Now with "cut optimization". I'm still unsure if it cuts "all there was to
cut". Again only marginal improvement and even some overhead. :-(
TrG:0:TranslateGameExtra:0:connect5... tested connect5 in 1.715772 seconds
ok
TrG:0:TranslateGameExtra:1:connect4... tested connect4 in 1.254531 seconds
ok
TrG:0:TranslateGameExtra:2:pawn_whopping... tested pawn_whopping in
3.184812 seconds
ok
TrG:0:TranslateGameExtra:3:breakthrough... tested breakthrough in 27.095218
seconds
ok
TrG:0:TranslateGameExtra:4:pacman3p... FAIL
TrG:0:TranslateGameExtra:5:chinesecheckers3... tested chinesecheckers3 in
274.368936 seconds
ok
TrG:0:TranslateGameExtra:6:asteroids-scrambled... tested
asteroids-scrambled in 22.944988 seconds
ok
On Thu, May 23, 2013 at 11:38 PM, Lukasz Stafiniak <luk...@gm...>wrote:
> I've implemented path indexing but it helps only slightly, at best by 1/3
> and generally below 1/10 speedup.
>
> TrG:0:TranslateGameExtra:0:connect5... tested connect5 in 1.787425 seconds
> ok
> TrG:0:TranslateGameExtra:1:connect4... tested connect4 in 1.272966 seconds
> ok
> TrG:0:TranslateGameExtra:2:pawn_whopping... tested pawn_whopping in
> 3.238722 seconds
> ok
> TrG:0:TranslateGameExtra:3:breakthrough... tested breakthrough in
> 30.075195 seconds
> ok
> TrG:0:TranslateGameExtra:4:pacman3p... FAIL
> TrG:0:TranslateGameExtra:5:chinesecheckers3... tested chinesecheckers3 in
> 271.586825 seconds
> ok
> TrG:0:TranslateGameExtra:6:asteroids-scrambled... tested
> asteroids-scrambled in 24.478126 seconds
> ok
>
>
>
> On Sun, May 19, 2013 at 9:22 PM, Lukasz Stafiniak <luk...@gm...>wrote:
>
>> Hi,
>>
>> We now have optimized clause reordering, and memoization (i.e. cache) in
>> Prolog interpreter (no optimizations yet in the saturation-absed engine).
>> Unfortunately, both techniques sometimes make things worse rather than
>> better, but in no case so far prohibitively worse. Even more unfortunately,
>> memoization does not help enough in case of "amazons.gdl".
>>
>> Cheers.
>>
>
>
|
