3 Commits

Author SHA1 Message Date
Nikolas Klauser
8fc2538f33
Reapply "[libc++] Optimize num_put integral functions" (#131613) (#133572)
This reverts commit d1156fcb56891fb1a426c3e8331a51d47f98a1b8.

This patch fixes the reported incorrect formatting changes and adds
tests for them. The performance should be unaffected, since there are no
significant changes required to fix the bugs.

Specifically, a `>` was changed to a `>=` to also add a `+` in the zero
case, and we're checking for zero now before printing the octal and
hexadecimal prefixes.

Closes #131710
2025-04-11 15:35:58 +02:00
Alexander Kornienko
d1156fcb56
Revert "[libc++] Optimize num_put integral functions" (#131613)
Reverts llvm/llvm-project#120859

This change breaks formatting of `0` with `std::showbase` + `std::hex`
or `std::oct`, as well as `+0` with `std::showpos`. I believe the new
behavior is violating the standard. See
https://github.com/llvm/llvm-project/pull/120859#issuecomment-2723970242
and later comments for details and explanation.
2025-03-18 02:05:51 +01:00
Nikolas Klauser
15edf8725a
[libc++] Optimize num_put integral functions (#120859)
```
-------------------------------------------------------
Benchmark                              old          new
-------------------------------------------------------
BM_num_put<bool>                   76.2 ns      32.0 ns
BM_num_put<long>                   76.9 ns      33.1 ns
BM_num_put<long long>              77.9 ns      34.2 ns
BM_num_put<unsigned long>          78.4 ns      33.1 ns
BM_num_put<unsigned long long>     78.0 ns      34.4 ns
BM_num_put<double>                  224 ns       228 ns
BM_num_put<long double>             239 ns       230 ns
BM_num_put<const void*>            68.7 ns      35.1 ns
```

Fixes #40109.
2025-03-05 14:18:49 +01:00