TIL that I can use Perl's Benchmark module to time and compare the performance of different commands in an OS-agnostic way, ie as long as Perl is installed.
For example, to benchmark curl, wget and httpie you could simply run:
$ perl -MBenchmark=:all \
-E '$cmd_wget = sub { system("wget https://google.com > /dev/null 2>&1") };' \
-E '$cmd_curl = sub { system("curl https://google.com > /dev/null 2>&1") };' \
-E '$cmd_httpie = sub { system("https https://google.com > /dev/null 2>&1") };' \
-E '$timeresult = timethese(15, { "wget" => $cmd_wget, "curl" => $cmd_curl, "httpie" => $cmd_httpie });' \
-E 'cmpthese($timeresult)'
which on my old T530 produces:
Benchmark: timing 15 iterations of curl, httpie, wget...
curl: 2 wallclock secs ( 0.00 usr 0.00 sys + 0.42 cusr 0.11 csys = 0.53 CPU) @ 28.30/s (n=15)
httpie: 8 wallclock secs ( 0.00 usr 0.01 sys + 4.63 cusr 0.79 csys = 5.43 CPU) @ 2.76/s (n=15)
wget: 3 wallclock secs ( 0.00 usr 0.00 sys + 0.53 cusr 0.19 csys = 0.72 CPU) @ 20.83/s (n=15)
Rate httpie wget curl
httpie 2.76/s -- -87% -90%
wget 20.8/s 654% -- -26%
curl 28.3/s 925% 36% --
Very handy indeed ❤
See also: The shell's own
timescommand (where available), and/or the external commandtime, which is the fore-runner of all of these.Those won't run a command a number of times or provide a pretty table like Perl's benchmarker will, but if you want to roll your own, those would be the place to start.
(I'm a big fan of Perl, but know that not everyone's ready for it yet!)