Drop flat caches #3009
Closed
Drop flat caches #3009
Conversation
This file contains 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
|
Code Climate has analyzed commit 984c318 and detected 4 issues on this pull request. Here's the issue category breakdown:
View more on Code Climate. |
dapplion
force-pushed
the
dapplion/drop-flat-caches
branch
from
0537bee
to
b953734
Compare
dapplion
force-pushed
the
dapplion/drop-flat-caches
branch
from
b953734
to
4e4e2be
Compare
dapplion
force-pushed
the
dapplion/drop-flat-caches
branch
from
b29db4f
to
b11e6e8
Compare
|
This pull request introduces 1 alert when merging b11e6e8 into 9fb1c42 - view on LGTM.com new alerts:
|
|
| Benchmark suite | Current: c4c95f9 | Previous: 241540f | Ratio |
|---|---|---|---|
| getPubkeys - persistent - req 1000 vs - 250000 vc | 5.1941 ms/op | 22.033 us/op | 235.74 |
🚀🚀 Significant benchmark improvement detected
| Benchmark suite | Current: c4c95f9 | Previous: 241540f | Ratio |
|---|---|---|---|
| phase0 processEffectiveBalanceUpdates - 250000 worstcase 0.5 | 39.428 ms/op | 3.4839 s/op | 0.01 |
| phase0 processSlashings - 250000 worstcase | 6.9683 ms/op | 100.08 ms/op | 0.07 |
Full benchmark results
| Benchmark suite | Current: c4c95f9 | Previous: 241540f | Ratio |
|---|---|---|---|
| state hashTreeRoot - No change | 777.00 ns/op | 878.00 ns/op | 0.88 |
| state hashTreeRoot - 1 full validator | 776.00 ns/op | 868.00 ns/op | 0.89 |
| state hashTreeRoot - 32 full validator | 771.00 ns/op | 883.00 ns/op | 0.87 |
| state hashTreeRoot - 512 full validator | 739.00 ns/op | 873.00 ns/op | 0.85 |
| state hashTreeRoot - 1 validator.effectiveBalance | 752.00 ns/op | 866.00 ns/op | 0.87 |
| state hashTreeRoot - 32 validator.effectiveBalance | 768.00 ns/op | 848.00 ns/op | 0.91 |
| state hashTreeRoot - 512 validator.effectiveBalance | 744.00 ns/op | 878.00 ns/op | 0.85 |
| state hashTreeRoot - 1 balances | 748.00 ns/op | 868.00 ns/op | 0.86 |
| state hashTreeRoot - 32 balances | 779.00 ns/op | 868.00 ns/op | 0.90 |
| state hashTreeRoot - 512 balances | 755.00 ns/op | 851.00 ns/op | 0.89 |
| state hashTreeRoot - 250000 balances | 768.00 ns/op | 882.00 ns/op | 0.87 |
| processSlot - 1 slots | 62.649 us/op | 64.710 us/op | 0.97 |
| processSlot - 32 slots | 2.8962 ms/op | 3.1689 ms/op | 0.91 |
| getCommitteeAssignments - req 1 vs - 250000 vc | 5.1063 ms/op | 6.6336 ms/op | 0.77 |
| getCommitteeAssignments - req 100 vs - 250000 vc | 7.5203 ms/op | 9.2732 ms/op | 0.81 |
| getCommitteeAssignments - req 1000 vs - 250000 vc | 8.4459 ms/op | 9.6479 ms/op | 0.88 |
| altair processBlock - 250000 vs - 7PWei normalcase | 160.66 ms/op | 142.61 ms/op | 1.13 |
| altair processBlock - 250000 vs - 7PWei worstcase | 377.92 ms/op | 342.08 ms/op | 1.10 |
| altair processEpoch - pyrmont_e62330 | 1.6495 s/op | 747.24 ms/op | 2.21 |
