This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [patchv2 2/2] Accelerate lookup_symbol_aux_objfile 14.5x [Re: [patch 0/2] Accelerate symbol lookups 15x]
- From: Doug Evans <xdje42 at gmail dot com>
- To: Jan Kratochvil <jan dot kratochvil at redhat dot com>
- Cc: "gdb-patches at sourceware dot org" <gdb-patches at sourceware dot org>
- Date: Sun, 26 Oct 2014 23:02:43 -0700
- Subject: Re: [patchv2 2/2] Accelerate lookup_symbol_aux_objfile 14.5x [Re: [patch 0/2] Accelerate symbol lookups 15x]
- Authentication-results: sourceware.org; auth=none
- References: <20141020214410 dot GA22011 at host2 dot jankratochvil dot net> <CAP9bCMQ7EXyXJiqK4j2UA9YgxkQiNFFqJPOpbPXH8-YZMRLh2w at mail dot gmail dot com> <20141023182434 dot GA31412 at host2 dot jankratochvil dot net> <CAP9bCMTrzU7srWLfiS2814nfBEySJ-i6yKy7AKHBvXbUoLa-rQ at mail dot gmail dot com> <20141024073308 dot GA20087 at host2 dot jankratochvil dot net> <CAP9bCMSrz7HswXRckCvTy7tcWrNktbbYm6hXcT3GxLCZ1zNTwA at mail dot gmail dot com> <m3k33mjbw8 dot fsf at sspiff dot org>
On Sun, Oct 26, 2014 at 10:54 PM, Doug Evans <xdje42@gmail.com> wrote:
> [...]
> For me, without your dictionary hash caching patch
> I see 13.8M calls to iter_match_first_hashed for "p/r var".
> 13.8 million. Yikes.
> With your dictionary hash caching patch that reduces to 31.
> With the appended experiment it is reduced to 55.
> Not as good, though the difference from 13.8M is in the noise at this
> point :-).
Sorry, poor wording.
The latter number, 55, is the number of calls to
iter_match_first_hashed with the experimental patch.
The 13.8M number is the number of calls to iter_match_first_hashed
with your 1/2 patch and
the 31 number is the number of times a new hash is computed.