Author: Robert Hyatt
Date: 14:53:31 12/26/02
Go up one level in this thread
On December 26, 2002 at 11:26:20, Frank Phillips wrote: >On December 26, 2002 at 10:59:34, Robert Hyatt wrote: > >>On December 26, 2002 at 10:49:52, Frank Phillips wrote: >> >>>On December 26, 2002 at 09:33:43, Eugene Nalimov wrote: >>> >>>>From my experience: Intel's compiler shines where application spent majority of >>>>its time in several tight loops. When execution time is more-or-less evenly >>>>spread across the large application it's more important to get shorter code than >>>>to emit locally optimal but longer code. >>>> >>>>Thanks, >>>>Eugene >>> >>>This seems consistent with an Intel 1.1 MByte executable compared to around 400 >>>kBytes for gcc. >>> >>>Frank >> >> >>It is probably unrelated to that. The size difference is most likely a >>library issue. IE my intel executables on linux are in the same size range >>is the executables for gcc 3.x. But both use ld.so-type shared libraries that >>are _not_ a part of the executable file while it sits on disk. > > >The executables are for Linux. Linked with a standard install of the Intel >compiler; and not statically linked. Are you linking with another library? > >Frank > > Nope. I use redhat 8.0, and make crafty with the linux-icc-elf option, after doing the profile option of course. My executable is under 800KB... >> >>> >>>> >>>>On December 25, 2002 at 22:34:41, Robert Hyatt wrote: >>>> >>>>>On December 24, 2002 at 10:55:17, Vincent Diepeveen wrote: >>>>> >>>>>>On December 24, 2002 at 04:47:21, Frank Phillips wrote: >>>>>> >>>>>>>On December 23, 2002 at 12:12:29, Vincent Diepeveen wrote: >>>>>>> >>>>>>>>On December 23, 2002 at 12:01:10, Robert Hyatt wrote: >>>>>>>> >>>>>>>>You forget the crucial data point and that's that you have >>>>>>>>no AMD K7s out there. >>>>>>> >>>>>>>Intel is about 30% faster than gcc3.2 (and gcc2.95 and gcc 2.96) with profile >>>>>>>guided optimisation for me on my AMD Athlons (Palomino and Thoroughbred). Not >>>>>>>all of this can be due to incompetence, I suggest. >>>>>>> >>>>>>>Frank >>>>>> >>>>>>you must be using bitboards then. No other option possible. >>>>>>profile guided optimization speeds me up 20% at k7 with gcc 3.x >>>>>> >>>>> >>>>> >>>>>How can you possibly say "no other option possible"??? >>>>> >>>>>Intel _also_ does profile-guided optimizations. But I can think of _lots_ >>>>>of reasons why he might get better results than you get. Starting with his >>>>>programming style which might be _better_...
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.