This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH] GDB/testsuite: Bump up `match_max'
- From: "Maciej W. Rozycki" <macro at codesourcery dot com>
- To: Doug Evans <dje at google dot com>
- Cc: Joel Brobecker <brobecker at adacore dot com>, Tom Tromey <tromey at redhat dot com>, gdb-patches <gdb-patches at sourceware dot org>
- Date: Tue, 20 May 2014 01:46:28 +0100
- Subject: Re: [PATCH] GDB/testsuite: Bump up `match_max'
- Authentication-results: sourceware.org; auth=none
- References: <alpine dot DEB dot 1 dot 10 dot 1405172136000 dot 12061 at tp dot orcam dot me dot uk> <87bnutzwbj dot fsf at fleche dot redhat dot com> <20140519142324 dot GB22822 at adacore dot com> <20140519143702 dot GC22822 at adacore dot com> <CADPb22SVw2kVVMrTVXTQWN1chb4V-+N_=TXHq7J30Sx5B8FY_w at mail dot gmail dot com>
On Mon, 19 May 2014, Doug Evans wrote:
> >> > I wonder whether you timed the test suite?
> >> > The expect man page says:
> >> >
> >> > This may be changed with the function match_max. (Note that excessively
> >> > large values can slow down the pattern matcher.)
> >> >
> >> > If it is notably slower then it would be better to rewrite the macro
> >> > tests to avoid this need.
> >>
> >> Funny you would say that! I was reviewing the patch, and decided to
> >> do exactly that. Ran into trouble (fresh install), but almost there...
> >
> > Here are the results. As I hoped, it doesn't seem to introduce
> > any noticeable difference (at -j16 on an 8-thread machine).
> >
> > Before: 1093.79s user 153.20s system 589% cpu 3:31.68 total
> > After: 1097.58s user 155.08s system 589% cpu 3:32.39 total
>
> fwiw, I did several runs of before/after with the testsuite running
> serially and didn't find any statistical difference.
> All runs were in the range 14:07s to 14:25s elapsed, and sometimes
> with-patch was faster.
> Not unexpected I guess - most of the time what's actually in the
> buffer is pretty small, much less than the buffer size, so other
> factors would (generally) have more of an influence on run time.
Have we reached consensus? At this point of discussion I don't have
anything to add -- all has been already written AFAICT. Thank you all for
verifying the change.
Maciej