[GitHub] [arrow] wesm removed a comment on pull request #7356: ARROW-9034: [C++] Implement "BinaryBitBlockCounter", add single-word functions to BitBlockCounter

[GitBox <gi...@apache.org>]

wesm removed a comment on pull request #7356:
URL: https://github.com/apache/arrow/pull/7356#issuecomment-639547819


   Benchmarks
   
   ```
   BitBlockCounterSum/8                           1687631 ns    1687634 ns        407   592.545M items/s
   BitBlockCounterSum/64                           730805 ns     730798 ns        970    1.3363G items/s
   BitBlockCounterSum/512                          179964 ns     179949 ns       3923   5.42689G items/s
   BitBlockCounterSum/4096                          75167 ns      75165 ns       9461   12.9922G items/s
   BitBlockCounterSum/32768                         58679 ns      58679 ns      12065   16.6425G items/s
   BitBlockCounterSum/65536                         57921 ns      57922 ns      12311     16.86G items/s
   BitBlockCounterSumWithOffset/8                 1672146 ns    1672131 ns        419   598.039M items/s
   BitBlockCounterSumWithOffset/64                 735521 ns     735515 ns        944   1.32773G items/s
   BitBlockCounterSumWithOffset/512                186470 ns     186468 ns       3782   5.23716G items/s
   BitBlockCounterSumWithOffset/4096                82928 ns      82928 ns       8544    11.776G items/s
   BitBlockCounterSumWithOffset/32768               66855 ns      66855 ns      10402   14.6072G items/s
   BitBlockCounterSumWithOffset/65536               66821 ns      66822 ns      10131   14.6145G items/s
   BitBlockCounterInlineSum/8                     1563707 ns    1563716 ns        448   639.502M items/s
   BitBlockCounterInlineSum/64                     596145 ns     596143 ns       1166   1.63813G items/s
   BitBlockCounterInlineSum/512                    122461 ns     122461 ns       5554    7.9745G items/s
   BitBlockCounterInlineSum/4096                    30327 ns      30327 ns      22765   32.2007G items/s
   BitBlockCounterInlineSum/32768                   16484 ns      16484 ns      40881   59.2429G items/s
   BitBlockCounterInlineSum/65536                   16145 ns      16145 ns      43742   60.4885G items/s
   BitBlockCounterInlineSumWithOffset/8           1577564 ns    1577548 ns        441   633.895M items/s
   BitBlockCounterInlineSumWithOffset/64           609330 ns     609319 ns       1091   1.60271G items/s
   BitBlockCounterInlineSumWithOffset/512          126930 ns     126931 ns       5580   7.69365G items/s
   BitBlockCounterInlineSumWithOffset/4096          33720 ns      33720 ns      20848   28.9608G items/s
   BitBlockCounterInlineSumWithOffset/32768         20190 ns      20190 ns      34575   48.3693G items/s
   BitBlockCounterInlineSumWithOffset/65536         19914 ns      19914 ns      35329   49.0382G items/s
   BitBlockCounterFourWordsSum/8                  1557230 ns    1557208 ns        456   642.175M items/s
   BitBlockCounterFourWordsSum/64                  755518 ns     755523 ns        936   1.29257G items/s
   BitBlockCounterFourWordsSum/512                 314926 ns     314928 ns       2237   3.10091G items/s
   BitBlockCounterFourWordsSum/4096                 62012 ns      62011 ns      10790   15.7483G items/s
   BitBlockCounterFourWordsSum/32768                18307 ns      18307 ns      38225   53.3436G items/s
   BitBlockCounterFourWordsSum/65536                16755 ns      16755 ns      42106   58.2835G items/s
   BitBlockCounterFourWordsSumWithOffset/8        1554037 ns    1554017 ns        455   643.494M items/s
   BitBlockCounterFourWordsSumWithOffset/64        771938 ns     771925 ns        878    1.2651G items/s
   BitBlockCounterFourWordsSumWithOffset/512       311751 ns     311753 ns       2200   3.13249G items/s
   BitBlockCounterFourWordsSumWithOffset/4096       65821 ns      65821 ns      10635   14.8366G items/s
   BitBlockCounterFourWordsSumWithOffset/32768      21528 ns      21528 ns      32434    45.362G items/s
   BitBlockCounterFourWordsSumWithOffset/65536      20499 ns      20499 ns      35251   47.6401G items/s
   BinaryBitBlockCounterSum/8                     1369784 ns    1369771 ns        529   730.048M items/s
   BinaryBitBlockCounterSum/64                      75780 ns      75780 ns       9360   12.8867G items/s
   BinaryBitBlockCounterSum/512                     56490 ns      56491 ns      12401   17.2871G items/s
   BinaryBitBlockCounterSum/4096                    55213 ns      55213 ns      12094   17.6871G items/s
   BinaryBitBlockCounterSum/32768                   56260 ns      56260 ns      12740    17.358G items/s
   BinaryBitBlockCounterSum/65536                   57650 ns      57650 ns      12731   16.9394G items/s
   BinaryBitBlockCounterSumWithOffset/8           1433256 ns    1433222 ns        479   697.728M items/s
   BinaryBitBlockCounterSumWithOffset/64            96996 ns      96996 ns       7346    10.068G items/s
   BinaryBitBlockCounterSumWithOffset/512           75189 ns      75188 ns       9420   12.9882G items/s
   BinaryBitBlockCounterSumWithOffset/4096          74190 ns      74191 ns       9414   13.1628G items/s
   BinaryBitBlockCounterSumWithOffset/32768         74885 ns      74885 ns       9362   13.0409G items/s
   BinaryBitBlockCounterSumWithOffset/65536         74462 ns      74462 ns       9425   13.1148G items/s
   ```
   
   I'm surprised that the binary sum (BinaryBitBlockCounterSum) seems faster than the unary sum (BitBlockCounterSum). I'm going to scrutinize the benchmark to check whether I've implemented something incorrectly. 


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org