chore: Optimistic E3 Execution for Poseidon2 Chip

[nyunyunyunyu]

Add a flag to execute E3 optimistically. When the flag is enabled, verify_batch skips all verification logic and postpone all poseidon2 hash computation to the tracegen phase.

Reth-benchmark shows 20% execution time reduction. Tracegen time slightly increases but the increment is less than the saving.

Close INT-4374

Per opcode stats on my laptop(variance is large but we could see a significant difference):

Before:

GROUP	FREQUENCY	TOTAL EXEC TIME	AVG TIME	Percentage
ADD	599175	46411254	77.46	9.30%
BBE4DIV	19534	7959238	407.46	1.60%
BBE4MUL	52005	9243036	177.73	1.85%
COMP_POS2	27	53917	1,996.93	0.01%
DIV	2184	390832	178.95	0.08%
FE4ADD	27755	3580894	129.02	0.72%
FE4SUB	6709	1523758	227.12	0.31%
FRI_REDUCED_OPENING	17500	47049340	2,688.53	9.43%
HINT_STOREW	54961	6639322	120.80	1.33%
JAL	20264	1670740	82.45	0.33%
LOADW	332530	56095576	168.69	11.25%
MUL	146789	11367159	77.44	2.28%
NativeBranchEqualOpcode(BEQ)	8951	570160	63.70	0.11%
NativeBranchEqualOpcode(BNE)	151700	10170358	67.04	2.04%
PERM_POS2	1968	3516015	1,786.59	0.70%
PHANTOM	16541	1531674	92.60	0.31%
PUBLISH	36	16165	449.03	0.00%
RANGE_CHECK	26909	1374440	51.08	0.28%
STOREW	72946	10643024	145.90	2.13%
SUB	61182	7812138	127.69	1.57%
VERIFY_BATCH	2700	271206706	100,446.93	54.37%
		498825746		100.00%

After:

GROUP	FREQUENCY	TOTAL EXEC TIME	AVG TIME	Percentage
ADD	599175	50,137,976	83.68	11.65%
BBE4DIV	19534	10,487,243	536.87	2.44%
BBE4MUL	52005	11,279,418	216.89	2.62%
COMP_POS2	27	109,915	4,070.93	0.03%
DIV	2184	502,509	230.09	0.12%
FE4ADD	27755	4,140,336	149.17	0.96%
FE4SUB	6709	1,979,492	295.05	0.46%
FRI_REDUCED_OPENING	17500	57,441,000	3,282.34	13.34%
HINT_STOREW	54961	9,498,320	172.82	2.21%
JAL	20193	1,504,236	74.49	0.35%
LOADW	332530	60,762,686	182.73	14.12%
MUL	146789	12,791,836	87.14	2.97%
NativeBranchEqualOpcode(BEQ)	8951	662,178	73.98	0.15%
NativeBranchEqualOpcode(BNE)	151700	10,591,935	69.82	2.46%
PERM_POS2	1968	4,060,309	2,063.17	0.94%
PHANTOM	16541	2,082,586	125.90	0.48%
PUBLISH	36	233,085	6,474.58	0.05%
RANGE_CHECK	26909	1,326,437	49.29	0.31%
STOREW	72946	10,698,539	146.66	2.49%
SUB	61182	9,573,439	156.47	2.22%
VERIFY_BATCH	2700	170,589,925	63,181.45	39.63%
		430,453,400		100.00%

[codspeed-hq]

CodSpeed WallTime Performance Report

Merging #1896 will not alter performance

_{Comparing chore/e3-poseidon (b98c4a3) with feat/new-execution (ffc9f3e)}

⚠️ 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

✅ 24 untouched benchmarks

[jonathanpwang]

Please provide an update on the measured time in insn/s of just verifybatch before and after this PR

[github-actions]

group	app.proof_time_ms	app.cycles	app.cells_used	leaf.proof_time_ms	leaf.cycles	leaf.cells_used
verify_fibair	(-9 [-0.8%]) 1,055	322,700	17,672,408	-	-	-
fibonacci	2,110	3,000,420	51,503,521	2,932	1,248,029	70,625,644
regex	6,894	8,217,194	163,669,874	8,656	3,326,693	234,118,948
ecrecover	(+23 [+2.1%]) 1,138	274,568	8,596,011	9,951	2,934,871	246,667,012
pairing	(+99 [+2.7%]) 3,714	3,725,930	93,105,105	4,361	2,010,448	138,831,313
fib_e2e	19,462	24,000,420	410,933,023	16,196	7,462,555	435,494,883
kitchen_sink	14,392	309,192	899,580,290	21,492	7,952,478	748,994,058

Commit: b98c4a3

Benchmark Workflow

[jonathanpwang]

maybe you could have used RecordSeeker for this, not sure

[jonathanpwang]

LGTM