speed of printing output by the interface.

[Uri Blass]

In one of the games of movei it played under interface that is not winboard
Movei printed the output that is at the end of this post(last line that was printed to the logfile showed that it used 0.02 seconds).

Movei used only 0.02 seconds based on it's logfile but I saw that in the game it used 4 seconds.

The time that it used based on the interface caused it to lose the game on time later.

I will probably change movei to print less information(for example not to print new pv unless the pv was changed or it searched enough nodes) but
the question is how much time do interfaces that are not winboard use for printing output.

I think that the time that they use for printing output may be significant factor for rating of programs in blitz under different interfaces.


1 98 0 58 d3d8
2 68 0 80 d3d8
2 56 0 118 d3d8 e2e3
2 57 0 155 d3g3
2 75 0 186 d3g3 a5a7
2 76 0 212 d3h3
2 75 0 215 d3g3 a5a7
2 76 0 241 d3c3
2 75 0 244 d3g3 a5a7
2 76 0 270 d3b3
2 75 0 273 d3g3 a5a7
2 76 0 325 f6e6
2 86 0 347 f6e6 a5a7
2 86 0 371 f6e6 a5a7
3 94 0 514 f6e6 a5a7 e6e5 a7e7 e5f4
3 94 0 631 f6e6 a5a7 e6e5 a7e7 e5f4
4 83 0 986 f6e6 a5a7 e6e5 a7e7 e5f4 e7f7 f4e5
4 84 0 1353 d3g3
4 85 1 1754 d3g3 a5a7 f6f5 a7f7 f5e5
4 86 1 2205 d3b3
4 85 1 2368 d3g3 a5a7 f6f5 a7f7 f5e5
4 85 1 2849 d3g3 a5a7 f6f5 a7f7 f5e5
5 0 1 2850 d3g3
5 0 1 2850 d3g3
5 0 1 2869 d3g3
6 0 1 2870 d3g3
6 0 1 2889 d3g3
7 0 1 2890 d3g3
7 0 1 2909 d3g3
8 0 1 2910 d3g3
8 0 1 2929 d3g3
9 0 1 2930 d3g3
9 0 1 2949 d3g3
10 0 1 2950 d3g3
10 0 1 2969 d3g3
11 0 1 2970 d3g3
11 0 1 2989 d3g3
12 0 1 2990 d3g3
12 0 1 3009 d3g3
13 0 1 3010 d3g3
13 0 1 3029 d3g3
14 0 1 3030 d3g3
14 0 1 3049 d3g3
15 0 1 3050 d3g3
15 0 1 3069 d3g3
16 0 1 3070 d3g3
16 0 1 3089 d3g3
17 0 1 3090 d3g3
17 0 1 3109 d3g3
18 0 1 3110 d3g3
18 0 1 3129 d3g3
19 0 1 3130 d3g3
19 0 1 3149 d3g3
20 0 1 3150 d3g3
20 0 1 3169 d3g3
21 0 1 3170 d3g3
21 0 1 3189 d3g3
22 0 1 3190 d3g3
22 0 1 3209 d3g3
23 0 1 3210 d3g3
23 0 1 3229 d3g3
24 0 1 3230 d3g3
24 0 1 3249 d3g3
25 0 1 3250 d3g3
25 0 1 3269 d3g3
26 0 1 3270 d3g3
26 0 1 3289 d3g3
27 0 1 3290 d3g3
27 0 1 3309 d3g3
28 0 1 3310 d3g3
28 0 1 3329 d3g3
29 0 1 3330 d3g3
29 0 1 3349 d3g3
30 0 1 3350 d3g3
30 0 1 3369 d3g3
31 0 1 3370 d3g3
31 0 1 3389 d3g3
32 0 1 3390 d3g3
32 0 1 3409 d3g3
33 0 1 3410 d3g3
33 0 1 3429 d3g3
34 0 1 3430 d3g3
34 0 1 3449 d3g3
35 0 1 3450 d3g3
35 0 1 3469 d3g3
36 0 1 3470 d3g3
36 0 1 3489 d3g3
37 0 1 3490 d3g3
37 0 1 3509 d3g3
38 0 1 3510 d3g3
38 0 1 3529 d3g3
39 0 1 3530 d3g3
39 0 1 3549 d3g3
40 0 1 3550 d3g3
40 0 1 3569 d3g3
41 0 1 3570 d3g3
41 0 1 3589 d3g3
42 0 1 3590 d3g3
42 0 1 3609 d3g3
43 0 1 3610 d3g3
43 0 1 3629 d3g3
44 0 1 3630 d3g3
44 0 1 3649 d3g3
45 0 1 3650 d3g3
45 0 1 3669 d3g3
46 0 1 3670 d3g3
46 0 1 3689 d3g3
47 0 1 3690 d3g3
47 0 1 3709 d3g3
48 0 1 3710 d3g3
48 0 1 3729 d3g3
49 0 1 3730 d3g3
49 0 1 3749 d3g3
50 0 1 3750 d3g3
50 0 1 3769 d3g3
51 0 1 3770 d3g3
51 0 1 3789 d3g3
52 0 1 3790 d3g3
52 0 1 3809 d3g3
53 0 1 3810 d3g3
53 0 1 3829 d3g3
54 0 1 3830 d3g3
54 0 1 3849 d3g3
55 0 1 3850 d3g3
55 0 1 3869 d3g3
56 0 1 3870 d3g3
56 0 1 3889 d3g3
57 0 1 3890 d3g3
57 0 1 3909 d3g3
58 0 1 3910 d3g3
58 0 1 3929 d3g3
59 0 1 3930 d3g3
59 0 1 3949 d3g3
60 0 1 3950 d3g3
60 0 1 3969 d3g3
61 0 1 3970 d3g3
61 0 1 3989 d3g3
62 0 1 3990 d3g3
62 0 1 4009 d3g3
63 0 1 4010 d3g3
63 0 1 4029 d3g3

0.00/63 time= 0.02 nps= 251812 d3g3

[Naum]

First the basic question. Do you flush the output?

If you do, then it's a GUI bug. Which GUI?

Naum actually measures GUI delay and adjusts its time management based on it. If I remember correctly, delay for Winboard and Arena is very small, but for the Fritz8 GUI can be large (up to 1sec).
4sec delay is too large. It indicates a bug somewhere.

Alex

[Uri Blass]

First the basic question. Do you flush the output?

If you do, then it's a GUI bug. Which GUI?

Naum actually measures GUI delay and adjusts its time management based on it. If I remember correctly, delay for Winboard and Arena is very small, but for the Fritz8 GUI can be large (up to 1sec).
4sec delay is too large. It indicates a bug somewhere.

Alex

I do not know the gui
This game is one of the games of the CEGT blitz games and I understood that 2 gui were used for it (fritz and arena at the same time)

from the email that I got from Wilhelm Hudetz:

"Movei run in two interfaces. Chessbase an Arena, on an AMD X2 4400 dual."

4 seconds is simply written in the pgn that I download when the time in the logfile is 0.02 seconds.
The delay on my fritz8 gui seems to be 1 second inspite of having only AMD 3000 but maybe there was bigger delay when you use dual and try to play 2 matches with ponder off at the same time.

I cannot test it because I have not a dual.
Note that Movei's logfile showed that Movei played 2 games on the same time and I asked wilhelm to put the exe file that run in chessbase gui and the exe file that run in arena gui in different folders in the future in order to have better logfile because I do not like to see a logfile that talk about 2 games at the same time.

Wilhelm told me that he is going to do it.

I do not think that with a dual using 2 exe file of movei on the same time
for ponder off games should be a problem and I assume that 2 different exe file should use the different processors but I am not sure about it because I can imagine a problem if 2 exe files use the same processor at the same time.

Of course if one processor use the programs that Fritz ask to use and one processor use the programs that arena ask to use there is no problem in the games but I can imagine the following problem(maybe it cannot happen but I am not sure about it because I know nothing about dual and how they are used but I will describe it):

there are 2 games:

a against b
c against d

a to move use processor 1
c to move use processor 2

c play a move and now it is d to move
c does not stop immediatly to use processor 2 and write some information to the logfile.

It takes less than 0.01 second but d at the same time
needs to choose a processor when both processors are busy.

d "thinks" that processor 2 is busy so
d choose the wrong processor (processor 1) and now both a and d use it at the same time.

Uri

[Naum]

You can verify if something like that is happening if you check depth and time info.

I don't think it's a good idea to use a dual machine in that way (2 GUIs running at the same time). Esspecialy if you use the Chessbase GUIs which is so big and unpredictable.
I guess that 4sec delay in such a setup is quite possible and may impact results a lot. Esspecialy for blitz games.