djbsort-speed - Man Page
NAME
djbsort-speed - run some djbsort speed tests
SYNOPSIS
djbsort-speed
DESCRIPTION
djbsort-speed runs some tests of the speed of the djbsort library.
The djbsort-speed output format is subject to change. Examples of output lines with the current format, from a djbsort-speed run on an AMD Zen 3 CPU core, include the following:
sort_int32 selected implementation avx2 compiler gcc -Wall -fPIC -fwrapv -O2 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -mavx -mbmi -mbmi2 -mpopcnt -mavx2 -mtune=haswell; gcc (Debian 12.2.0-14+deb12u1) 12.2.0; Copyright (C) 2022 Free Software Foundation, Inc.; This is free software; see the source for copying conditions. There is NO; warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. sort_int32 selected 1597 8769 +1335+200-12+36-20-70+20-61-25+57-55+26+4-22+10+0+26-1-70-41+53-59-25+57-55+26+4-22+10+0+26-1
These lines say that the automatically selected implementation for djbsort_int32 was the avx2 implementation compiled with the clang options shown, taking an interquartile mean (IQM) of 8769 CPU cycles to sort an int32[1597] array. The numbers 8769 +1335+200+... show the timings for a series of consecutive operations: the IQM was 8769 cycles, the first run took 1335 cycles more than the IQM, the second run took 200 cycles more than the IQM, etc.