perf(levm): use specialized PUSH1 and PUSH2 implementations #3262

edg-l · 2025-06-23T06:43:47Z

Motivation
According to stats from @azteca1998 PUSH2 and PUSH1 are widely used:

Loaded 903636264 rows (3447.10MiB)
Stats (of 903636264 records):
  0xf1: count=   730979  t_min=  2278  t_max=1512728  t_avg=110877.43  t_acc=81049072024  CALL
  0x61: count=131856777  t_min=   136  t_max= 549032  t_avg=   189.29  t_acc=24959614846  PUSH2
  0x56: count= 78745029  t_min=   170  t_max=1488792  t_avg=   243.75  t_acc=19194034756  JUMP
  0x60: count= 86327863  t_min=   136  t_max= 837080  t_avg=   199.78  t_acc=17246262544  PUSH1
  0x5b: count=107216057  t_min=   102  t_max= 267308  t_avg=   159.43  t_acc=17093508806  JUMPDEST
  0x50: count= 86546732  t_min=   102  t_max= 353260  t_avg=   174.49  t_acc=15101132640  POP
  0x57: count= 53096953  t_min=   102  t_max=1382576  t_avg=   233.40  t_acc=12393069292  JUMPI
  0x81: count= 55585321  t_min=   102  t_max= 267410  t_avg=   192.79  t_acc=10716509980  DUP2
  0x01: count= 56493418  t_min=   102  t_max=1431060  t_avg=   189.52  t_acc=10706399944  ADD
  0x91: count= 31380921  t_min=   102  t_max= 146030  t_avg=   205.38  t_acc= 6444862520  SWAP2

Furthermore i keep seeing U256::from_big_endian taking quite some time on samply so I made specialized PUSH1 and PUSH2 implementations that avoid that, also using fixed size arrays.

Benchmarks:

Hoodi 11k:

main 9m10.471s
pr 8m25.933s

Description

Closes #issue_number

Benchmark Results Comparison

Benchmark Results: Factorial

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`levm_Factorial_pr`	634.2 ± 7.3	629.6	654.2	2.71 ± 0.04
`levm_Factorial`	726.1 ± 5.2	722.5	740.1	3.11 ± 0.03
`levm_FactorialRecursive_pr`	3.567 ± 0.021	3.541	3.604	2.22 ± 0.05
`levm_FactorialRecursive`	3.828 ± 0.035	3.775	3.889	2.39 ± 0.03
`levm_Fibonacci_pr`	629.2 ± 6.4	625.7	646.9	2.99 ± 0.03
`levm_Fibonacci`	727.7 ± 6.5	722.3	743.9	3.47 ± 0.03
`levm_ManyHashes_pr`	14.9 ± 0.2	14.7	15.3	1.70 ± 0.03
`levm_ManyHashes`	16.3 ± 0.1	16.2	16.4	1.87 ± 0.02
`levm_BubbleSort_pr`	5.065 ± 0.023	5.034	5.107	1.58 ± 0.01
`levm_BubbleSort`	5.508 ± 0.035	5.489	5.603	1.71 ± 0.02
`levm_ERC20Transfer_pr`	461.5 ± 1.3	459.7	463.4	1.87 ± 0.03
`levm_ERC20Transfer`	487.9 ± 2.4	484.1	491.0	1.99 ± 0.01
`levm_ERC20Mint_pr`	306.8 ± 8.9	300.1	328.5	2.22 ± 0.07
`levm_ERC20Mint`	320.1 ± 1.5	317.9	322.6	2.31 ± 0.05
`levm_ERC20Approval_pr`	1.779 ± 0.023	1.763	1.838	1.69 ± 0.02
`levm_ERC20Approval`	1.850 ± 0.011	1.837	1.873	1.76 ± 0.02

According to the samply this makes op_push nearly negligible (from 30% to 0%)

github-actions · 2025-06-23T06:45:45Z

Lines of code report

Total lines added: 39
Total lines removed: 0
Total lines changed: 39

Detailed view

+---------------------------------------------------+-------+------+
| File                                              | Lines | Diff |
+---------------------------------------------------+-------+------+
| ethrex/crates/vm/levm/src/execution_handlers.rs   | 208   | +2   |
+---------------------------------------------------+-------+------+
| ethrex/crates/vm/levm/src/opcode_handlers/push.rs | 78    | +37  |
+---------------------------------------------------+-------+------+

github-actions · 2025-06-23T06:54:32Z

Benchmark Results Comparison

PR Results

Benchmark Results: Factorial

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`revm_Factorial`	234.1 ± 1.5	232.2	237.1	1.00
`levm_Factorial`	673.3 ± 5.4	668.1	683.9	2.88 ± 0.03

