feat: use memmap instead of paged_vecs on platforms that support it

[pjabbarzade]

resolve INT-3754

reth benchmark comparison: https://github.com/axiom-crypto/openvm-reth-benchmark/actions/runs/15863851635

[codspeed-hq]

CodSpeed WallTime Performance Report

Merging #1734 will improve performances by 44.58%

_{Comparing feat/new-execution-mmap (bb09764) with feat/new-execution (6446607)}

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

Summary

⚡ 20 improvements

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
⚡	benchmark_execute[bubblesort]	330.4 ms	166.4 ms	+98.57%
⚡	benchmark_execute[factorial_iterative_u256]	742.4 ms	443.8 ms	+67.28%
⚡	benchmark_execute[fibonacci_iterative]	351.6 ms	174.3 ms	×2
⚡	benchmark_execute[fibonacci_recursive]	447.6 ms	223 ms	×2
⚡	benchmark_execute[keccak256]	343.6 ms	165.5 ms	×2.1
⚡	benchmark_execute[keccak256_iter]	489.7 ms	247.2 ms	+98.1%
⚡	benchmark_execute[quicksort]	352.6 ms	171.7 ms	×2.1
⚡	benchmark_execute[revm_snailtracer]	279.6 ms	139.5 ms	×2
⚡	benchmark_execute[sha256]	335.7 ms	166.5 ms	×2
⚡	benchmark_execute[sha256_iter]	500.2 ms	263.2 ms	+90.1%
⚡	benchmark_execute_metered[bubblesort]	429.8 ms	243.5 ms	+76.52%
⚡	benchmark_execute_metered[factorial_iterative_u256]	1,098.4 ms	759.7 ms	+44.58%
⚡	benchmark_execute_metered[fibonacci_iterative]	489.2 ms	301.5 ms	+62.24%
⚡	benchmark_execute_metered[fibonacci_recursive]	601.5 ms	369.8 ms	+62.66%
⚡	benchmark_execute_metered[keccak256]	455 ms	249.9 ms	+82.08%
⚡	benchmark_execute_metered[keccak256_iter]	681.5 ms	429.4 ms	+58.69%
⚡	benchmark_execute_metered[quicksort]	458.4 ms	262.6 ms	+74.58%
⚡	benchmark_execute_metered[revm_snailtracer]	290.5 ms	140.4 ms	×2.1
⚡	benchmark_execute_metered[sha256]	445.5 ms	249.2 ms	+78.76%
⚡	benchmark_execute_metered[sha256_iter]	698.1 ms	441.4 ms	+58.13%

[codspeed-hq]

CodSpeed Instrumentation Performance Report

Merging #1734 will improve performances by 11.69%

_{Comparing feat/new-execution-mmap (b647e68) with feat/new-execution (70ae977)}

Summary

⚡ 20 improvements

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
⚡	benchmark_execute[bubblesort]	1.6 s	1.1 s	+45.43%
⚡	benchmark_execute[factorial_iterative_u256]	4.1 s	3.2 s	+30.38%
⚡	benchmark_execute[fibonacci_iterative]	1.6 s	1.1 s	+48.05%
⚡	benchmark_execute[fibonacci_recursive]	2.2 s	1.5 s	+42.81%
⚡	benchmark_execute[keccak256]	1.7 s	1.2 s	+44.91%
⚡	benchmark_execute[keccak256_iter]	2.5 s	1.8 s	+36.95%
⚡	benchmark_execute[quicksort]	1.8 s	1.2 s	+44.06%
⚡	benchmark_execute[revm_snailtracer]	910.5 ms	541.5 ms	+68.14%
⚡	benchmark_execute[sha256]	1.7 s	1.2 s	+43.84%
⚡	benchmark_execute[sha256_iter]	2.8 s	2.1 s	+32.91%
⚡	benchmark_execute_metered[bubblesort]	2.6 s	2.1 s	+19.97%
⚡	benchmark_execute_metered[factorial_iterative_u256]	7.1 s	6.4 s	+11.69%
⚡	benchmark_execute_metered[fibonacci_iterative]	2.6 s	2.2 s	+20.87%
⚡	benchmark_execute_metered[fibonacci_recursive]	3.5 s	3 s	+18%
⚡	benchmark_execute_metered[keccak256]	2.9 s	2.4 s	+19.46%
⚡	benchmark_execute_metered[keccak256_iter]	4.1 s	3.5 s	+16.67%
⚡	benchmark_execute_metered[quicksort]	2.9 s	2.4 s	+19.08%
⚡	benchmark_execute_metered[revm_snailtracer]	925.9 ms	540.7 ms	+71.25%
⚡	benchmark_execute_metered[sha256]	2.9 s	2.4 s	+19.22%
⚡	benchmark_execute_metered[sha256_iter]	4.5 s	3.9 s	+14.86%

[jonathanpwang]

The micro-benchmarks have a much higher tracegen time because the memory image to clone for the initial merkle tree is now bigger. This is only a significant factor because the microbenchmarks have such small memory allocated.

[jonathanpwang]

LGTM

The sdk and cli tests are now very slow, maybe we can increase the test-threads a bit if the OOM is not a concern.

[jonathanpwang]

The merkle memory finalization is slower in micro-benchmarks, and there are ways to address this (notably the merkle tree construction can be updated to directly use a segment tree now). However to avoid making this PR balloon even more and also to prevent merge conflicts, we will merge it first.

[github-actions]

group	app.proof_time_ms	app.cycles	app.cells_used	leaf.proof_time_ms	leaf.cycles	leaf.cells_used
verify_fibair	(+171 [+13.6%]) 1,426	322,699	17,340,154	-	-	-
fibonacci	(+395 [+18.2%]) 2,565	1,500,277	50,589,503	3,604	1,248,092	69,834,746
regex	(+674 [+9.9%]) 7,489	4,165,432	166,519,456	10,659	3,349,040	228,918,339
ecrecover	(+542 [+48.2%]) 1,666	137,260	8,151,353	11,193	2,934,925	241,888,862
pairing	(+595 [+16.1%]) 4,302	1,862,964	97,508,739	5,515	2,010,485	134,810,861

Commit: bb09764

Benchmark Workflow