Author: Robert Hyatt
Date: 14:19:49 12/14/03
Go up one level in this thread
On December 14, 2003 at 05:38:25, Russell Reagan wrote: >On December 14, 2003 at 05:07:10, Gian-Carlo Pascutto wrote: > >>On December 14, 2003 at 04:43:20, Russell Reagan wrote: >> >>>I used to use Microsoft Visual C++ 6 Professional (because I got it for $5 from >>>my school, it's normally several hundred at least). I quit using it when I saw >>>that the executable that gcc/g++ created was significantly faster than the >>>executable that MSVC++ 6 Pro created. >> >>Academic versions usually lack the optimizer. >> >>A complete MSVC6 should be competitive or better than any current version of GCC >>for most thing. If not you messed up somewhere. > >Well it is doing some optimization, because when I turn on optimizations it is >significantly faster than without, but still not as fast as a simple gcc -O2 or >-O3. You could be right though. It might be handicapped in some way. gcc 3.x is significantly better than older gcc versions, but nowhere near the Intel compiler for me... > >Here is a quick example using TSCP. > >MSCV 6 Pro default Debug build > >Nodes: 550778 >Best time: 5608 ms >Nodes per second: 98212 (Score: 0.404) > > >MSVC 6 Pro default Release build > >Nodes: 550778 >Best time: 2534 ms >Nodes per second: 217355 (Score: 0.894) > > >MSVC 6 Pro Release build after some tinkering with optimization settings > >Nodes: 550778 >Best time: 2494 ms >Nodes per second: 220841 (Score: 0.908) > >GCC 3.3.1 Cywgin: gcc > >Nodes: 550778 >Best time: 4514 ms >Nodes per second: 122015 (Score: 0.502) > > >GCC 3.3.1 Cywgin: gcc -O1 > >Nodes: 550778 >Best time: 2698 ms >Nodes per second: 204143 (Score: 0.840) > > >GCC 3.3.1 Cygwin: gcc -O2 > >Nodes: 550778 >Best time: 2438 ms >Nodes per second: 225913 (Score: 0.929) > > >GCC 3.3.1 Cygwin: gcc -O3 > >Nodes: 550778 >Best time: 2437 ms >Nodes per second: 226006 (Score: 0.929) > > >Best times: >MSVC6Pro 2494 ms >GCC3.3.1 2437 ms > >So there it is roughly equal, although I imagine I could get some more out of >gcc by using some more specific optimization options. > >I think the case where I saw gcc give significantly better results was in a >bitboard program I wrote that was somewhat unconventional. It could have been >some rare thing in that program that I was doing that MSVC didn't like that gcc >knew how to handle. Who knows.
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.