[libcxx-commits] [PATCH] D101752: Speedup to_string for integers using zero-copy.

[Roman Koshelev via Phabricator via libcxx-commits]

Roman-Koshelev added a comment.

Benchmarks

$ g++ -v

  gcc version 10.3.0 (Ubuntu 10.3.0-1ubuntu1)

$ g++ main.cpp -O3 -lbenchmark -std=c++20 
$ ./a.out

  2021-05-07T07:54:15+03:00
  Running ./a.out
  Run on (4 X 3000 MHz CPU s)
  CPU Caches:
    L1 Data 32 KiB (x4)
    L1 Instruction 32 KiB (x4)
    L2 Unified 256 KiB (x4)
    L3 Unified 12288 KiB (x4)
  Load Average: 0.43, 0.45, 0.41
  ---------------------------------------------------------------
  Benchmark                     Time             CPU   Iterations
  ---------------------------------------------------------------
  to_stringNewBench/1         693 ns          692 ns       971467
  to_stringNewBench/2         933 ns          932 ns       749060
  to_stringNewBench/3        2008 ns         2006 ns       342603
  to_stringNewBench/4        2034 ns         2032 ns       340914
  to_stringNewBench/5        1992 ns         1990 ns       347027
  to_stringNewBench/6        1909 ns         1908 ns       345394
  to_stringNewBench/7        1947 ns         1945 ns       346533
  to_stringNewBench/8        1993 ns         1991 ns       356201
  to_stringNewBench/9        1913 ns         1911 ns       362352
  to_stringNewBench/10       1958 ns         1956 ns       358725
  to_stringNewBench/11       1965 ns         1963 ns       366209
  to_stringNewBench/12       1971 ns         1969 ns       354935
  to_stringNewBench/13       1927 ns         1925 ns       349851
  to_stringNewBench/14       2013 ns         2011 ns       358991
  to_stringNewBench/15       2021 ns         2019 ns       356885
  to_stringNewBench/16       2011 ns         2009 ns       350729
  to_stringOldBench/1        1038 ns         1037 ns       694926
  to_stringOldBench/2        1168 ns         1167 ns       590311
  to_stringOldBench/3        1601 ns         1600 ns       438899
  to_stringOldBench/4        1560 ns         1559 ns       442588
  to_stringOldBench/5        4362 ns         4359 ns       161358
  to_stringOldBench/6        4306 ns         4303 ns       162674
  to_stringOldBench/7        4375 ns         4372 ns       158870
  to_stringOldBench/8        4370 ns         4367 ns       166312
  to_stringOldBench/9        4334 ns         4331 ns       160176
  to_stringOldBench/10       4402 ns         4399 ns       157898
  to_stringOldBench/11       4401 ns         4397 ns       158140
  to_stringOldBench/12       4330 ns         4327 ns       160746
  to_stringOldBench/13       4388 ns         4385 ns       159924
  to_stringOldBench/14       4302 ns         4299 ns       158747
  to_stringOldBench/15       4358 ns         4354 ns       158021
  to_stringOldBench/16       4322 ns         4319 ns       159411

$ clang++ -v

  Ubuntu clang version 12.0.0-1ubuntu1

$ clang++ main.cpp -O3 -lbenchmark -std=c++20 
$ ./a.out

  2021-05-07T07:55:43+03:00
  Running ./a.out
  Run on (4 X 3000 MHz CPU s)
  CPU Caches:
    L1 Data 32 KiB (x4)
    L1 Instruction 32 KiB (x4)
    L2 Unified 256 KiB (x4)
    L3 Unified 12288 KiB (x4)
  Load Average: 0.55, 0.52, 0.44
  ---------------------------------------------------------------
  Benchmark                     Time             CPU   Iterations
  ---------------------------------------------------------------
  to_stringNewBench/1         689 ns          688 ns       985090
  to_stringNewBench/2         835 ns          834 ns       845049
  to_stringNewBench/3        1350 ns         1348 ns       522957
  to_stringNewBench/4        1364 ns         1362 ns       518457
  to_stringNewBench/5         939 ns          938 ns       735028
  to_stringNewBench/6         963 ns          962 ns       710654
  to_stringNewBench/7         978 ns          977 ns       700846
  to_stringNewBench/8         972 ns          971 ns       686915
  to_stringNewBench/9         929 ns          928 ns       715882
  to_stringNewBench/10       1002 ns         1001 ns       708947
  to_stringNewBench/11       1009 ns         1008 ns       651967
  to_stringNewBench/12       1040 ns         1039 ns       640582
  to_stringNewBench/13        996 ns          995 ns       682899
  to_stringNewBench/14       1026 ns         1025 ns       683722
  to_stringNewBench/15        977 ns          976 ns       710880
  to_stringNewBench/16        965 ns          964 ns       712499
  to_stringOldBench/1         970 ns          969 ns       697907
  to_stringOldBench/2        1066 ns         1064 ns       629877
  to_stringOldBench/3        1489 ns         1488 ns       466870
  to_stringOldBench/4        1482 ns         1480 ns       469958
  to_stringOldBench/5        4080 ns         4076 ns       170472
  to_stringOldBench/6        4219 ns         4214 ns       167871
  to_stringOldBench/7        4049 ns         4045 ns       169507
  to_stringOldBench/8        4205 ns         4190 ns       165934
  to_stringOldBench/9        4094 ns         4091 ns       167660
  to_stringOldBench/10       4041 ns         4038 ns       168373
  to_stringOldBench/11       4120 ns         4117 ns       170489
  to_stringOldBench/12       4084 ns         4080 ns       170699
  to_stringOldBench/13       4020 ns         4017 ns       171939
  to_stringOldBench/14       4125 ns         4122 ns       171680
  to_stringOldBench/15       4091 ns         4088 ns       169939
  to_stringOldBench/16       4085 ns         4082 ns       170872


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D101752/new/

https://reviews.llvm.org/D101752