Author: Robert Hyatt
Date: 08:24:36 05/07/04
Go up one level in this thread
On May 07, 2004 at 04:13:57, martin fierz wrote: >On May 06, 2004 at 20:02:50, Robert Hyatt wrote: > >>On May 06, 2004 at 19:03:48, martin fierz wrote: >> >>>aloha! >>> >>>bob posted some crafty logfiles running a 24-position test set on his ftp site >>>(for anyone else crazy enough to repeat what i did: >>>ftp.cis.uab.edu/pub/hyatt/smpdata) >>> >>>these are logfiles of crafty running as single CPU, dual, or quad; on opterons. >>>i took the last completed ply on the single CPU set for each position (marked by >>>-> in the logfile, i hope...), wrote down the time to complete this ply, and did >>>this for all logfiles. there are 9 of these, 4 repeats for 2 and 4 CPUs. i >>>computed the speedup for time-to-finish-ply-X for each of the multi-CPU runs >>>with the following results: >>> >>>2 CPUs: >>>1.961 +- 0.093 >>>1.888 +- 0.074 >>>1.846 +- 0.078 >>>1.763 +- 0.084 >>> >>>4 CPUs: >>>3.15 +- 0.15 >>>3.29 +- 0.20 >>>3.06 +- 0.12 >>>3.19 +- 0.13 >>> >> >>That is higher than my number although I only checked log 1 for mt=4. Did you >>compute the speedup for each position, then add and divide by 24? > >yes, i did. > >> If so, I'm >>not a fan of that way. A long search on an efficient position skews the >>results. I prefer to take the total time for each run and use that... > >i'm not a fan of doing that, because in that case, some weird positions don't >produce PVs after 1 minute, while others produce a PV after close to 5 minutes, >and one position gets more weight than others... >i don't quite understand your point, because the length of a run on an efficient >position does not matter in my way of computing things. THink about what is happening. If one position takes 4 minutes, another takes 1, you are counting them equally. In a game the one that takes 4 minutes is more important to the speedup because it takes 4x as much of the total game time as the 1 minute position. As I said, either way is ok. They weight different positions differently. Your way says each position is "equal". My way says longer positions are "more equal"... :) > >i was interested in the speedup as such per position, to see how variable it is >for any given position. to see that, i have to do it this way. you lose this >information when you lump everything together, and have no means to calculate a >variability in the end... or none that i can think of within a few seconds while >i'm typing this :-) I agree for looking at the variability issue... > >cheers > martin
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.