Closed joaquintides closed 1 year ago
An automated preview of the documentation is available at https://187.unordered.prtest.cppalliance.org/libs/unordered/doc/html/unordered.html
Looks good to me, although I'm not sure we should be merging to develop now. Maybe wait for the beta to go out in case there are critical fixes we need to apply.
The release notes added in the PR target 1.83, I just submitted it because it's ready but we can wait of course.
An automated preview of the documentation is available at https://187.unordered.prtest.cppalliance.org/libs/unordered/doc/html/unordered.html
An automated preview of the documentation is available at https://187.unordered.prtest.cppalliance.org/libs/unordered/doc/html/unordered.html
Merging #187 (1c5640c) into develop (9aedb95) will increase coverage by
0.00%
. The diff coverage is100.00%
.
Latest commit (1c5640cfbe84213b9d2a7eff68e2db744cfa3817) didn't affect performance: https://github.com/boostorg/boost_unordered_benchmarks/commit/90e178376ffefe73d3e6fc7fa24ef70a26d429a6
Latest commit (1c5640c) didn't affect performance: boostorg/boost_unordered_benchmarks@90e1783
Well, my logic was more about the code self-documenting the assumptions.
It wasn't clear to me why we needed the null pointer check before calling ++it
.
@pdimov @joaquintides
Now that the Boost beta is out, should we feel free to go and merge this PR?
FWIW, this feature is already integrated in feature/cfoa, so we can either drop this PR or merge and rebase that branch.
We can merge it to develop, I suppose, just not to master.
Changed iterator increment code for "regular layout" group implementations (basically, those for SSE2 and Neon). Iteration speed improves dramatically, look for "Running erasure" in:
https://github.com/boostorg/boost_unordered_benchmarks/commit/46129697b05fab67c42cfa2b918a6e40af26d6e6