Benchmark Results: Factorial - Recursive

Command	Mean [s]	Min [s]	Max [s]	Relative
`revm_FactorialRecursive`	1.616 ± 0.038	1.534	1.678	1.00
`levm_FactorialRecursive`	3.634 ± 0.025	3.606	3.678	2.25 ± 0.05

Benchmark Results: Fibonacci

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`revm_Fibonacci`	208.9 ± 1.4	205.9	210.5	1.00
`levm_Fibonacci`	675.0 ± 1.7	673.6	679.1	3.23 ± 0.02

Benchmark Results: ManyHashes

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`revm_ManyHashes`	9.0 ± 0.1	8.8	9.2	1.00
`levm_ManyHashes`	15.0 ± 0.1	14.9	15.2	1.67 ± 0.03

Benchmark Results: BubbleSort

Command	Mean [s]	Min [s]	Max [s]	Relative
`revm_BubbleSort`	3.191 ± 0.022	3.170	3.222	1.00
`levm_BubbleSort`	5.162 ± 0.047	5.118	5.282	1.62 ± 0.02

Benchmark Results: ERC20 - Transfer

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`revm_ERC20Transfer`	244.1 ± 1.5	242.5	246.6	1.00
`levm_ERC20Transfer`	474.9 ± 4.9	468.8	484.0	1.95 ± 0.02

Benchmark Results: ERC20 - Mint

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`revm_ERC20Mint`	136.7 ± 0.8	135.6	138.4	1.00
`levm_ERC20Mint`	304.2 ± 2.8	299.9	310.2	2.23 ± 0.02

Benchmark Results: ERC20 - Approval

Command	Mean [s]	Min [s]	Max [s]	Relative
`revm_ERC20Approval`	1.062 ± 0.084	1.032	1.303	1.00
`levm_ERC20Approval`	1.787 ± 0.006	1.781	1.797	1.68 ± 0.13

Main Results

Benchmark Results: Factorial

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`revm_Factorial`	234.0 ± 6.6	230.6	252.7	1.00
`levm_Factorial`	718.6 ± 5.2	715.2	732.3	3.07 ± 0.09

Benchmark Results: Factorial - Recursive

Command	Mean [s]	Min [s]	Max [s]	Relative
`revm_FactorialRecursive`	1.588 ± 0.039	1.511	1.640	1.00
`levm_FactorialRecursive`	3.826 ± 0.037	3.784	3.891	2.41 ± 0.06

Benchmark Results: Fibonacci

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`revm_Fibonacci`	207.3 ± 0.4	206.4	207.8	1.00
`levm_Fibonacci`	726.3 ± 5.5	722.0	740.9	3.50 ± 0.03

Benchmark Results: ManyHashes

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`revm_ManyHashes`	8.7 ± 0.1	8.6	8.8	1.00
`levm_ManyHashes`	15.8 ± 0.1	15.6	15.9	1.81 ± 0.02

Benchmark Results: BubbleSort

Command	Mean [s]	Min [s]	Max [s]	Relative
`revm_BubbleSort`	3.200 ± 0.025	3.172	3.258	1.00
`levm_BubbleSort`	5.384 ± 0.041	5.350	5.480	1.68 ± 0.02

Benchmark Results: ERC20 - Transfer

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`revm_ERC20Transfer`	243.7 ± 2.6	241.8	248.6	1.00
`levm_ERC20Transfer`	492.7 ± 3.4	489.1	500.0	2.02 ± 0.03

Benchmark Results: ERC20 - Mint

Command	Mean [ms]	Min [ms]	Max [ms]	Relative
`revm_ERC20Mint`	136.9 ± 0.7	136.1	138.4	1.00
`levm_ERC20Mint`	321.9 ± 4.0	318.1	329.0	2.35 ± 0.03

Benchmark Results: ERC20 - Approval

Command	Mean [s]	Min [s]	Max [s]	Relative
`revm_ERC20Approval`	1.062 ± 0.043	1.040	1.184	1.00
`levm_ERC20Approval`	1.864 ± 0.012	1.850	1.878	1.75 ± 0.07

Arkenan

Approved with a comment

crates/vm/levm/src/opcode_handlers/push.rs

github-actions · 2025-06-23T12:06:54Z

Benchmark Block Execution Results Comparison Against Main

Command	Mean [s]	Min [s]	Max [s]	Relative
`base`	227.127 ± 0.927	225.807	228.843	1.01 ± 0.01
`head`	225.031 ± 0.782	224.062	226.714	1.00

crates/vm/levm/src/opcode_handlers/push.rs

Co-authored-by: Jeremías Salomón <[email protected]>

perf(levm): use specialized PUSH1 and PUSH2 implementations

e881c69

edg-l added the performance label Jun 23, 2025

edg-l marked this pull request as ready for review June 23, 2025 09:40

edg-l requested a review from a team as a code owner June 23, 2025 09:40

changelog

f6b4c9e

Oppen approved these changes Jun 23, 2025

View reviewed changes

Merge branch 'main' into improve_push1_2

769d0cf

Arkenan approved these changes Jun 23, 2025

View reviewed changes

crates/vm/levm/src/opcode_handlers/push.rs Outdated Show resolved Hide resolved

docs

2f9be6c

JereSalo reviewed Jun 23, 2025

View reviewed changes

crates/vm/levm/src/opcode_handlers/push.rs Outdated Show resolved Hide resolved

JereSalo approved these changes Jun 23, 2025

View reviewed changes

Update crates/vm/levm/src/opcode_handlers/push.rs

81a4d29

Co-authored-by: Jeremías Salomón <[email protected]>

edg-l enabled auto-merge June 23, 2025 13:05

Merge branch 'main' into improve_push1_2

01a305d

edg-l added this pull request to the merge queue Jun 23, 2025

Merged via the queue into main with commit bca73af Jun 23, 2025
33 of 36 checks passed

edg-l deleted the improve_push1_2 branch June 23, 2025 13:50

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

perf(levm): use specialized PUSH1 and PUSH2 implementations #3262

perf(levm): use specialized PUSH1 and PUSH2 implementations #3262

Uh oh!

edg-l commented Jun 23, 2025 •

edited

Loading

Uh oh!

github-actions bot commented Jun 23, 2025

Uh oh!

github-actions bot commented Jun 23, 2025 •

edited

Loading

Uh oh!

Arkenan left a comment

Uh oh!

Uh oh!

github-actions bot commented Jun 23, 2025 •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

Uh oh!

perf(levm): use specialized PUSH1 and PUSH2 implementations #3262

perf(levm): use specialized PUSH1 and PUSH2 implementations #3262

Uh oh!

Conversation

edg-l commented Jun 23, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchmark Results Comparison

Benchmark Results: Factorial

Uh oh!

github-actions bot commented Jun 23, 2025

Lines of code report

Uh oh!

github-actions bot commented Jun 23, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchmark Results Comparison

PR Results

Benchmark Results: Factorial

Benchmark Results: Factorial - Recursive

Benchmark Results: Fibonacci

Benchmark Results: ManyHashes

Benchmark Results: BubbleSort

Benchmark Results: ERC20 - Transfer

Benchmark Results: ERC20 - Mint

Benchmark Results: ERC20 - Approval

Main Results

Benchmark Results: Factorial

Benchmark Results: Factorial - Recursive

Benchmark Results: Fibonacci

Benchmark Results: ManyHashes

Benchmark Results: BubbleSort

Benchmark Results: ERC20 - Transfer

Benchmark Results: ERC20 - Mint

Benchmark Results: ERC20 - Approval

Uh oh!

Arkenan left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

github-actions bot commented Jun 23, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchmark Block Execution Results Comparison Against Main

Uh oh!

Uh oh!

Uh oh!

Uh oh!

edg-l commented Jun 23, 2025 •

edited

Loading

github-actions bot commented Jun 23, 2025 •

edited

Loading

github-actions bot commented Jun 23, 2025 •

edited

Loading