On 2021/08/24 05:06, Greg Wooledge wrote:
Looks like the efficiency of "read -ra" vs. a shell loop just about makes
up for the system calls used for the here string (f6 and f7 are almost
tied in overall speed, with f6 just a *tiny* bit faster). Good to know.
----
If you set your TIMEFORMAT -- I put this in my login scripts:
export TIMEFORMAT="%2Rsec %2Uusr %2Ssys (%P%% cpu)"
the 4th field can give a feel for parallelism:
fn1() { /usr/bin/ls -1 /usr/bin >/tmp/tt && wc </tmp/tt && rm /tmp/tt ;
}
fn2() { /usr/bin/ls -1 /usr/bin | wc ; }
time fn1 >/dev/null
0.01sec 0.01usr 0.00sys (101.56% cpu)
time fn2 >/dev/null
0.01sec 0.00usr 0.00sys (134.03% cpu)
