4 Commits

Author	SHA1	Message	Date
Nikolas Klauser	07a0e0fb31	[libc++] Remove benchmarks for ranges algorithms that have a std equivalent (#176138) ... We're currently running all the algorithms benchmarks for the `std` and `ranges` variants, even though almost all the algorithms share the same code. This makes running the benchmarks on a large set of commits very slow and costly. This reduced running the `algorithm/` subdirectory from ~4 hours to roughtly 2.5 hours. Fixes #175973	2026-01-19 16:04:32 +01:00
Nikolas Klauser	4fc5b6d8c4	[libc++] Optimize {std,ranges}::for_each for iterating over __trees (#164405) ... This patch optimizes how `for_each` iterates over trees by using recursion and storing pointers to the next nodes on the stack. This avoids pointer chasing through the `__parent_` pointer, reducing cache misses. It also makes use of the compiler being able tail-call optimize the recursive function, removing back-tracking the iterators have to do. ``` Benchmark old new Difference % Difference -------------------------------------------------- -------------- -------------- ------------ -------------- rng::for_each(map<int>)/32 35.19 26.67 -8.52 -24.21% rng::for_each(map<int>)/50 64.13 40.68 -23.45 -36.57% rng::for_each(map<int>)/8 5.06 6.49 1.43 28.21% rng::for_each(map<int>)/8192 22893.89 9266.68 -13627.21 -59.52% rng::for_each(map<int>::iterator)/32 35.51 26.88 -8.63 -24.31% rng::for_each(map<int>::iterator)/50 64.39 41.24 -23.15 -35.95% rng::for_each(map<int>::iterator)/8 5.12 5.93 0.81 15.80% rng::for_each(map<int>::iterator)/8192 21283.14 9736.83 -11546.31 -54.25% rng::for_each(multimap<int>)/32 35.22 26.61 -8.61 -24.45% rng::for_each(multimap<int>)/50 64.10 40.07 -24.03 -37.49% rng::for_each(multimap<int>)/8 5.08 6.69 1.61 31.70% rng::for_each(multimap<int>)/8192 23130.44 9026.16 -14104.28 -60.98% rng::for_each(multimap<int>::iterator)/32 35.40 25.08 -10.32 -29.15% rng::for_each(multimap<int>::iterator)/50 64.19 38.15 -26.04 -40.56% rng::for_each(multimap<int>::iterator)/8 5.04 5.25 0.22 4.31% rng::for_each(multimap<int>::iterator)/8192 22875.97 9392.08 -13483.89 -58.94% rng::for_each(multiset<int>)/32 35.82 27.11 -8.72 -24.33% rng::for_each(multiset<int>)/50 62.92 41.59 -21.32 -33.89% rng::for_each(multiset<int>)/8 4.79 6.79 2.00 41.70% rng::for_each(multiset<int>)/8192 22642.68 9280.95 -13361.73 -59.01% rng::for_each(multiset<int>::iterator)/32 35.76 26.71 -9.04 -25.28% rng::for_each(multiset<int>::iterator)/50 63.44 39.00 -24.44 -38.53% rng::for_each(multiset<int>::iterator)/8 4.90 5.21 0.30 6.18% rng::for_each(multiset<int>::iterator)/8192 19930.45 9867.60 -10062.85 -50.49% rng::for_each(set<int>)/32 35.90 27.30 -8.60 -23.96% rng::for_each(set<int>)/50 63.15 40.75 -22.40 -35.47% rng::for_each(set<int>)/8 4.77 6.83 2.06 43.23% rng::for_each(set<int>)/8192 20262.77 9381.57 -10881.20 -53.70% rng::for_each(set<int>::iterator)/32 36.02 26.42 -9.60 -26.64% rng::for_each(set<int>::iterator)/50 63.29 37.97 -25.32 -40.01% rng::for_each(set<int>::iterator)/8 4.72 5.22 0.50 10.50% rng::for_each(set<int>::iterator)/8192 20041.91 9831.91 -10210.00 -50.94% ```	2025-12-12 09:05:51 +01:00
Peng Liu	9827440f1e	[libc++] Optimize ranges::{for_each, for_each_n} for segmented iterators (#132896) ... Previously, the segmented iterator optimization was limited to `std::{for_each, for_each_n}`. This patch extends the optimization to `std::ranges::for_each` and `std::ranges::for_each_n`, ensuring consistent optimizations across these algorithms. This patch first generalizes the `std` algorithms by introducing a `Projection` parameter, which is set to `__identity` for the `std` algorithms. Then we let the `ranges` algorithms to directly call their `std` counterparts with a general `__proj` argument. Benchmarks demonstrate performance improvements of up to 21.4x for ``std::deque::iterator`` and 22.3x for ``join_view`` of ``vector<vector<char>>``. Addresses a subtask of #102817.	2025-06-18 12:22:47 -04:00
Louis Dionne	b3a4bf9d8f	[libc++] Refactor and add benchmarks from [alg.nonmodifying] (#128206)	2025-03-19 00:08:46 -04:00