Author: Dave Gomboc
Date: 14:47:03 06/02/99
Go up one level in this thread
On June 02, 1999 at 17:35:02, Gregor Overney wrote: >cut...cut...cut... > >>First block is for Dann's latest 16.8 exe on his site (mt=2): >> >>Total Nodes: 66918014 >>Raw nodes per second: 264205 >>C.A.P. ratio: 3.302571 >>SMP time-to-ply measurement: 3.666667 >> >>This is for the non-multiprocessor version: >> >>Total Nodes: 59979081 >>Raw nodes per second: 166696 >>C.A.P. ratio: 2.083707 >>SMP time-to-ply measurement: 5.202899 >> >>Especially the last item is a bit puzzling, because this is a single-cpu version >>of Crafty. Dann? >> >>The speedup from 1 to 2 processors on whatever position(s) are searched in >>"bench" is 1.585, this sounds pretty low to me... >> >>Dave > >On my dual p6/200, I see more than 90% improvement when running 'bench' on two >processors (mt=2). I assume you do not have any background tasks running that >require a lot of CPU cycles. Did you recompile 16.8 using /Ox /G6 /Gr /DCPU=2 >/MT /DSMP /DVC_INLINE_ASM? I have noticed that /Oa is counterproductive if I >applied it for the whole Crafty source. So I do not use it (although it's >working now). > >Just to make sure, you are using SP3 for VC++6. Right? > >Gregor The binaries (both of them) were from Dann's FTP site. He would know the exact compiler options he used, but I don't. I had a couple of windows open, but didn't have any compute-intensive tasks running, and didn't use the computer while bench was running. Dave
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.