feat(criterion_compat): fork criterion and add walltime support

[not-matthias]

No description provided.

[codspeed-hq]

CodSpeed Instrumentation Performance Report

Merging #85 will degrade performances by 16.51%

_{Comparing cod-240-walltime-criterion-support (9102482) with main (17a5e27)}

Summary

⚡ 1 improvements
❌ 6 regressions
✅ 156 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
❌	from_elem[16384]	19 µs	19.6 µs	-3.12%
❌	from_elem[2048]	3.8 µs	4.1 µs	-5.48%
❌	from_elem[4096]	6.6 µs	7.2 µs	-8.77%
❌	from_elem[8192]	10.7 µs	11.4 µs	-5.48%
❌	from_elem_decimal[2048]	3.4 µs	4.1 µs	-16.51%
⚡	mut_borrow	895.8 ns	866.7 ns	+3.37%
❌	recursive_memoized[BTreeMap<u64, u64>, 0]	371.7 ns	400.8 ns	-7.28%

[codspeed-hq]

CodSpeed Walltime Performance Report

Merging #85 will degrade performances by 72.66%

_{Comparing cod-240-walltime-criterion-support (9102482) with main (17a5e27)}

Summary

⚡ 1 improvements
❌ 16 regressions
✅ 92 untouched benchmarks
🆕 36 new benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
🆕	sum_fold	N/A	304 ns	N/A
🆕	build_vec	N/A	204 ns	N/A
🆕	Iterative	N/A	21 ns	N/A
🆕	Recursive	N/A	52.3 µs	N/A
🆕	Iterative[20]	N/A	21 ns	N/A
🆕	Iterative[21]	N/A	22 ns	N/A
🆕	Recursive[20]	N/A	52.3 µs	N/A
🆕	Recursive[21]	N/A	84.7 µs	N/A
🆕	fibonacci_custom_measurement	N/A	421 ns	N/A
🆕	large_drop	N/A	169.9 µs	N/A
🆕	small_drop	N/A	< 1 ns	N/A
🆕	large_setup	N/A	13 ns	N/A
🆕	small_setup	N/A	1 ns	N/A
🆕	iter_with_setup	N/A	39 ns	N/A
🆕	iter	N/A	< 1 ns	N/A
🆕	iter_batched_large_input	N/A	< 1 ns	N/A
🆕	iter_batched_per_iteration	N/A	40 ns	N/A
🆕	iter_batched_ref_large_input	N/A	< 1 ns	N/A
🆕	iter_batched_ref_per_iteration	N/A	40 ns	N/A
🆕	iter_batched_ref_small_input	N/A	< 1 ns	N/A
...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

[GuillaumeLagrange]

Biggest change is how we handle the fork which feels a bit uncomfortable, I would like us to be able to have more control over clippy to make diff completely minimal with original criterion

[GuillaumeLagrange]

very small comments, we can merge after this !

[GuillaumeLagrange]

LGTM!

Let's wait for the display issue with 0 time benchmarks before merging, I just want to sanity check that the perf regression reported by codspeed is expected/known and not caused by changes here, but can't really do it without the report page

[not-matthias]

Verified execution time of benchmarks with 0s, by comparing the measured walltime with local results:

• small_drop 143.00 ps
• iter 128.60 ps
• iter_batched_large_input 89.355 ps
• iter_batched_ref_large_input 87.253 ps
• etc.

Looks good, so I don't think we have a bug in there.

[art049]

Just some small structure details. Otherwise looks good

[not-matthias]

Link for the uv-walltime-criterion integration: https://codspeed.io/AvalancheHQ/uv-criterion-walltime/runs

[art049]

lets go!