omar

I ran some indicative general Dear ImGui perf tests. - There aren't designed as sprintf tests, more like testing average use of dear imgui. "perf_demo_all" is opening all of dear...

Thanks for looking! If at a glance it looks like something you think I'll be able to build/run I don't mind trying. (`omar AT dearimgui dot org`)

I ran the tests sent by Sean (they more or less run through all possible values for many types). Here's one of the run: (`stbsp_sprintf ` = vanilla, `stbsp_sprintf2` my...

> The way you swapped the goto-Labels and the two checks in the first loop, we won't check for end-of-string anymore, if the "if (((v ^ 0x25252525) - 0x01010101) &...

> Kind of depressing since it means strlen() on whichever platform you're using isn't already doing four-at-a-time (or even SIMD). VS2015 and VS2019 "Release" build are inlining a dumb version...

I duplicated the copying loop (~42 lines) to provide both a zero-terminated and a pointer-terminated version. `fmt_end` gets set to ~0 when using a zero-terminated function. The flag parsing code...

> as is having our own strlen implementation. That part got actually removed in v5, since the zero-terminating case doesn't need it anymore. (fyi stb_printf already had `stbsp__strlen_limited()`, the one...

> I'd recommend only implementing versions of nz functions that take an explicit output length, but moreover, they probably should write nz output. So we probably need a way to...

Current behavior seems fine, and I recall the thread arguing for matching standard (I probably voted for that). IMHO the only value/purpose of adding nz output support would to save...

Will look into that (may do a separate PR to be able to get that one out of the way soon).