Faster IT

I think for consistency reasons this switch should also affect the bar mode. I'm not against it, it's small and fits a use case of seeing how much memory is...

Can we have one "format times" code as in #1583 and this combined?

https://en.wikipedia.org/wiki/Microsecond ... "Its symbol is μs, sometimes simplified to us when Unicode is not available." No need to ignore the unit. It's 2025 and we have `#ifdef HAVE_LIBNCURSESW` :tada: .

Fails CI: ``` XUtils.c:426:6: error: 'HAVE_STRNLEN' is not defined, evaluates to 0 [-Werror,-Wundef] 426 | # if HAVE_STRNLEN | ^ 1 error generated.``

Building a htop with your patch ``` * 58524a3b - (HEAD -> main) Merge branch 'meter-human-time-unit' of htop-1 (13 minutes ago) [Daniel Lange] |\ | * 602d4f40 - Rework humanTimeUnit()...

That only shows me a usage "%". But in any case why is `humanTimeUnit()` not even used consistently throughout the GPU meter where it was conceived? I think you need...

What changed between `htop` did not do this and now `htop` does? Which version of `htop` do you run? Did you try to compile the latest commit from the repo?

@BenBE: Nope, you got that well. @jb-boin's comment only applies to his patch shortening strings as per the last line of your first comment above.

With your forkbomb script running: ``` Samples: 150K of event 'cycles', 4000 Hz, Event count (approx.): 26606134128 lost: 0/0 drop: 0/0 Overhead Shared Object Symbol 15,90% htop [.] Hashtable_isConsistent 7,40%...

Thank you @natoscott. This is what a non-debug build gets: ``` Samples: 101K of event 'cycles', 4000 Hz, Event count (approx.): 18467791692 lost: 0/0 drop: 0/0 Overhead Shared Object Symbol...