| pyrmont_e62330 - altair beforeProcessEpoch | 362.20 ms/op | 123.62 ms/op | 2.93 |
| pyrmont_e62330 - altair processJustificationAndFinalization | 138.10 us/op | 132.25 us/op | 1.04 |
| pyrmont_e62330 - altair processInactivityUpdates | 76.505 ms/op | 86.294 ms/op | 0.89 |
| pyrmont_e62330 - altair processRewardsAndPenalties | 107.43 ms/op | 180.41 ms/op | 0.60 |
| pyrmont_e62330 - altair processRegistryUpdates | 27.897 us/op | 12.213 us/op | 2.28 |
| pyrmont_e62330 - altair processSlashings | 6.0700 us/op | 3.6660 us/op | 1.66 |
| pyrmont_e62330 - altair processEth1DataReset | 5.8720 us/op | 3.0810 us/op | 1.91 |
| pyrmont_e62330 - altair processEffectiveBalanceUpdates | 818.25 ms/op | 30.376 ms/op | 26.94 |
| pyrmont_e62330 - altair processSlashingsReset | 30.324 us/op | 16.377 us/op | 1.85 |
| pyrmont_e62330 - altair processRandaoMixesReset | 43.430 us/op | 23.443 us/op | 1.85 |
| pyrmont_e62330 - altair processHistoricalRootsUpdate | 7.7440 us/op | 3.8510 us/op | 2.01 |
| pyrmont_e62330 - altair processParticipationFlagUpdates | 45.611 ms/op | 53.844 ms/op | 0.85 |
| pyrmont_e62330 - altair processSyncCommitteeUpdates | 5.0740 us/op | 2.5080 us/op | 2.02 |
| pyrmont_e62330 - altair afterProcessEpoch | 202.25 ms/op | 149.46 ms/op | 1.35 |
| altair processInactivityUpdates - 250000 normalcase | 208.40 ms/op | 292.83 ms/op | 0.71 |
| altair processInactivityUpdates - 250000 worstcase | 211.27 ms/op | 290.45 ms/op | 0.73 |
| altair processParticipationFlagUpdates - 250000 anycase | 83.820 ms/op | 107.39 ms/op | 0.78 |
| altair processRewardsAndPenalties - 250000 normalcase | 190.83 ms/op | 361.20 ms/op | 0.53 |
| altair processRewardsAndPenalties - 250000 worstcase | 189.73 ms/op | 393.33 ms/op | 0.48 |
| altair processSyncCommitteeUpdates - 250000 | 417.93 ms/op | 494.75 ms/op | 0.84 |
| Tree 40 250000 create | 489.23 ms/op | 710.92 ms/op | 0.69 |
| Tree 40 250000 get(125000) | 256.37 ns/op | 1.8362 us/op | 0.14 |
| Tree 40 250000 set(125000) | 1.7247 us/op | 1.9296 us/op | 0.89 |
| Tree 40 250000 toArray() | 41.463 ms/op | 48.023 ms/op | 0.86 |
| Tree 40 250000 iterate all - toArray() + loop | 40.546 ms/op | 48.148 ms/op | 0.84 |
| Tree 40 250000 iterate all - get(i) | 105.87 ms/op | 514.92 ms/op | 0.21 |
| MutableVector 250000 create | 22.868 ms/op | 30.630 ms/op | 0.75 |
| MutableVector 250000 get(125000) | 14.214 ns/op | 17.025 ns/op | 0.83 |
| MutableVector 250000 set(125000) | 689.45 ns/op | 848.56 ns/op | 0.81 |
| MutableVector 250000 toArray() | 7.8583 ms/op | 10.788 ms/op | 0.73 |
| MutableVector 250000 iterate all - toArray() + loop | 7.8008 ms/op | 11.250 ms/op | 0.69 |
| MutableVector 250000 iterate all - get(i) | 4.2874 ms/op | 4.8219 ms/op | 0.89 |
| Array 250000 create | 6.1262 ms/op | 7.1532 ms/op | 0.86 |
| Array 250000 clone - spread | 2.7747 ms/op | 5.0036 ms/op | 0.55 |
