Stephan T. Lavavej comments

Results 354 comments of


                                            Stephan T. Lavavej

trafficstars

Vectorize `std::search` of 1 and 2 bytes elements with `pcmpestri`

:warning: Note to self, check the benchmarks on my machine.

Vectorize `std::search` of 1 and 2 bytes elements with `pcmpestri`

This PR isn't too massive, I think it would make sense to add large needle logic here.

Vectorize `std::search` of 1 and 2 bytes elements with `pcmpestri`

/azp run

Vectorize `std::search` of 1 and 2 bytes elements with `pcmpestri`

I'm reviewing this and will push changes when I'm ready, including simple merge conflict resolutions.

Vectorize `std::search` of 1 and 2 bytes elements with `pcmpestri`

Benchmark results on my 5950X, split into separate tables for 1 and 2 bytes versus 4 and 8 bytes: Benchmark | `main` | PR | Speedup (Old/New) -------------------------------------------|---------|---------|------------------ `c_strstr/0` |...

Vectorize `std::search` of 1 and 2 bytes elements with `pcmpestri`

Thanks, makes sense!

Vectorize `std::search` of 1 and 2 bytes elements with `pcmpestri`

I'm mirroring this to the MSVC-internal repo - please notify me if any further changes are pushed.

Vectorize `std::search` of 1 and 2 bytes elements with `pcmpestri`

:mag: :detective: :mag_right: ===

`<stacktrace>`: investigate sporadic failure of `CaptureStackBackTrace`

OS-37394205 "RtlCaptureStackBackTrace returns empty back trace" was wontfixed. Perhaps we need (ugh) retry logic, if it simply returns 0 with a low failure rate.

`<stacktrace>`: investigate sporadic failure of `CaptureStackBackTrace`

Yeah, that's the Windows-internal bug database, apologies for not mentioning that. The bug report essentially described what we've observed; there was no further discussion when it was wontfixed :crying_cat_face: