__gnu_parallel::_Settings(3cxx)

🌸 NAME 🌸

Welcome to the __gnu_parallel::_Settings(3cxx) page, bestie — part 1 of 2, and it's giving "control room for parallel mode."

__gnu_parallel::_Settings — the run-time settings class for parallel mode, packing all the tunable parameters into one iconic little vibe console.

---

✨ SYNOPSIS ✨

#include <settings.h>

🎀 Static Public Member Functions 🎀

• static const _Settings & get () throw () — get the global settings, no cap.
• static void set (_Settings &) throw () — set the global settings. Periodt.

💖 Public Attributes 💖

• _SequenceIndex accumulate_minimal_n — minimal input size for accumulate.
• unsigned int adjacent_difference_minimal_n — minimal input size for adjacent_difference.
• _AlgorithmStrategy algorithm_strategy — the manpage gives this one literally just the name, bestie.
• unsigned int cache_line_size — an overestimation of the cache line size, used to dodge false sharing — i.e. elements from different threads are at least this amount apart.
• _SequenceIndex count_minimal_n — minimal input size for count and count_if.
• _SequenceIndex fill_minimal_n — minimal input size for fill.
• _FindAlgorithm find_algorithm — again, no extra blurb on this page; just the name.
• double find_increasing_factor — block size increase factor for find.
• _SequenceIndex find_initial_block_size — initial block size for find.
• _SequenceIndex find_maximum_block_size — maximal block size for find.
• float find_scale_factor — block size scale-down factor with respect to current position.
• _SequenceIndex find_sequential_search_size — start by looking for this many elements sequentially, for find.
• _SequenceIndex for_each_minimal_n — minimal input size for for_each.
• _SequenceIndex generate_minimal_n — minimal input size for generate.
• unsigned long long L1_cache_size — size of the L1 cache in bytes (underestimation).
• unsigned long long L2_cache_size — size of the L2 cache in bytes (underestimation).
• _SequenceIndex max_element_minimal_n — minimal input size for max_element.
• _SequenceIndex merge_minimal_n — minimal input size for merge.
• unsigned int merge_oversampling — oversampling factor for merge.
• _SplittingAlgorithm merge_splitting — no description here, just the name.
• _SequenceIndex min_element_minimal_n — minimal input size for min_element.
• _MultiwayMergeAlgorithm multiway_merge_algorithm — name only, bestie.
• int multiway_merge_minimal_k — oversampling factor for multiway_merge.
• _SequenceIndex multiway_merge_minimal_n — minimal input size for multiway_merge.
• unsigned int multiway_merge_oversampling — oversampling factor for multiway_merge.
• _SplittingAlgorithm multiway_merge_splitting — just the name on this one.
• _SequenceIndex nth_element_minimal_n — minimal input size for nth_element.
• _SequenceIndex partial_sort_minimal_n — minimal input size for partial_sort.
• _PartialSumAlgorithm partial_sum_algorithm — name only, no extra tea.
• float partial_sum_dilation — ratio for partial_sum. Assume 'sum and write result' to be this factor slower than just 'sum'.
• unsigned int partial_sum_minimal_n — minimal input size for partial_sum.
• double partition_chunk_share — chunk size for partition, relative to input size. If > 0.0, this value overrides partition_chunk_size.
• _SequenceIndex partition_chunk_size — chunk size for partition.
• _SequenceIndex partition_minimal_n — minimal input size for partition.
• _SequenceIndex qsb_steals — the number of stolen ranges in load-balanced quicksort.
• unsigned int random_shuffle_minimal_n — minimal input size for random_shuffle.
• _SequenceIndex replace_minimal_n — minimal input size for replace and replace_if.
• _SequenceIndex search_minimal_n — minimal input size for search and search_n.
• _SequenceIndex set_difference_minimal_n — minimal input size for set_difference.
• _SequenceIndex set_intersection_minimal_n — minimal input size for set_intersection.
• _SequenceIndex set_symmetric_difference_minimal_n — minimal input size for set_symmetric_difference.
• _SequenceIndex set_union_minimal_n — minimal input size for set_union.
• _SortAlgorithm sort_algorithm — name only, iykyk.
• _SequenceIndex sort_minimal_n — minimal input size for parallel sorting.
• unsigned int sort_mwms_oversampling — oversampling factor for parallel std::sort (MWMS).
• unsigned int sort_qs_num_samples_preset — such many samples to take to find a good pivot (quicksort).
• _SequenceIndex sort_qsb_base_case_maximal_n — maximal subsequence __length to switch to unbalanced __base case. Applies to std::sort with dynamically load-balanced quicksort.
• _SplittingAlgorithm sort_splitting — just the name, bestie.
• unsigned int TLB_size — size of the Translation Lookaside Buffer (underestimation).
• _SequenceIndex transform_minimal_n — minimal input size for parallel std::transform.
• _SequenceIndex unique_copy_minimal_n — minimal input size for unique_copy.
• _SequenceIndex workstealing_chunk_size — no description on this page, just the name.

