Skip to content

Latest commit

 

History

History
31 lines (19 loc) · 1.56 KB

README.md

File metadata and controls

31 lines (19 loc) · 1.56 KB
Raw

Solutions for 1 Billion Record Challenge in Perl

  • 1brc-perl-semipar.pl reads file and forks when number of lines is reached, uses Parallel::ForkManager

  • 1brc-pfm-mestia.pl splits file into chunks and forks code for each section - higher memory usage but can utilize more cpu cores, bottleneck is obviously the I/O. Has a problem when requesting more cores than available, buffer passed to proc_chunc() gets rundomly truncated.

  • 1brc-mce-mestia.pl MCE implementation

Source file

compile and run create-sample.c for the sample file. Code is available here: https://github.com/dannyvankooten/1brc

Some stats ( running on AMD EPYC 7702P 64-Core Processor and a SSD pool )

 64 cores; real: 0m23.951s, user: 23m4.662s,  sys: 0m13.939s ; time perl 1brc-mce-mestia.pl ./measurements.txt 64
 8 cores ; real: 1m59.429s, user: 15m36.224s, sys: 0m7.145s  ; time perl 1brc-mce-mestia.pl ./measurements.txt 8
 8 cores ; real: 2m0.195s,  user: 15m22.971s, sys: 0m27.302s ; time perl 1brc-pfm-mestia.pl ./measurements.txt 8

Probably the cleanest solution without modules like P::FM or MCE, https://github.com/adriaandens/1brc.pl/blob/main/forkie.pl

another MCE option: https://github.com/janlimpens/1brc-perl/blob/main/calculate_average_mce.pl

See also: