Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

4043 filter smt lemmas use only opaque use equations symmetrically #4054

Open
wants to merge 7 commits into
base: master
Choose a base branch
from

Conversation

jberthold
Copy link
Member

Related: #4034

Builds upon prior PR in #4048

Booster SMT bindings change as follows:

  • SMT lemmas are held in a lookup table separate from the sort and function declarations. The lookup is by SymbolName, and returns lemmas that (transitively) may relate to the given symbol
  • this PR only This lookup table only considers symbols with smt-lib (called opaque symbols). Symbols with smt-hook are left out because no lemmas should be necessary for those symbols (the solver should know all valuable lemmas about its built-in operations)
  • When checking predicates or computing a model, only those lemmas are added which relate to an opaque symbol used in one of the predicates

This is hopefully reducing load on the SMT solver to enable solving problems that came up in engagement proofs and led to timeouts.

@jberthold
Copy link
Member Author

Same picture as before for kevm: sum-to-n slowdown

Test 4043-filter-smt-lemmas-use-only-opaque-use-equations-symmetrically time master-4e95286ea time (4043-filter-smt-lemmas-use-only-opaque-use-equations-symmetrically/master-4e95286ea) time
benchmarks/ecrecoverloop02-sig1-invalid-spec.k 65.03 67.57 0.9624093532632826
benchmarks/bytes00-spec.k 39.02 40.39 0.9660807130477842
examples/sum-to-n-spec.k 96.61 71.21 1.356691475916304
TOTAL 200.66000000000003 179.17 1.1199419545682874

@jberthold
Copy link
Member Author

jberthold commented Oct 3, 2024

kontrol tests, run sequentially (PYTEST_PARALLEL=1) but still too noisy. (and some artefacts of the time comparison script)

Test 4043-filter-smt-lemmas-use-only-opaque-use-equations-symmetrically time master-4e95286ea time (4043-filter-smt-lemmas-use-only-opaque-use-equations-symmetrically/master-4e95286ea) time
kontrol/src/tests/integration/test_foundry_prove.py::test_foundry_remove_node 5.47 10.65 0.5136150234741783
kontrol/src/tests/integration/test_foundry_prove.py::test_constructor_with_symbolic_args 35.32 56.15 0.6290293855743544
ImmutableVarsTest.test_run_deployment(uint256) 97.5 132.43 0.736238012534924
StoreTest.testGasStoreWarmUp() 14.8 18.59 0.7961269499731038
FreshCheatcodes.test_int128() 11.53 14.14 0.8154172560113153
ExpectRevertTest.test_expectRevert_true() 14.15 16.3 0.8680981595092024
ExpectRevertTest.testFail_expectRevert_false() 16.7 19.13 0.872974385781495
EmitContractTest.testExpectEmitLessTopics() 14.11 16.11 0.8758535071384234
ConstructorTest.run_constructor() 10.63 12.04 0.8828903654485051
SetUpTest.testSetupData() 27.55 30.8 0.8944805194805194
BlockParamsTest.testChainId(uint256) 12.71 14.06 0.9039829302987198
ExpectRevertTest.test_ExpectRevert_increasedDepth() 17.39 19.18 0.9066736183524505
AssertTest.test_failing_branch(uint256) 44.93 49.51 0.9074934356695618
kontrol/src/tests/integration/test_foundry_prove.py::test_foundry_show_with_hex_encoding 4.63 4.87 0.9507186858316221
AccountParamsTest.testNonceSymbolic(uint64) 22.33 23.22 0.9616709732988803
BlockParamsTest.testWarp(uint256)-trace_options2 13.93 14.44 0.9646814404432134
AddrTest.test_builtInAddresses() 11.46 10.95 1.0465753424657536
SetUpDeployTest.test_extcodesize() 39.28 37.4 1.0502673796791444
ChainIdTest.test_chainid_setup() 26.8 25.35 1.057199211045365
CoinBaseTest.test_coinbase_setup() 25.48 24.01 1.0612244897959182
]) 30.64 28.0 1.0942857142857143
ExpectRevertTest.test_expectRevert_bytes4() 18.1 16.27 1.1124769514443762
AssertTest.test_revert_branch(uint256,uint256) 49.8 44.52 1.118598382749326
ExpectRevertTest.testFail_ExpectRevert_failAndSuccess() 18.83 16.81 1.1201665675193337
ExpectRevertTest.test_expectRevert_returnValue() 21.35 19.02 1.1225026288117772
AssumeTest.test_assume_false(uint256,uint256) 45.97 40.52 1.1345014807502467
EmitContractTest.testExpectEmitDoNotCheckData() 16.25 14.29 1.1371588523442968
StartPrankTestOrigin.test_startprank_origin_setup() 31.81 27.59 1.152953968829286
FreshCheatcodes.test_bool() 13.78 11.74 1.1737649063032367
kontrol/src/tests/integration/test_foundry_prove.py::test_foundry_resume_proof 36.17 30.47 1.1870692484410896
],uint256)) 27.66 23.3 1.1871244635193132
kontrol/src/tests/integration/test_foundry_prove.py::test_foundry_auto_abstraction 39.44 32.3 1.2210526315789474
LabelTest.testLabel() 13.87 11.01 1.259763851044505
AssertTest.testFail_assert_true() 39.92 30.06 1.3280106453759148
StoreTest.testStoreLoadNonExistent() 16.02 11.44 1.4003496503496504
kontrol/src/tests/integration/test_foundry_prove.py::test_foundry_merge_loop_heads 85.74 52.86 1.6220204313280362
TOTAL 972.05 959.53 1.0130480547768177

@jberthold jberthold requested review from geo2a and PetarMax October 3, 2024 09:07
@jberthold jberthold marked this pull request as ready for review October 3, 2024 09:07
@ehildenb
Copy link
Member

ehildenb commented Oct 3, 2024

Do we have a test merged into Kontrol or KEVM which demonstrates where this causes a performance improvement? We should definitely have that, especially if it is good for a client engagement.

@jberthold
Copy link
Member Author

Do we have a test merged into Kontrol or KEVM which demonstrates where this causes a performance improvement? We should definitely have that, especially if it is good for a client engagement.

Completely agree, especially since we see that slowdown in the sum-to-n. I changed the algorithm for @PetarMax who will hopefully be able to extract some tests from our recent engagement where he saw speed-up with this.

@geo2a geo2a removed their request for review November 25, 2024 07:47
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants