objdump *really* slow on RH8 machine?

objdump *really* slow on RH8 machine?

[Emile Snyder]

Hi all,

Has anyone experienced really slow performance from objdump with a gcc3.x
based system vs. older?

We've been using objdump as part of a little helper script to get human
readable backtrace info from crashes in a statically built binary on our
project.  Currently we build on Red Hat 7.1 boxes (stock gcc 2.96, stock
objdump 2.11.93.0.2), our application executable is ~30MB, and 'objdump -d
--line-numbers <program>' takes ~1minute and produces about ~90MB of
output.

Building on a stock RH 8.0 box (gcc 3.2, objdump 2.13) the executable size
is more like 55MB, and the objdump run takes on the order of hours.
vmstat shows that the CPU is essentially pegged, but there's no swapping
going on or anything.

I've built the 2.13.2pre2 binutils on one of the RH7.1 boxes, and the
objdump from it also only takes ~1minute.

Both machines are 600MHz PIIIs that seem comprable in performance for
other tasks.

Has anyone else experienced this?  Am I asking for help in the right
formum, or should I be asking the gcc people?  Any pointers at all would
be appreciated.

thanks all for your time,
-emile

Re: objdump *really* slow on RH8 machine?

[Nick Clifton]

Hi Emile,

> Has anyone experienced really slow performance from objdump with a
> gcc3.x based system vs. older?

Noone else has reported the kind of slow downs that you mention :-(

> Has anyone else experienced this?  Am I asking for help in the right
> formum, or should I be asking the gcc people?

This is the right forum.

> Any pointers at all would be appreciated.

Do you have any clue where objdump is pegged ?  If you build objdump
with profilng enabled you may be able to find out which routine(s) are
eating up the CPU's time.

Cheers
        Nick