Author: Dann Corbit
Date: 21:08:48 02/23/01
Go up one level in this thread
On February 24, 2001 at 00:01:33, Dan Newman wrote: [snip] >Heh. I bought the Intel compiler (version 5.0) on the basis of reports of >better performance than MSVC, but on my program I get a 24% hit instead... >I imagine what may happen is that if you optimize for one compiler, you >end up doing worse on another. Generally I've found MSVC to be 20-30% >better than anything else I've tried (Watcom, gcc, Intel). > >(It could be that I just haven't found the right switches yet.) > >The Intel compiler did find a few minor "bugs" that have been slipping past >MSVC--got 4 warnings that really needed fixing--so it's not a total loss :). Here's another benefit: Compiler using the Intel compiler with -Qprof_usex and run your program for a long time doing EPD test suites and then run it as a chess engine. Run profmerge.exe and proforder.exe to create a link order file. Take the link order file and add in the link line for the MS VC++ compiler: /ORDER:@proford.txt and that will almost always provide a speedup for the Microsoft compiler. You can use /Qip and /Qipo with the link order file, but don't specify /Qprof_use.
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.