Author: Vincent Diepeveen
Date: 06:44:18 05/11/02
Go up one level in this thread
On May 10, 2002 at 23:20:32, Robert Hyatt wrote: >On May 09, 2002 at 22:18:18, Vincent Diepeveen wrote: > >>On May 09, 2002 at 10:21:28, Robert Hyatt wrote: >> >>>On May 08, 2002 at 23:54:50, Vincent Diepeveen wrote: >>> >>>>Sometimes i hate company politics. A superb compiler intel c++ >>>>they weakened it for K7. Where 5.0.1 compiler was pretty fast >>>>for the superb K7 processors, the 6.0 compiler is deliberatly >>>>doing bad for K7. >>> >>> >>>I doubt it is "deliberate". Why would Intel worry about optimizing for >>>the K7 when they don't build them? >>> >>>Wouldn't make any sense at all, IMHO... >> >>because 5.0.1 is faster on k7 than 6.01 >> > > >So what? I have occasionally done things to Crafty to make it faster on >a pentium, which also makes it slower on _other_ architectures. But since >I do 99% of my testing on Intel, this is the way things work out. I could >see lots of places where optimizing for intel would break AMD processor >pipes at times... The inverse is also true, but since AMD is not writing >a compiler for their chip, they don't have something to make them look >their best... > >Which will always hurt them on SPEC of course... > AMD must be busy with their own compiler, otherwise the whole Hammer chip is a major joke; only alternative could be visual m$ works on a compiler for it. From a company viewpoint having your own compiler is smartest as it can run at any OS. > > > >>> >>> >>> >>>> >>>>intel c++ 6.0 linux KDE3 MP 1.4Ghz run single cpu >>>>is 2.4% slower than msvc 6.0 sp4+processorpack. >>>> >>>>Please take into account this is WITH profile information. >>>>msvc is already faster without taking advantage of that. >>>> >>>>God, sometimes i hate company politics. >>>> >>>>Last so many years msvc did shit litterary regarding to improve >>>>compiler (at least released nearly nothing). In total from >>>>4.5 to 6.0 sp4 procpack it got faster 3% from which the processor >>>>pack is 2%. >>>> >>>>Intel c++ which is taking use of all kind of new tricks is >>>>actually 8.5% faster on a P3, but 2.4% slower on a K7. >>>> >>>>Sometimes you can dominate because others make major blunders... >>>>...of course intel doesn't sell many compiler versions, so >>>>i understand the math. but sometimes i just hate company politics. >>>>couldn't they make it *equally* fast to the msvc compiler for example?
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.