Sahnvour

Zig automatically inserts indirect constraint for outputs that are not annotated as return (`->`), I don't know why nor if it was required at some point. But before that, your...

Just saw Andrew's tweets, good work everyone. In case that's of any interest, I had some hashmap benchmarks at https://github.com/Sahnvour/zig-containers/blob/master/benchmarks/martinus_map.zig that tried to reproduce the ones mentionned by @squeek502 at...

@lemmi In my experience, going for a similar design to google's hashtable (8bit meta data per element) with scalar code was faster. I think the SIMD is beneficial to them...

I'm adding some benchmarks that can take up to a few tens of seconds. Should we make them as atomic as possible, or is it ok to have them all...

> I'm not very familiar with Wyhash and how well it avoids the DoS issues that SipHash attempts to avoid. Looks like the author there ran into the same issue...

Again, if you're willing to put up a PR to fix these issues, don't hesitate. Just be sure that they don't break anything 🙂

You're right there isn't a compelling reason for using singletons. I wouldn't write it this way nowadays :) I'm not actively maintaining this code anymore but feel free to propose...

Yes, see for example https://github.com/Sahnvour/PathFinder/blob/master/examples/image.cpp#L138 . In these loops, for each pixel we look at its (up to) 9 neighbours and add them as children, ie. accessible nodes. If you...

You need to deallocate every array stored in `squares[x]` and `squares` itself. See https://github.com/Sahnvour/PathFinder/blob/03bac25118aaa65822d036ac9f3b82124b681ad9/examples/image.cpp#L171-L173 If your memory usage increases every time, this also means that you are allocating and constructing...

You're missing angle brackets in the error messages (`std::vectorTAlgorithm::node_type*`), is that a formatting error ?