Emilio Cobos Álvarez
bb460ea208
Hashing char by char is unnecessarily expensive, this hashes four bytes at a time for those. I did that rather than keeping size_t because HashBytes(uint64_t) ends up in two AddUint32ToHash() anyways, and that way we avoid the python re-implementation from depending on the word size of the target. For wide strings, keep the current char-by-char version for now, but reduce the amount of times this algorithm is duplicated and we rely on it matching: * HashUTF8AsUTF16() * JS atoms * xpcom atoms. Use constexpr to simplify the static atoms set up. Differential Revision: https://phabricator.services.mozilla.com/D306312
This is DMD. See https://firefox-source-docs.mozilla.org/performance/memory/dmd.html for details on how to use it.