Add perf profiling to tinybench

[GuillaumeLagrange]

https://github.com/CodSpeedHQ/codspeed-node/pull/51/files#diff-37f58390c47ed415f315435a8b448d14d00f4fb914e205d86410d694c3215d94R28-R38

Explains the 4 added walltime benches (fixes a bug where only last bench result of a bid was written to json)

[codspeed-hq]

CodSpeed Instrumentation Performance Report

Merging #51 will degrade performances by 84.69%

_{Comparing cod-1221-support-perf-profiling-with-codspeed-node (e31ee8b) with main (81f0dc9)}

Summary

⚡ 7 improvements
❌ 11 regressions
✅ 64 untouched

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

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
❌	switch 1	30 µs	123.9 µs	-75.76%
⚡	test_iterative_fibo_10	450.4 µs	299.4 µs	+50.4%
⚡	test_iterative_fibo_100	194.6 µs	34 µs	×5.7
⚡	test_recursive_cached_fibo_30	205.7 µs	48.9 µs	×4.2
❌	test_recursive_fibo_10	46 µs	138.4 µs	-66.81%
⚡	test sync baz 10	201.3 µs	40.9 µs	×4.9
❌	test_recursive_fibo_20	1.5 ms	1.8 ms	-15.84%
❌	test_iterative_fibo_10	29.1 µs	190.3 µs	-84.69%
⚡	test_iterative_fibo_100	282.3 µs	214.2 µs	+31.76%
❌	test_recursive_cached_fibo_30	47.2 µs	227.5 µs	-79.25%
❌	test_recursive_fibo_10	46 µs	139 µs	-66.92%
❌	test sync baz 10	39.4 µs	200 µs	-80.31%
❌	switch 1	30.7 µs	125.2 µs	-75.47%
❌	switch 1	30.7 µs	123.7 µs	-75.2%
❌	short body	102.5 µs	121.1 µs	-15.37%
❌	switch 1	30.9 µs	93.3 µs	-66.94%
⚡	switch 2	25.5 µs	8.1 µs	×3.2
⚡	wait 1ms	24.7 µs	19.8 µs	+24.92%

[codspeed-hq]

CodSpeed WallTime Performance Report

Merging #51 will degrade performances by 50%

_{Comparing cod-1221-support-perf-profiling-with-codspeed-node (e31ee8b) with main (81f0dc9)}

Summary

❌ 18 regressions
✅ 26 untouched
🆕 4 new

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

Benchmarks breakdown

	Benchmark	BASE	HEAD	Change
❌	switch 1	156 ns	300 ns	-48%
❌	switch 2	156 ns	300 ns	-48%
❌	test_iterative_fibo_10	216 ns	372 ns	-41.94%
❌	test_iterative_fibo_100	504 ns	648 ns	-22.22%
❌	test_recursive_cached_fibo_10	984 ns	1,164 ns	-15.46%
❌	test_recursive_cached_fibo_30	3.1 µs	3.5 µs	-11.11%
❌	test sync baz 10	156 ns	300 ns	-48%
❌	test sync baz 100	288 ns	432 ns	-33.33%
❌	test_iterative_fibo_10	228 ns	396 ns	-42.42%
❌	test_iterative_fibo_100	504 ns	672 ns	-25%
❌	test_recursive_cached_fibo_10	996 ns	1,164 ns	-14.43%
❌	test sync baz 10	168 ns	312 ns	-46.15%
❌	test sync baz 100	288 ns	432 ns	-33.33%
❌	switch 1	144 ns	276 ns	-47.83%
❌	switch 2	156 ns	312 ns	-50%
❌	switch 1	156 ns	264 ns	-40.91%
❌	switch 2	168 ns	300 ns	-44%
🆕	long body	N/A	217.6 µs	N/A
🆕	short body	N/A	1.8 µs	N/A
🆕	switch 1	N/A	144 ns	N/A
...	...	...	...	...

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

[adriencaccia]

Almost there 💪

[art049]

Do we need that much duplication between instrumented.ts and walltime.ts? Since we already have unavoidable duplication between sync and async, it gets a bit hard to verify the logic

[GuillaumeLagrange]

Do we need that much duplication between instrumented.ts and walltime.ts? Since we already have unavoidable duplication between sync and async, it gets a bit hard to verify the logic

I've adressed this in d9c546a

This commit also removes the setupCore issue you pointed out

[GuillaumeLagrange]

@art049 the x1000 in instrumented runs are fixed. We were actually measuring the garbage collector call because of a refactoring error. Changes look more in line with what's expected now

[art049]

finally