Author: Robert Hyatt
Date: 11:58:27 09/04/02
Go up one level in this thread
On September 04, 2002 at 14:16:03, Vincent Diepeveen wrote:
>On September 03, 2002 at 22:34:29, martin fierz wrote:
>
>A number of things
> a) bob says he modified node counts now, so we cannot
> check how cray blitz scaled at all, because there is not
> a single log file saved anymore of it (words of bob) in 2002.
> It all 'by accident' dissappeared.
Node counts are irrelevant, since time is a function of node counts and
vice-versa.
The "accident" was well-documented years ago, although I didn't think about
all the logs, and other things I had been saving for nearly 25 years, at the
time...
> b) bob has a cray executable as he said in CCC, NWO has cray machines,
> why doesn't he ship a cray executable to me?
> When i offered that he said he didn't have them anymore, not
> source anymore.
No source is around. What would you do with a Cray executable? The only
one I have is for a C90, and won't run on others, and it is for a specific
version of unicos that is almost 10 years old. Think you can find something
to run it on? If so, let me know...
> But an email from june Bob says he does have source code, so why
> hide it? I'm interested in how he did the parallel thing. Don't worry
>
I said I had source code. I have a thick book with all 50,000 lines of
fortran and 25,000+ lines of assembly. How do you want that delivered and
who is going to pay for it? And what would you do with it?
I have _zero_ stuff on disk. I might be able to find a pre-1990 version of
the code somewhere, I don't know as I haven't started asking around. There are
_no_ copies since 1990 around because I kept them here, and managed to lose them
after a hardware problem.
I found a few copies of pre-1990 code but none would run. We had some
interesting bugs that showed up when moving to newer machines and I was
not interested in trying to go back and reconstruct what we found and then
fix very old versions. You always asked about the SE version, which was
1994 version. It's gone, sadly. The 1993 version used for the DTS paper
was pre-SE and was actually pretty solid... But trying to reconstruct that
from a 1990 version would be a task that I'm not willing to invest the time
in with the machines "gone".
for assembly or fortran, i CAN read assembly and fortran.
>
>--------------------------------------
>Return-Path: <hyatt@cis.uab.edu>
>X-XS4ALL-To: <diep@maildrop.xs4all.nl>
>X-Authentication-Warning: crafty.cis.uab.edu: hyatt owned process doing -bs
>Date: Mon, 17 Jun 2002 21:47:11 -0500 (CDT)
>From: "Robert M. Hyatt" <hyatt@cis.uab.edu>
>X-X-Sender: <hyatt@crafty>
>To: Vincent Diepeveen <diep@xs4all.nl>
>Subject: Re: advice supercomputer
>
>On Tue, 18 Jun 2002, Vincent Diepeveen wrote:
>
>> Bob, please don't shout about icc.
>>
>> first of all a 2001 diep version plays there.
>> secondly it plays with a random book without learning.
>> third at inferior hardware.
>>
>> but most important, it's not slow level. it's not 3 minutes a move
>> what gets played there.
>
>
>The recent games I watched were 30 30 and 60 30 time controls.
>Very similar to ICCA tournaments. The hardware being used for some
>was a dual AMD which was as fast as my quad xeon...
>
>
>
>>
>> So don't shout about ICC. that's nonsense.
>>
>> Now imagine that my tournament book is nearly that good and the engine
>> is better.
>>
>> Anyway, you don't have source code from cray blitz anymore,
>> so it's hard to talk about how cray blitz would do.
>
>Sorry, but I still have source code for Cray Blitz. Including
>20,000 lines of assembly language...
>
>Therefore I _know_ how it will do because I occasionally get time on
>a Cray to fiddle around, and Cray Blitz is the first thing I fiddle
>with given the opportunity.
>
>
>
>
>>hi vincent,
>>
>>i cannot understand a word of what you say. what i see in the table is that all
>>times given in the table, with the exception of the 1-processor column, are NOT
>>measured numbers, but calculated numbers with the help of a factor which is
>>rounded to one decimal place.
>>
>>>So Bob *had* to fake the outputs of 1-8 processors or his 16 processor
>>>thing would look silly though it wasn't at all.
>>
>>these "calculated" times are there in all columns - 2,4,8,16 processors. the
>>16-processor time is also calculated from a 1-decimal place factor, just like
>>the 2,4,8 columns. therefore, there is nothing *in this table* to suggest that
>>the times in the 16-processor column are more or less "real" than the times in
>>the other columns.
>>
>>so....
>>
>>>>times. if you invent yourself a speedup number and
>>>>calculate based upon that the time, then your whole
>>>>thing is a big lie simply.
>>...bob explained: he had some original times, calculated the speedup, wrote that
>>down, and for presentation in the paper, he didnt print the real times, but the
>>original 1-processor time divided by the speedup. of course, if a student of
>>mine did anything like that with his numbers i'd give him a serious talking-to,
>>but if you accept the hypothesis that this is how he arrived at these numbers,
>>then it is just really bad style, but not a big lie.
>>
>>>>It is provable that all search times from 1-8 cpu's
>>>>at all tests are completely not true. they are about
>>>>a factor 2 too fast in order to let the 16 processor
>>>>look good.
>>how do you know that? if it is provable, then prove it...
>>please explain that more clearly! i can definitely understand that you see a
>>problem with the data in the table, but with all the rest, i don't understand
>>:-(
>>
>>aloha
>> martin
>
>Robert Hyatt Computer and Information Sciences
>hyatt@cis.uab.edu University of Alabama at Birmingham
>(205) 934-2213 115A Campbell Hall, UAB Station
>(205) 934-5473 FAX Birmingham, AL 35294-1170
>-------------------------
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.