llvm-project

History

Connector Switch 2d106844e7

[libcxx] Optimize ranges::fold_left_with_iter for segmented iterators (#177853 )

Part of https://github.com/llvm/llvm-project/issues/102817.

This patch attempts to optimize the performance of
`ranges::fold_left_with_iter` for segmented iterators.

- before

```
# | rng::fold_left(vector<int>)/8             2.78 ns         2.78 ns    241953718
# | rng::fold_left(vector<int>)/32            12.2 ns         12.2 ns     57579851
# | rng::fold_left(vector<int>)/50            19.2 ns         19.2 ns     36487764
# | rng::fold_left(vector<int>)/8192          3226 ns         3226 ns       216811
# | rng::fold_left(vector<int>)/1048576     441842 ns       441839 ns         1592
# | rng::fold_left(deque<int>)/8              2.83 ns         2.83 ns    243888678
# | rng::fold_left(deque<int>)/32             16.6 ns         16.6 ns     42297458
# | rng::fold_left(deque<int>)/50             22.3 ns         22.3 ns     31387998
# | rng::fold_left(deque<int>)/8192           2492 ns         2492 ns       281637
# | rng::fold_left(deque<int>)/1048576      324936 ns       324936 ns         2154
# | rng::fold_left(list<int>)/8               2.54 ns         2.54 ns    275946635
# | rng::fold_left(list<int>)/32              16.2 ns         16.2 ns     42901634
# | rng::fold_left(list<int>)/50              54.7 ns         54.7 ns     12767450
# | rng::fold_left(list<int>)/8192           15154 ns        15154 ns        56744
# | rng::fold_left(list<int>)/1048576      4976906 ns      4976867 ns          158
```

- after

```
# | rng::fold_left(vector<int>)/8             2.74 ns         2.74 ns    255954900
# | rng::fold_left(vector<int>)/32            12.1 ns         12.1 ns     57843462
# | rng::fold_left(vector<int>)/50            19.2 ns         19.2 ns     36422594
# | rng::fold_left(vector<int>)/8192          3202 ns         3202 ns       218265
# | rng::fold_left(vector<int>)/1048576     435718 ns       435709 ns         1609
# | rng::fold_left(deque<int>)/8              2.52 ns         2.52 ns    277288254
# | rng::fold_left(deque<int>)/32             14.1 ns         14.1 ns     52244463
# | rng::fold_left(deque<int>)/50             16.2 ns         16.2 ns     43131857
# | rng::fold_left(deque<int>)/8192           1695 ns         1695 ns       415620
# | rng::fold_left(deque<int>)/1048576      277729 ns       277731 ns         2532
# | rng::fold_left(list<int>)/8               2.55 ns         2.55 ns    277025050
# | rng::fold_left(list<int>)/32              16.2 ns         16.2 ns     43058857
# | rng::fold_left(list<int>)/50              54.7 ns         54.7 ns     12705516
# | rng::fold_left(list<int>)/8192           15236 ns        15235 ns        56840
# | rng::fold_left(list<int>)/1048576      4827263 ns      4827147 ns          152
```

2026-02-05 21:12:36 +08:00

alg.c.library

…

alg.modifying.operations

[libc++][ranges] implement ranges::shift_left (#83231 )

2026-01-25 10:35:43 +00:00

alg.nonmodifying

[libcxx] Optimize ranges::fold_left_with_iter for segmented iterators (#177853 )

2026-02-05 21:12:36 +08:00

alg.sorting

[libc++][pstl] Generic implementation of parallel std::is_sorted (#176129 )

2026-01-27 13:43:23 -05:00