| Array 250000 get(125000) | 0.90200 ns/op | 1.6940 ns/op | 0.53 |
| Array 250000 set(125000) | 0.95500 ns/op | 1.0100 ns/op | 0.95 |
| Array 250000 iterate all - loop | 138.58 us/op | 209.91 us/op | 0.66 |
| aggregationBits - 2048 els - readonlyValues | 247.70 us/op | 324.58 us/op | 0.76 |
| aggregationBits - 2048 els - zipIndexesInBitList | 55.685 us/op | 40.774 us/op | 1.37 |
| ssz.Root.equals | 1.6440 us/op | 1.8000 us/op | 0.91 |
| ssz.Root.equals with valueOf() | 1.8230 us/op | 2.0150 us/op | 0.90 |
| byteArrayEquals with valueOf() | 1.7250 us/op | 1.9280 us/op | 0.89 |
| phase0 processBlock - 250000 vs - 7PWei normalcase | 16.653 ms/op | 18.496 ms/op | 0.90 |
| phase0 processBlock - 250000 vs - 7PWei worstcase | 107.56 ms/op | 112.05 ms/op | 0.96 |
| phase0 afterProcessEpoch - 250000 vs - 7PWei | 325.20 ms/op | 300.73 ms/op | 1.08 |
| phase0 beforeProcessEpoch - 250000 vs - 7PWei | 934.29 ms/op | 819.64 ms/op | 1.14 |
| phase0 processEpoch - mainnet_e58758 | 1.3111 s/op | 1.2654 s/op | 1.04 |
| mainnet_e58758 - phase0 beforeProcessEpoch | 833.89 ms/op | 675.01 ms/op | 1.24 |
| mainnet_e58758 - phase0 processJustificationAndFinalization | 132.64 us/op | 130.34 us/op | 1.02 |
| mainnet_e58758 - phase0 processRewardsAndPenalties | 113.62 ms/op | 255.13 ms/op | 0.45 |
| mainnet_e58758 - phase0 processRegistryUpdates | 100.61 us/op | 143.50 us/op | 0.70 |
| mainnet_e58758 - phase0 processSlashings | 6.2150 us/op | 3.1810 us/op | 1.95 |
| mainnet_e58758 - phase0 processEth1DataReset | 5.8680 us/op | 3.1100 us/op | 1.89 |
| mainnet_e58758 - phase0 processEffectiveBalanceUpdates | 35.007 ms/op | 32.417 ms/op | 1.08 |
| mainnet_e58758 - phase0 processSlashingsReset | 44.548 us/op | 19.907 us/op | 2.24 |
| mainnet_e58758 - phase0 processRandaoMixesReset | 55.024 us/op | 27.815 us/op | 1.98 |
| mainnet_e58758 - phase0 processHistoricalRootsUpdate | 8.5330 us/op | 3.7160 us/op | 2.30 |
| mainnet_e58758 - phase0 processParticipationRecordUpdates | 37.289 us/op | 16.688 us/op | 2.23 |
| mainnet_e58758 - phase0 afterProcessEpoch | 289.54 ms/op | 239.52 ms/op | 1.21 |
| phase0 processEffectiveBalanceUpdates - 250000 normalcase | 30.061 ms/op | 54.146 ms/op | 0.56 |
| phase0 processEffectiveBalanceUpdates - 250000 worstcase 0.5 | 39.428 ms/op | 3.4839 s/op | 0.01 |
| phase0 processRegistryUpdates - 250000 normalcase | 110.75 us/op | 71.990 us/op | 1.54 |
| phase0 processRegistryUpdates - 250000 badcase_full_deposits | 4.4748 ms/op | 6.3869 ms/op | 0.70 |
| phase0 processRegistryUpdates - 250000 worstcase 0.5 | 2.9396 s/op | 3.1595 s/op | 0.93 |
| phase0 getAttestationDeltas - 250000 normalcase | 64.208 ms/op | 59.576 ms/op | 1.08 |
| phase0 getAttestationDeltas - 250000 worstcase | 63.557 ms/op | 58.934 ms/op | 1.08 |
| phase0 processSlashings - 250000 worstcase | 6.9683 ms/op | 100.08 ms/op | 0.07 |
