Looking for partner?

[BubbaTough]

Howdy,

When I started LearningLemming a couple of years ago my plan was:

1. write a reasonably strong 1 cpu version (hopefully near Crafty 1cpu level, which was the gold standard when I started playing with this stuff)
2. confirm strength via independent testing
3. add multiprocessor support
4. release a public version
5. write a couple papers describing the ideas/techniques used
6. pick a new project

I was not sure I would ever get past stage 1, but I think I am just about there and am embarking on step 2 with the help of CCRL, which brings me to step 3.

With the impending birth of my first child, however, it looks like I may not have enough time to do a responsible job adding multiprocessor support. To address this, I am considering taking on a co-author, if there is an experienced person interested in taking on this role. If this is the case, please message or email me and we can discuss. A few relevant notes:

a) I am not interested in a "quick and dirty" implementation, well, "dirty" is ok but it must be effective/efficient.
b) I have no plans to ever do anything commercial, even if the program becomes unexpectedly successful.
c) as primary author, I would prefer to maintain version control, as well as be the primary decision maker regarding if and when source code is made public.
d) The current source code is very very ugly C code with lots of bugs. If this scares you don't volunteer.
e) I am not committed to this path yet, so don't feel bad if you make a generous offer and I don't take you up on it.
f) if the author of an existing program becomes co-author of learninglemming, and wants to try ideas from learninglemming in their program, this is fine with me as long as I am given credit as a coauthor, have access to your source code, and the relationship is symmetrical. It is even fine for you to cut and paste my code into your program (within reason). This is true even for authors who make their source code publicly available.
g) I want to avoid the situation where I take on a coauthor of an existing program who ends up spending all their time trying ideas from learninglemming in their own program instead of working on adding multi-processor support to learninglemming. For this reason, only authors of programs with existing multi-processor support will be considered, and the co-author will be expected to initially spend a majority of their chess programming time on LearningLemming. People that have demonstrated expertise in parallel programming outside the domain of chess and have never programmed computer chess would also be considered.

Anyway, given all this, I doubt any will volunteer, but it doesn't hurt to ask .

-Sam

[Edsel Apostol]

Hi Sam,

I am volunteering, but just like what I have said on the other thread that I will first implement SMP in my engine before I could start working with yours.

I am now in the process of studying SMP algorithms. After the CCT, Zach Wegner, a participant also there with his engine ZCT, will release its source code, according to him. I am awaitng that release as it has a real DTS support.

He also help add SMP support for Toga but I'm not sure if it is the same Toga as the latest 1.4beta5c SMP.

Maybe you could recruit him too to work with Learning Lemming.

[BubbaTough]

That is good news. I look forward to seeing how things go with adding SMP to your own engine. I am also glad to see you are still interested given all the details listed in this post. Of course, neither of us are commited to anything right now, but I certainly find the prospect of LearningLemming continuing to make progress even after my time for work on it drops near zero.

-Sam