5-1 EGTB's

[vb4]

Hi John,

Did I hear you say this has got to stop?? No not yet please <S>. Thanks for all your work regarding the completion of the 5-1 egtb's. I see your desire to eventually start working on the 7 egtb's!! I cant imagine how long that may take or the size of it! Do you have a generator that you believe will be able to generate these faster then what has been done in the past? Just curious. Anyone who even whispers a suggestion of working on those gets a lot of attention around here <S>. Look forward to continuing updates John.

Les

[jkominek]

I wasn't announcing my retirement quite yet. Just expressing my desire that loading of the tables take less time, as well as much less memory for the decompression data structures. (And hinting that a unified API might prove problematic.)

Speedwise, tbgen2 is a little faster than tbgen, considerably faster than DeKoning and Edwards, and considerably slower than Muller. At the impressively high rate of solving 1 million positions per second -- an order of magnitude faster than what tbgen2 achieves -- generating the 7-man tables will take 20 years. Cutting-edge tablebase work has a long history of demanding parallel computation.

john

[vb4]

Hi John,

Just wondering how things are going with your generation of the remaining 5-1 egtb's and when you think they might be completed. Are they presenting any difficulties? Just curious.

Thanks,

Les

[jkominek]

Hi,

Unfortunately I have had to set aside chess activities for the past month and a half. My regular job has been requiring extra time, including weekends. I'd actually been intending to post a short note of apology here, but procrastination won the day. I'm glad that you asked though. I don't mind the "just wondering" inquiries.

john

[vb4]

Thanks John for he reply,

Les

[gambit3]

*sticks up hand*

ooo! ooo! mr. kotter! mr. kotter!

being a rather late re-entrant into this nature of discussion (last time was two years and a rather interesting online debate with guyhaw ago), i wouldn't be able to offer any suggestions, having had basically NO time to program of late, let alone look at the tbgen source again.

i CAN, however, offer complete 7 man fegs (short only 6v1 cause feg can only do 5 men per side) and with your source, or better yet an x64 .exe, can attempt to compile 5v1s nalimov style on a 3G athlon with only 1.5G RAM :>

i like a challenge that'll overload my computer rather than me :> just another exabyte of storage to buy when attempting to generate 7 man nalimovs.....