Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: TSCP nps test (Mac drools so far)...

Author: Dan Newman

Date: 15:27:02 07/24/99

Go up one level in this thread


On July 23, 1999 at 19:26:02, Dann Corbit wrote:

>On July 23, 1999 at 19:18:36, Dan Newman wrote:
>
>>Here's what I get with Dann's Win32 binary on a P6/200:
>>
>>tscp> on
>>ply      nodes  score  pv
>>  1         21     48  d2d4
>>  2         93      0  d2d4 d7d5
>>  3        996     35  d2d4 d7d5 b1c3
>>  4       5952      5  e2e4 d7d5 f1b5 c8d7 b5d3
>>  5      32924     35  e2e4 b8c6 b1c3 e7e5 g1f3
>>  6     203273     13  e2e4 b8c6 d2d4 d7d5 e4d5 d8d5
>>time: 8.620000 sec., nodes=203273, NPS = 23581.554524
>>tscp>
>>
>>(which is quite a bit better than the 16knps I got with Watcom...)
>
>Keep in mind that I'm a dirty, rotten cheater.  I created a source file which
>#includes all of the original source files.  I told the compiler to inline
>whenever it felt like it.  With a tiny source code like this, that tends to be a
>very good idea.  I also targeted the P6 chip instruction set, which gives a big
>boost.  Similar options with other compilers may give similar results.
>
>OTOH, the Microsoft compiler does emit great binaries.  You might try to
>recompile with maximum optimization, but give it TSCP.C instead of the separate
>source files.  I bet your compiler will nudge closer to the better mark.

I just tried that and got an extra 200 or so nps.  I then told the (Watcom 11.0)
compiler to inline and got about 15.5 knps, so I told it to massively inline
(which, with the optimization on, took several minutes) and got 15.2 knps...
bad trend.

I don't think Watcom does P6 (at least I don't see any switches for that) and
it looks as if it never will -- I just got a letter yesterday from Sybase saying
that they are discontinuing development on Watcom :(.  (They say it's "fully"
developed, IIRC.)

What compiler are you using?  I assume MSVC 6.0.  (I have 5.0 and it seems to be
about 15% faster than Watcom 11.0 on integer stuff and about 15% slower on
floating point.)  Is 6.0 a lot better than 5.0?  If it is, I think I'll upgrade.

-Dan.



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.