bench_static: Add benchmarks for p{exp,dep}s{,_1} #14

jepler commented 2026年03月28日 01:09:19 +01:00

Typical output, i5-1235U, performance core

estimated overhead 1.195 msec ( 1.195 nsec per iteration ) last: d423b7ad3868912d
pext: 1.579 msec ( 1.579 nsec per iteration ) last: 0002263039ed1a3f
pdep: 1.556 msec ( 1.556 nsec per iteration ) last: a589e3daa1ddc7bd
pexts (warmup) 5.930 msec ( 5.930 nsec per iteration ) last: 0002263039ed1a3f
pexts: 2.139 msec ( 2.139 nsec per iteration ) last: 0002263039ed1a3f
pdeps: 1.976 msec ( 1.976 nsec per iteration ) last: a589e3daa1ddc7bd
pexts_1: 1.791 msec ( 1.791 nsec per iteration ) last: 0000000e53d5f2c7
pdeps_1: 1.698 msec ( 1.698 nsec per iteration ) last: 002b43089394d373
best impl: 1

I found that a warmup run on the large array was necessary.
perhaps some other way of pre-faulting could be done instead.

Typical output, i5-1235U, performance core ``` estimated overhead 1.195 msec ( 1.195 nsec per iteration ) last: d423b7ad3868912d pext: 1.579 msec ( 1.579 nsec per iteration ) last: 0002263039ed1a3f pdep: 1.556 msec ( 1.556 nsec per iteration ) last: a589e3daa1ddc7bd pexts (warmup) 5.930 msec ( 5.930 nsec per iteration ) last: 0002263039ed1a3f pexts: 2.139 msec ( 2.139 nsec per iteration ) last: 0002263039ed1a3f pdeps: 1.976 msec ( 1.976 nsec per iteration ) last: a589e3daa1ddc7bd pexts_1: 1.791 msec ( 1.791 nsec per iteration ) last: 0000000e53d5f2c7 pdeps_1: 1.698 msec ( 1.698 nsec per iteration ) last: 002b43089394d373 best impl: 1 ``` I found that a warmup run on the large array was necessary. perhaps some other way of pre-faulting could be done instead.

jjl commented 2026年03月28日 10:56:38 +01:00

I found that a warmup run on the large array was necessary. perhaps some other way of pre-faulting could be done instead.

lol, did you just make a million entry array of test data? that's going to be more problematic than a small one i would have thought.

> I found that a warmup run on the large array was necessary. perhaps some other way of pre-faulting could be done instead. lol, did you just make a million entry array of test data? that's going to be more problematic than a small one i would have thought.

jepler commented 2026年03月31日 16:53:51 +02:00

Yeah it's shameful. I can rework it to be less shameful if you want, or I can close it up.

Yeah it's shameful. I can rework it to be less shameful if you want, or I can close it up.

jjl commented 2026年03月31日 18:41:03 +02:00

ah fuck it, these are only really for completeness anyway.

ah fuck it, these are only really for completeness anyway.