Hi there.

Nobody thought to mention that static linking can dramatically increase
performance,
or to put it the other way around, dynamic linking can incur serious
runtime penalties.


I don't want to encourage everyone to start static linking everywhere
just to get a
few percentage points in efficiency, obviously your mileage may vary,
but basically
my treedb project got 100 times faster and meta-treedb got at least 20
times faster,

through static linking on x86_64.

You can get these from SourceForge.

These do little more than function calls and memory access through
memory maps,

so maybe they're unusual in this respect.

I've yet to upload the versions (including v3c) that build static
libraries and the

static tests, but I thought it was worth mentioning.

Philip


--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: 4DAA0912.6070005@philipashmore.com">http://lists.debian.org/4DAA0912.6070005@philipashmore.com

On 2011-04-16, Philip Ashmore <contact@philipashmore.com> wrote:
> Nobody thought to mention that static linking can dramatically
> increase performance, or to put it the other way around, dynamic
> linking can incur serious runtime penalties.

At the expence of more RAM in use if multiple instances of the problem
are running because the library pages can't be shared anymore.

But of course linking might make smarter decisions.

Kind regards
Philipp Kern


--
To UNSUBSCRIBE, email to debian-devel-REQUEST@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Archive: slrniqk3hp.447.trash@kelgar.0x539.de">http://lists.debian.org/slrniqk3hp.447.trash@kelgar.0x539.de