---

💖 Detailed Description 💖

So _Settings is the class for run-time settings for the parallel mode, and it includes all the tunable parameters. All of them. Every single knob. It's the main character of "where do I change that value," fr fr.

---

🫶 Member Function Documentation 🫶

🌟 static const _Settings & __gnu_parallel::_Settings::get () [static] 🌟

static const _Settings & get () throw () — get the global settings. That's the whole job, and it ate.

The page says this is referenced by:

• __gnu_parallel::__find_template()
• __gnu_parallel::__find_template()
• __gnu_parallel::__find_template()
• __gnu_parallel::__for_each_template_random_access_workstealing()
• __gnu_parallel::__parallel_nth_element()
• __gnu_parallel::__parallel_partial_sum()
• __gnu_parallel::__parallel_partial_sum_linear()
• __gnu_parallel::__parallel_partition()
• __gnu_parallel::__parallel_random_shuffle_drs()
• __gnu_parallel::__parallel_sort()
• __gnu_parallel::__parallel_sort()
• __gnu_parallel::__parallel_sort_qs_conquer()
• __gnu_parallel::__qsb_local_sort_with_helping()
• __gnu_parallel::__sequential_random_shuffle()
• __gnu_parallel::multiway_merge_sampling_splitting()
• __gnu_parallel::parallel_multiway_merge()
• __gnu_parallel::parallel_sort_mwms()
• __gnu_parallel::parallel_sort_mwms_pu()

🌈 static void __gnu_parallel::_Settings::set (_Settings &) [static] 🌈

static void set (_Settings &) throw () — set the global settings. Simple, clean, iconic.

✨ Member Data Documentation ✨

Okay bestie, this is the good part — all the actual knobs and dials on __gnu_parallel::_Settings. Every single one, no skips, we're doing the full tour 💅

