Bebe, (Be-Be)
a chess program running on a customized, special purpose bit slicehardware developed by Tony and Linda Scherzer as testbed for their company's SYS-10, Inc. mini- and mainframe processors [1] . It was originally written on a Z80 basis, as BB-1 already standby at the ACM 1978[2][3], and was ported to run on a SYS-10 computer [4] , based on AMD 2900 bit slicers [5] .
After Tony Scherzer's early death in January 1995 [8] , the work on Bebe discontinued, while SYS-10, Inc. continued, and is now a company providing custom software applications located in South Elgin, Illinois[9] , owned by Linda Scherzer since 1978 [10] .
At ACM 1986, after winning ChipTest, Merlin and Recom, Bebe lost to Belle, the eventual winner. In the battle for the second place, Bebe was defeated by Lachex. Both programs went in a sequence of forced moves where both sides had only one single good choice along the way. Neither program had any idea about who would come up ahead in the end, until suddenly after playing along the forced line, both programs realized that Bebe was lost. In a discussion with Feng-hsiung Hsu and others, Tony Scherzer made the statement that the old idea of selective pruning was dead, replaced by the new idea of selective extensions . According to Hsu, that was how the idea of Singular Extensions began [15] .
A relatively new program and machine, Bebe has recently acquired a provisional UCSF rating of 1810 based on play in one tournament. Bebe defeated an Expert in that tournament. Tony Scherzer's brainchild examines 10,000 nodes/sec or about 2,000,000 in a three minute move. The program is small, requiring only 10k 16 bit words. The program has no book. It uses iterative deepening and is written in assembly language.
ACM 1986[20] : Tony Scherzer and Linda Scherzer, SYS-10 Inc., Hoffman Estates, Illinois, Bebe Chess Engine 40K Nodes/Sec
1989
from the WCCC 1989 booklet [21] : Tony Scherzer and Linda Scherzer, SYS-10 Inc., Hoffman Estates, Illinois, Bebe Chess Engine, 40K Nodes/Sec
In early 1980 SYS-10 tried new hardware techniques needed for their mini/mainframe processor in co-processors for BeBe's CPU. Each co-processor takes over a specific function from the main CPU.
The first co-processor does the complete task of move list generation. The actual unit is divided into two processors which function in parallel: one that finds pieces and one that calculate and stores moves. This parallelism provides results more than 25 times faster than software.
A second co-processor performs the position scoring function. The scorer "looks at" the output of the move generator and uses the moves to calculate values for piece position mobility and co-operation. The scorer functions in parallel with the move generator.
The co-processors perform move list generation, and some of the board scoring functions.
The self-activated parallel processor determines of either king is in check and determines the attack-defender count for any square. Because it self starts, the answers for both kings are ready before the software can ask the question.
^ Linda Scherzer (1984): "We justify this as our research and development department. We try out the firmware in Bebe, then design it for commercial products".
^Feng-hsiung Hsu (2002). Behind Deep Blue: Building the Computer that Defeated the World Chess Champion, Princeton University Press, ISBN 0-691-09065-3, pp. 54-55
a chess program running on a customized, special purpose bit slice hardware developed by Tony and Linda Scherzer as testbed for their company's SYS-10, Inc. mini- and mainframe processors [1] . It was originally written on a Z80 basis, as BB-1 already standby at the ACM 1978 [2] [3], and was ported to run on a SYS-10 computer [4] , based on AMD 2900 bit slicers [5] .
While the hardware did no longer improve significantly since the mid 80s, the software and opening book did. Bebe approached parallelism in generating attack and defend maps, move generation, move ordering, and evaluation, but otherwise had a serial search. Its implementation of Iterative deepening was quite unique. To avoid the odd-even effect, Bebe searched only odd depths with an increment of two plies at the root [6] . It further introduced a learning approach utilizing a persistent hash table, explained in award winning paper [7].
After Tony Scherzer's early death in January 1995 [8] , the work on Bebe discontinued, while SYS-10, Inc. continued, and is now a company providing custom software applications located in South Elgin, Illinois [9] , owned by Linda Scherzer since 1978 [10] .
Table of Contents
Photos & Games
Achievements
Bebe was one of the strongest chess entities in the 80s. At its first appearance, the WCCC 1980 in Linz, it finished best placed "Micro" [13] . Four draws in four rounds, playing programs like Chess 4.9 and Nuchess. The following three World Computer Chess Championships were even more successful. Bebe was first runner up at the WCCC 1983 in New York, tied for first place at the WCCC 1986 in Cologne, and finally was runner up again in Edmonton at the WCCC 1989. As mentioned by Robert Hyatt, Bebe "helped" Cray Blitz to win both the WCCC 1983 and the WCCC 1986 [14] .Bebe further played 13 consecutive ACM North American Computer Chess Championships, including the already mentioned WCCC in New York, from 1980 until 1993, again runner up in 1981 (tied), 1984 and 1985.
Selective Deepening
At ACM 1986, after winning ChipTest, Merlin and Recom, Bebe lost to Belle, the eventual winner. In the battle for the second place, Bebe was defeated by Lachex. Both programs went in a sequence of forced moves where both sides had only one single good choice along the way. Neither program had any idea about who would come up ahead in the end, until suddenly after playing along the forced line, both programs realized that Bebe was lost. In a discussion with Feng-hsiung Hsu and others, Tony Scherzer made the statement that the old idea of selective pruning was dead, replaced by the new idea of selective extensions . According to Hsu, that was how the idea of Singular Extensions began [15] .Descriptions
1980
from the ACM 1980 booklet [16] :Tony Scherzer, SYS-10, Inc., Hoffman Estates, Illinois, Bebe Chess Machine on site (32K bytes, 16 bits, 6,250,000 inst/sec)
1981-1986
1989
from the WCCC 1989 booklet [21] :Tony Scherzer and Linda Scherzer, SYS-10 Inc., Hoffman Estates, Illinois, Bebe Chess Engine, 40K Nodes/Sec
The first co-processor does the complete task of move list generation. The actual unit is divided into two processors which function in parallel: one that finds pieces and one that calculate and stores moves. This parallelism provides results more than 25 times faster than software.
A second co-processor performs the position scoring function. The scorer "looks at" the output of the move generator and uses the moves to calculate values for piece position mobility and co-operation. The scorer functions in parallel with the move generator.
BeBe operates at four distinct levels:
Mephisto Best-Publication Award
The fourth Mephisto Best-Publication Award for publications between April 01, 1990 and March 31, 1991 is awarded to Tony and Linda Scherzer and Dean Tjaden for their article Learning in Bebe published in Computers, Chess, and Cognition [22], explaining the implementation of a persistent hash table inside BeBe. The jury consisted of Jaap van den Herik, David Levy, Tony Marsland, Jonathan Schaeffer and Ken Thompson. The image by Jaap van den Herik from the Award giving ceremony during ACM 1991, November 20, 1991, Albuquerque, New Mexico, published in ICCA Journal, Vol. 14, No. 4 [23], the same issue also containes a condensed version of the paper [24] .See also
Publications
Forum Posts
External Links
Chess Program
Misc
References
What links here?
Up one level