Carsten Schlote

@CyberShadow XXHash seems to be slightly faster than Murmur. At least that is stated on https://github.com/Cyan4973/xxHash/README.md. There is also a SSE2 variant, which is much faster. So it won't hurt...

"slight faster" means 2.5 to 5 times faster than Murmur. XXHash is included in ZSTD, which itself gains popularity for being fast and efficient. We replaced MD5 with XXHash at...

Yes, I know, and the existing Dub xxhash package implements the old old 32 bit algorithm. Meanwhile, there are two new 64 bit modes variants, and a 128bit variant. The...

@ljmf00 Thanks for your thoughts. We had similiar discussions and considerations at work. XXHash is amazingly fast, but still reliable enough to checksum data. For stuff like toHash(), I think...

@ljmf00 I think, it shouldn't be too hard to create a D port of the C sources. With all the unittests already implemented and checksums compared against the reference C...

So porting the routines to D seems to be the best option, as it solves the problems with embedding C sources and get it working on all platforms. On the...

Latest problem with XXH3 variants: They needs 128bit arthmetics. With current C compilers this is no problem, as they already have 128 bit code available and builtin. In D this...

Ok, ported the C code to D. And Unittests still pass ;-) The result is still pretty much C-ish source code, which needs further cleanups and refactoring to match D...

First, thanks for the review and comments. It is always fun to see, that people do all that kind of work in their spare time. I would really like to...

The next steps are cleanups on documentation and code style. Maybe/Probably also some enhancements to unitests. After that, the PR should be ready to leave draft state.