| shuffle list - 16384 els | 13.861 ms/op | 16.521 ms/op | 0.84 |
| shuffle list - 250000 els | 198.80 ms/op | 238.73 ms/op | 0.83 |
| getPubkeys - index2pubkey - req 1000 vs - 250000 vc | 2.6477 ms/op | 2.4955 ms/op | 1.06 |
| getPubkeys - validatorsArr - req 1000 vs - 250000 vc | 8.4779 ms/op | 547.73 us/op | 15.48 |
| getPubkeys - persistent - req 1000 vs - 250000 vc | 5.1941 ms/op | 22.033 us/op | 235.74 |
| BLS verify - blst-native | 2.4524 ms/op | 2.3246 ms/op | 1.06 |
| BLS verifyMultipleSignatures 3 - blst-native | 5.3278 ms/op | 4.7670 ms/op | 1.12 |
| BLS verifyMultipleSignatures 8 - blst-native | 11.353 ms/op | 10.294 ms/op | 1.10 |
| BLS verifyMultipleSignatures 32 - blst-native | 39.234 ms/op | 37.278 ms/op | 1.05 |
| BLS aggregatePubkeys 32 - blst-native | 52.295 us/op | 50.523 us/op | 1.04 |
| BLS aggregatePubkeys 128 - blst-native | 221.07 us/op | 194.79 us/op | 1.13 |
| getAttestationsForBlock | 87.324 ms/op | 96.716 ms/op | 0.90 |
| CheckpointStateCache - add get delete | 13.841 us/op | 25.551 us/op | 0.54 |
| validate gossip signedAggregateAndProof - struct | 5.9145 ms/op | 5.5563 ms/op | 1.06 |
| validate gossip signedAggregateAndProof - treeBacked | 6.1840 ms/op | 5.5206 ms/op | 1.12 |
| validate gossip attestation - struct | 2.6394 ms/op | 2.6041 ms/op | 1.01 |
| validate gossip attestation - treeBacked | 2.8565 ms/op | 2.6225 ms/op | 1.09 |
by benchmarkbot/action
Simplify command Fix PR issues Use .rc Add more accurate epoch perf tests Re-arrange epoch tests with a real state Fix perf type issues Ensure state is valid Review all spec tests Fix hash computations Fix processRewardsAndPenalties Fix perf titles Re-org imports Fix get state script Perf test hashing Fix altair step Review stfn caches and performance Add SLOW CODE comments and whitespace Cache effectiveBalances Process altair attestations in batch Don't keep withdrawalCredentials and pubkey in validators flat Move flat arrays to epoch process only Add exitQueue churn cache Cache chrunLimit Optimize beacon state transition Move exitQueue cached values to epochCtx Improve error message in processAttestation Add comments to processBlock functions Allow to set validators and balances Convert from struct to tree properly Silence effectiveBalances type warnings Add data representation tradeoffs Add more docs Fix type errors More fixes Use better state assertion in spec tests Re-add delete condition Fix type issues Update perf tests Fix lint warnings
dapplion
force-pushed
the
dapplion/drop-flat-caches
branch
from
3660440
to
7bf07f0
Compare
10 tasks
|
Replaced by #3760 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Motivation
CachedBeaconState duplicates much of the data in the state. Lodestar struggles with too much memory, so it's imperative data is represented only once.
Description
WIP