRecordedApplications and EGra…
2501ab0 
…phWithRoot
@jonathanvdc
Refactor import statements in ApplierOps
a2a94d8
@jonathanvdc
Make IntRef class final
500718e
@jonathanvdc
Refactor processBlock to process stripes instead of contiguous blocks
a015f0e
@jonathanvdc
Enable search loop interchange
ba8a360
@jonathanvdc
Enhance users method to handle nodes without slots
2d6239d
@jonathanvdc jonathanvdc changed the title Fuse parallel rule iterations Replace Command with CommandSchedule Nov 9, 2025
@jonathanvdc jonathanvdc merged commit 2b9955e into main Nov 9, 2025
2 checks passed
@jonathanvdc jonathanvdc deleted the parallel-rule-iteration-fusion branch November 9, 2025 04:38
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@jonathanvdc