---

• _SequenceIndex __gnu_parallel::_Settings::accumulate_minimal_n — the minimal input size for accumulate. Below this, it's just not worth the parallel drama 🎀
• unsigned int __gnu_parallel::_Settings::adjacent_difference_minimal_n — minimal input size for adjacent_difference.
• unsigned int __gnu_parallel::_Settings::cache_line_size — an overestimation of the cache line size. It's used to dodge false sharing, i.e. making sure elements belonging to different threads sit at least this far apart 🦋
  • Referenced by __gnu_parallel::__for_each_template_random_access_workstealing().
• _SequenceIndex __gnu_parallel::_Settings::count_minimal_n — minimal input size for count and count_if.
• _SequenceIndex __gnu_parallel::_Settings::fill_minimal_n — minimal input size for fill.
• double __gnu_parallel::_Settings::find_increasing_factor — the block size increase factor for find.
• _SequenceIndex __gnu_parallel::_Settings::find_initial_block_size — the initial block size for find.
  • Referenced by __gnu_parallel::__find_template().
• _SequenceIndex __gnu_parallel::_Settings::find_maximum_block_size — the maximal block size for find.
• float __gnu_parallel::_Settings::find_scale_factor — block size scale-down factor with respect to the current position.
  • Referenced by __gnu_parallel::__find_template().
• _SequenceIndex __gnu_parallel::_Settings::find_sequential_search_size — start off by looking for this many elements sequentially, for find.
  • Referenced by __gnu_parallel::__find_template(), and __gnu_parallel::__find_template().
• _SequenceIndex __gnu_parallel::_Settings::for_each_minimal_n — minimal input size for for_each.
• _SequenceIndex __gnu_parallel::_Settings::generate_minimal_n — minimal input size for generate.
• unsigned long long __gnu_parallel::_Settings::L1_cache_size — size of the L1 cache in bytes (underestimation, so it lowballs it on purpose 🙈).
• unsigned long long __gnu_parallel::_Settings::L2_cache_size — size of the L2 cache in bytes (underestimation, same deal).
  • Referenced by __gnu_parallel::__parallel_random_shuffle_drs(), and __gnu_parallel::__sequential_random_shuffle().
• _SequenceIndex __gnu_parallel::_Settings::max_element_minimal_n — minimal input size for max_element.
• _SequenceIndex __gnu_parallel::_Settings::merge_minimal_n — minimal input size for merge.
• unsigned int __gnu_parallel::_Settings::merge_oversampling — oversampling factor for merge.
  • Referenced by __gnu_parallel::multiway_merge_sampling_splitting(), and __gnu_parallel::parallel_multiway_merge().
• _SequenceIndex __gnu_parallel::_Settings::min_element_minimal_n — minimal input size for min_element.
• int __gnu_parallel::_Settings::multiway_merge_minimal_k — oversampling factor for multiway_merge.
• _SequenceIndex __gnu_parallel::_Settings::multiway_merge_minimal_n — minimal input size for multiway_merge.
• unsigned int __gnu_parallel::_Settings::multiway_merge_oversampling — oversampling factor for multiway_merge.
• _SequenceIndex __gnu_parallel::_Settings::nth_element_minimal_n — minimal input size for nth_element.
  • Referenced by __gnu_parallel::__parallel_nth_element().
• _SequenceIndex __gnu_parallel::_Settings::partial_sort_minimal_n — minimal input size for partial_sort.
• float __gnu_parallel::_Settings::partial_sum_dilation — ratio for partial_sum. Basically it assumes that "sum and write result" is this factor slower than just plain "sum" 🍓
  • Referenced by __gnu_parallel::__parallel_partial_sum_linear().
• unsigned int __gnu_parallel::_Settings::partial_sum_minimal_n — minimal input size for partial_sum.
• double __gnu_parallel::_Settings::partition_chunk_share — chunk size for partition, relative to input size. Heads up: if it's > 0.0, this value overrides partition_chunk_size 👑
  • Referenced by __gnu_parallel::__parallel_partition().
• _SequenceIndex __gnu_parallel::_Settings::partition_chunk_size — chunk size for partition.
  • Referenced by __gnu_parallel::__parallel_partition().
• _SequenceIndex __gnu_parallel::_Settings::partition_minimal_n — minimal input size for partition.
  • Referenced by __gnu_parallel::__parallel_nth_element().
• _SequenceIndex __gnu_parallel::_Settings::qsb_steals — the number of stolen ranges in load-balanced quicksort. stealing is literally the whole personality of this algorithm ngl 💀
• unsigned int __gnu_parallel::_Settings::random_shuffle_minimal_n — minimal input size for random_shuffle.
• _SequenceIndex __gnu_parallel::_Settings::replace_minimal_n — minimal input size for replace and replace_if.
• _SequenceIndex __gnu_parallel::_Settings::search_minimal_n — minimal input size for search and search_n.
• _SequenceIndex __gnu_parallel::_Settings::set_difference_minimal_n — minimal input size for set_difference.
• _SequenceIndex __gnu_parallel::_Settings::set_intersection_minimal_n — minimal input size for set_intersection.
• _SequenceIndex __gnu_parallel::_Settings::set_symmetric_difference_minimal_n — minimal input size for set_symmetric_difference.
• _SequenceIndex __gnu_parallel::_Settings::set_union_minimal_n — minimal input size for set_union.
• _SequenceIndex __gnu_parallel::_Settings::sort_minimal_n — minimal input size for parallel sorting.
• unsigned int __gnu_parallel::_Settings::sort_mwms_oversampling — oversampling factor for parallel std::sort (MWMS).
  • Referenced by __gnu_parallel::parallel_sort_mwms(), and __gnu_parallel::parallel_sort_mwms_pu().
• unsigned int __gnu_parallel::_Settings::sort_qs_num_samples_preset — this many samples to take to find a good pivot (quicksort) 🎯
• _SequenceIndex __gnu_parallel::_Settings::sort_qsb_base_case_maximal_n — maximal subsequence __length to switch to the unbalanced __base case. Applies to std::sort with dynamically load-balanced quicksort.
  • Referenced by __gnu_parallel::__qsb_local_sort_with_helping().
• unsigned int __gnu_parallel::_Settings::TLB_size — size of the Translation Lookaside Buffer (underestimation).
  • Referenced by __gnu_parallel::__parallel_random_shuffle_drs(), and __gnu_parallel::__sequential_random_shuffle().
• _SequenceIndex __gnu_parallel::_Settings::transform_minimal_n — minimal input size for parallel std::transform.
• _SequenceIndex __gnu_parallel::_Settings::unique_copy_minimal_n — minimal input size for unique_copy.

---

🌸 Author 🌸

Generated automatically by Doxygen for libstdc++ from the source code. no cap, nobody hand-wrote this one, the machine did it 💖

---

okay that's the whole page bestie, you ate that 🫶✨ This lil doc shipped as part of libstdc++, in the __gnu_parallel::_Settings(3cxx) manual page (3cxx section) — go be iconic with your settings, mwah 💅👑