Theoretical Performance Question

[Brian Budge]

On Thu, Jan 27, 2011 at 3:11 PM, Brian D. McGrew <brian@visionpro.com> wrote:
> Greetings All!
>
> Currently we compile our software with (gcc version 4.1.2 20080704 (Red Hat
> 4.1.2-44)) which comes standard on CentOS 5.4-x86_64.  We're all straight
> C/C++ but we do a lot with X11/Xm/Xt (including some really, really
> old-school IPC with Xfunnels!)
>
> Deploying our software on the previous generation of Dell Precision T5500
> workstations (from last week) with whatever current BIOS version the box
> comes with is pig slow.  So, we'd downgrade the BIOS to Dell's A02 from last
> year and life was good, all our software ran as it should.
>
> Fast forward a week and Dell has changed the processor in these boxes along
> with the motherboard and BIOS revision (still calling it a T5500) and we're
> in a predicament.  We can't downgrade the BIOS anymore, it'll brick the box
> and our software is once again pig slow.
>
> So my question to the experts is...  Would upgrading to a more current
> compiler possibly get me any better performance???  As in taking advantage
> of the newer hardware, bios, dma, magic bus, etc???  Or, is the the compiler
> I'm using new enough that I should be looking other places for my hang
> ups???
>
> Thanks,
>
> -b
>

Hi Brian -

This is partially off topic, as mostly I don't think a gcc upgrade
will help you much.  Feel free to follow up with me off-list if you
want.

The fact that the BIOS version makes such a big difference is quite
frightening.  There's a good chance that newer gcc could generate
better code for your new processors, but I doubt very much that it
will make a difference for the other aspects you mention.

I also doubt that an upgrade would give you a large performance gain
unless you can verify that your critical path could be made much
better by utilizing your processors instructions better.

Do you have some idea what your bottleneck is?

  Brian