Re: profile-driven optimization and the linker?

[Dan Kegel <dank at kegel dot com>]

Dan Kegel wrote:
> I went looking for instances of people using gprof's --function-ordering
> option, but came up fairly empty handed.  Anyone
> have good references?  Here's the little bit I dug up:

One more example: mozilla does something like this (driven
not by gprof, but by some other execution trace analyzer);
the perl script that generates the linker script, along with
other interesting stuff, is the last attachment to
https://bugzilla.mozilla.org/show_bug.cgi?id=65845
- Dan

> Ben's post just now was pretty good, even though that was for data and 
> not code:
> http://gcc.gnu.org/ml/gcc/2004-11/msg01002.html
> Did reordering the global variables end up helping performance on any
> interesting tasks?
>
> http://gcc.gnu.org/ml/gcc-patches/2001-11/msg00935.html says
> "Steve Christiansen tried using gprof output to create a linker
> script that orders functions based on run-time call graphs
> and call counts, but couldn't show that it made a difference, based on 
> SPEC CPU2000 results."
> (Since the --function-ordering option was added to gprof at
> the end of 1995, I imagine Steve Christiansen used it.
> On the other hand, since glibc aborted with more
> than 64K symbols when run with -pg until late 2002,
> maybe he ran into trouble there.)
>
> http://vr3.uid0.sk/cd/Mailinglist_Archives/agenda-mail/agenda-user/200107/995378424.txt 
> says
> Jay Carlson gave it a try, but doesn't have any results.
>
> http://sources.redhat.com/binutils/docs-2.15/ld/Input-Section.html#Input%20Section 
>
> describes (just barely) how one writes a linker script
> to map a given list of input sections to an output section.
> That plus gcc's -ffunction-sections option are probably needed
> to use this.
> - Dan


--
Trying to get a job as a c++ developer?  See http://kegel.com/academy/getting-hired.html