Author: Uri Blass
Date: 12:43:24 01/07/03
Go up one level in this thread
On January 07, 2003 at 14:11:21, Miguel A. Ballicora wrote:
>On January 07, 2003 at 12:18:13, Uri Blass wrote:
>
>>On January 07, 2003 at 11:36:03, Uri Blass wrote:
>>
>>>On January 07, 2003 at 10:48:54, Dieter Buerssner wrote:
>>>
>>>>On January 07, 2003 at 10:41:18, Uri Blass wrote:
>>>>
>>>>>On January 07, 2003 at 09:31:35, Daniel Clausen wrote:
>>>>>
>>>>>>On January 07, 2003 at 09:16:21, Uri Blass wrote:
>>>>>>
>>>>>>>Here are some kind of remarks:
>>>>>>>
>>>>>>>C:\Program Files\Microsoft Visual Studio\VC98\INCLUDE\stdio.h(97): remark #344:
>>>>>>>typedef name has already been declared (with same type)
>>>>>>> typedef char * va_list;
>>>>>>
>>>>>>[snip]
>>>>>>
>>>>>>My guess is that the Intel compiler is more ANSI-compliant than the M$-compiler
>>>>>>and therefore doesn't like the M$-system header files too much.
>>>>
>>>>Yes. You just have to ignore those warnings.
>>>>
>>>>
>>>>>>>2)I get a lot of remarks for movei and the most popular remark is
>>>>>>>
>>>>>>>external definition with no prior declaration
>>>>>>
>>>>>>Not quite sure what the compiler means with that.. but maybe you call a function
>>>>>>XYZ and the compiler didn't see the declaration before of this function before
>>>>>>that. (like forgot to include the header-file where the function would/should be
>>>>>>declared)
>>>>>
>>>>>Is there a reason that I need to declare that function.
>>>>
>>>>>All calls for the functions are done after the function and in the same file.
>>>>
>>>>In this case, everything is ok. You might need to change
>>>>
>>>>int foo()
>>>>{ /* */ }
>>>>
>>>>to
>>>>
>>>>int foo(void)
>>>>{ /* */ }
>>>>
>>>>For the unspecified order of evaluation, you might want to show an example.
>>>>
>>>>Regards,
>>>>Dieter
>>>
>>>I got that warning in the following function that is supposed to give the change
>>>in the evaluation by a move.
>>>
>>>int evalmove(move_bytes m)
>>>{
>>> return evalmovewithoutpawns(m)+evalpawnchange(m);
>>>}
>>>
>>>evalmovewithoutpawns gives the change in the evaluation from the piece square
>>>table when evalpawnchange gives the change in the evaluation from the change in
>>>the pawn structure.
>>>
>>>Uri
>>
>>Another thing that I do not like is that some tests that I do suggest that
>>movei performs sligthly worse at very fast time control(1-10 seconds per game)
>>
>>I suspect that it is possible that the intel compiler made movei faster at slow
>>time control but slower at super bullet.
>>
>>I think that I will use the old compiler because the improvement from the intel
>>compiler seems to be minimal and I do not like the big number of warnings that I
>>get that may hide important warning.
>
>Once in a while, it is a VERY good idea to use all the compilers that you have
>and make sure that all the binaries produced work in the same way (i.e. they
>give the same amount of nodes etc.). You do not need to _choose_ one compiler
>and erase the rest from your computer.
>
>Miguel
The problem is that the intel compiler is only for evaluation until the end of
january.
I also suspect that the fact that some changes in my computer made the same
movei relatively weaker at very fast time control(I downloaded a big file of
Service pack 5 for microsoft visual C++ and maybe the big file make movei slower
only in the first milisecond).
I now tested a previous version that I liked it's results and it produced worse
results so it is possible that some change in my computer make movei slower at
very fast time control or maybe the computer at very fast time control does not
give the same cpu time for both programs and last time movei was lucky and today
movei is unlucky.
Statistical error is also a possibility but based on experience movei and Gerbil
are relatively deterministic even at very fast time control(I still test against
Gerbil at fast time control).
I remember that sune claimed that even with no book games do not happen twice at
1,2,3 second per game but it is not my experience and if I repeat a match of 40
games when movei always changes the first move games happen twice and even if
they do not happen exactly the same they may be often almost the same.
Uri
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.