[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

[Moderately OT] Compiler Optimisation (was Re: intel 7.0 compilers)



On Thu, 2002-12-19 at 09:51, John Schmidt wrote:
> On Thursday 19 December 2002 01:08 am, Oleg wrote:
> >
> > P.P.S To preemptively answer the inevitable question why I want Intel
> > compilers in addition to GCC: I'm interested in high-performance
> > numeric computing, looking at the code Intel compilers generate, and
> > comparing different compilers' performance.
> 
> Our university funded DOE ASCI center went through a similar comparison 
> with a large C++ code.  We actually had several Intel engineers bang 
> away on the compilers trying to tweak compiler flags and the code 
> itself.  Even turning on special undocumented flags, and tweaking the 
> code, the performance was no better than using gcc.  In fact the 
> optimizations they did to the code also helped gcc.  We didn't find 
> that the Intel compilers would yield any benefit.  Your mileage will 
> vary though.  Good luck
> 
> John Schmidt

I don't know how dated it is now, but I remember at one point the Watcom
compilers were no slouch at tight, fast code - IBM used them to compile
the Windows 3.1 support for OS/2 and got such improvements that Windows
running as just one task on an OS/2 box ran faster than MS Windows
running on MS-DOS. I remember the coding work from a couple decades back
when I worked there, and I'm curious if there are aspects of the now
open source code that could be constructively integrated in gcc (just
idle thoughts on my part - there is a reason why I do systems analysis
rather than down in the bits coding now, and why I'm not hacking patches
myself to gcc.)
-- 
Mark L. Kahnt, FLMI/M, ALHC, HIA, AIAA, ACS, MHP
ML Kahnt New Markets Consulting
Tel: (613) 531-8684 / (613) 539-0935
Email: kahnt@hosehead.dyndns.org

Attachment: signature.asc
Description: This is a digitally signed message part


Reply to: