New Architectures in Computer Chess

[Gerd Isenberg]

Congrats to Fritz Reul for his Ph.D. Thesis New Architectures in Computer Chess, which mainly refers to his innovative mailbox 8*8 board structure with a 15*12 board array, and doubly linked piece list for disjoint pieces, even light and dark bishops. Fritz also covers magic bitboards and SEE. He mentiones f.i. that using the homogeneous ~2MB array (Lasse Hansen's initial approach) with constant shifts (64-12) versus the denser Pradu's 800K, with individual sizes for each square but slightly longer code was about the same speed inside his chess program Loop Amsterdam. Reul's thesis deserves a recommendation for each computer chess programmer and people interested in computer chess programming.

Gerd

[mcostalba]

I cannot download

[Gerd Isenberg]

You cannot download the thesis, only toc and summary. You need to contact the author to order the book. Protective charge is 10€.

[mcostalba]

You cannot download the thesis, only toc and summary. You need to contact the author to order the book. Protective charge is 10€.

This deeply decreases my interest in the book

Sorry but I don't have any "Protective charge" to apply to any of my stuff to compensate that.

Thanks anyway
Marco

[Gerd Isenberg]

You cannot download the thesis, only toc and summary. You need to contact the author to order the book. Protective charge is 10€.

This deeply decreases my interest in the book

Sorry but I don't have any "Protective charge" to apply to any of my stuff to compensate that.

Thanks anyway
Marco

Your choice. I mean to print a book and to ship it, 10€ (at least inside Germany) is not that expensive and prime cost. It is absolutely worth the money. I don't receive any provision btw.

Gerd

[mcostalba]

Your choice. I mean to print a book and to ship it, 10€ (at least inside Germany) is not that expensive and prime cost. It is absolutely worth the money. I don't receive any provision btw.

Gerd

Does exist also email and pdf (BTW I assume there exsist a pdf version because the summary, contents, etc. are in pdf)

Of course I don't pretend to have a real book in paper and shipment box for free, I just would like to _read_ the thesis for free.

BTW mine and that of _all_ my university students friends was free and I never heard of someone that asks money to make people reading his thesis (normally happens the contrary )

Marco

[smrf]

This book has been sent to me this week. It is interesting to see, that an architecture using two double linked lists has found more implementations now very prominent in Reul's Loop. Using this within a post box related geometrie indeed is the superior approach at least to build a very fast move generator.

[Gerd Isenberg]

This book has been sent to me this week. It is interesting to see, that an architecture using two double linked lists has found more implementations now very prominent in Reul's Loop. Using this within a post box related geometrie indeed is the superior approach at least to build a very fast move generator.

Hi Reinhard,
it is clearly the more versatile board representation, if it is about other board sizes and any none 64-bit architecture, like 16-bit palm or wii station. Was the doubled linked piece list your invention? I am also interested in what you meant with monochrome algorithm recently in CSS forum (I don't like to post there, since I found all my (rare) old posts, no longer accessible without user account). Do you mean to flip the board and side to move after make/unmake?

Thanks,
Gerd

[smrf]

... Was the doubled linked piece list your invention?

SMIRF uses it from its beginning. I have mentioned that detail here sometimes. Though I have not seen that approach before, I am unsure whether possible parallel activities have existed. Another implementation detail in SMIRF has been, that both links combined with bit encoded piece properties do fit into one single integer.

I am also interested in what you meant with monochrome algorithm recently in CSS forum (I don't like to post there, since I found all my (rare) old posts, no longer accessible without user account). Do you mean to flip the board and side to move after make/unmake?

The idea is to no longer encode color within piece codes. It is about to use two boards, one active one passiv, using only the piece types: P, N, B, R, Q, K and Enemy. An additional horizontal mirroring moreover could avoid any distinction between the different pawn moving directions. Thus any procedure would be independent from the current color. An additional flipping within the position's hash code also would make stored information reusable for positions having changed colors.

[Dan Andersson]

I'm pretty pleased with Wii Chess. If only the Thesis was part of a promotional offering

MvH Dan Andersson