[bug-grep] Re: fgrep execution time

[David Bailey]

Stepan,

Very impressive! Thanks for the tip. 

> GNU grep 2.5.1 is awfully inefficient with non-C locales.
> 
> How would the times look like if you tried the following?
> 
> time yes "a very long file" | \
>     head -10000 | LC_ALL=C /usr/local/bin/fgrep -f /dev/null

Looks like the locale setting caused the problem.

When we set LC_ALL=C, the run times were similar to /bin/fgrep and 
/usr/xpg4/bin/grep -F.

prompt> setenv LANG ja
prompt> time yes "a very big file" | \
        head -100000 | /usr/local/bin/grep -f /dev/null
0.0u 0.0s 1:15 0% 0+0k 0+0io 0pf+0w

prompt> setenv LANG C
prompt> time yes "a very big file" | \
        head -100000 | /usr/local/bin/grep -f /dev/null
0.0u 0.0s 0:00 0% 0+0k 0+0io 0pf+0w

The third field in the output of the csh time command is wallclock time.

75 seconds when locale is ja (Japanese), less then 1 second when locale
is C.

BTW: We get the same results under Linux (RedHat EL 3.0 AS).

David M. Bailey

bailey-david.vcf
Description: David Bailey ÌJ[h