[bal-devnet-3] perf: exec parallel workers=numcpu#21296
Merged
Conversation
taratorio
added a commit
that referenced
this pull request
May 20, 2026
default **Previous run** `1779109056_a101d4ed_erigon-bal-full` — `numWorkers = runtime.NumCPU() / 2 = 8` (on this 16-core box). **New run** `1779246695_cc99a1f0_erigon-bal-full` — `numWorkers = runtime.NumCPU() = 16`. Both runs use the same `erigon-local:traced` image with the KZG async warmup + BAL `tx.Apply` read fix; only the worker count changed. Sorted by speedup desc. | Test | NumCPU/2 (MGas/s) | NumCPU (MGas/s) | Speedup | |---|---:|---:|---:| | `test_mod_arithmetic-MULMOD-mod191-120M` | 386.8 | 593.3 | **1.53×** | | `test_mod-SMOD-mod255-120M` | 538.6 | 794.3 | 1.47× | | `test_mod-SMOD-mod63-120M` | 639.6 | 929.2 | 1.45× | | `test_point_evaluation-120M` | 238.5 | 343.2 | 1.44× | | `test_mod-MOD-mod127-120M` | 532.9 | 711.1 | 1.33× | | `test_mod-MOD-mod191-120M` | 631.7 | 821.4 | 1.30× | | `test_mod_arithmetic-ADDMOD-mod191-120M` | 602.4 | 782.5 | 1.30× | | `test_mod-SMOD-mod127-120M` | 521.6 | 668.8 | 1.28× | | `test_mod-MOD-mod255-120M` | 639.2 | 803.9 | 1.26× | | `test_ecrecover-120M` | 362.0 | 455.0 | 1.26× | | `test_blake2f_benchmark-rounds6-120M` | 282.2 | 352.8 | 1.25× | | `test_mod-SMOD-mod191-120M` | 584.2 | 720.8 | 1.23× | | `test_blake2f_benchmark-rounds24-120M` | 305.6 | 356.1 | 1.17× | | `test_p256verify-120M` | 478.0 | 556.2 | 1.16× | | `test_blake2f_benchmark-rounds1-120M` | 293.2 | 336.3 | 1.15× | | `test_mod-MOD-mod63-120M` | 837.2 | 872.6 | 1.04× | | `test_blake2f_benchmark-rounds12-120M` | 318.0 | 313.9 | 0.99× | | **average MGas/s** | **481.8** | **612.4** | **1.27×** | Doubling the default worker count gives **~27% throughput improvement on average**, with the biggest wins on heavier mod-arithmetic and SMOD tests (45–53%) and the smallest impact on blake2f_rounds12 (basically noise).
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
2 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.
Benchmarkoor: NumCPU/2 (previous) vs NumCPU (new) for
Exec3WorkersdefaultPrevious run
1779109056_a101d4ed_erigon-bal-full—numWorkers = runtime.NumCPU() / 2 = 8(on this 16-core box).New run
1779246695_cc99a1f0_erigon-bal-full—numWorkers = runtime.NumCPU() = 16.Both runs use the same
erigon-local:tracedimage with the KZG async warmup + BALtx.Applyread fix; only the worker count changed. Sorted by speedup desc.test_mod_arithmetic-MULMOD-mod191-120Mtest_mod-SMOD-mod255-120Mtest_mod-SMOD-mod63-120Mtest_point_evaluation-120Mtest_mod-MOD-mod127-120Mtest_mod-MOD-mod191-120Mtest_mod_arithmetic-ADDMOD-mod191-120Mtest_mod-SMOD-mod127-120Mtest_mod-MOD-mod255-120Mtest_ecrecover-120Mtest_blake2f_benchmark-rounds6-120Mtest_mod-SMOD-mod191-120Mtest_blake2f_benchmark-rounds24-120Mtest_p256verify-120Mtest_blake2f_benchmark-rounds1-120Mtest_mod-MOD-mod63-120Mtest_blake2f_benchmark-rounds12-120MDoubling the default worker count gives ~27% throughput improvement on average, with the biggest wins on heavier mod-arithmetic and SMOD tests (45–53%) and the smallest impact on blake2f_rounds12 (basically noise).