Question about Program called Belle

[Carey]

Certainly the DT/DB guys gave Ken credit. In fact Ken was heavily involved in the original chiptest development. I had long conversations with him the first time I saw chiptest play in 1987 (I believe). He had a theory that going too deep would lead to very passive play since you could see all sorts of threats and try to parry them, when your opponent would not really see any of them. But in any case, I have seen Belle cited in their publications and mentioned in their discussions, even if the "press" failed to make mention of it.

It's also mentioned in Hsu's book. I just re-read that chapter about it. Hsu did have several bright ideas even for the initial chiptest that went beyond Belle.

He even spontaneously asked Ken Thompson to be on his thesis committee. Pretty gutsy.

Yet it is the popular press that usually wins in the long run. For example, Henry Ford is often credited with inventing the automobile,

Never heard that before. It was always he invented the afordable auto.

or at least the automobile assembly line. In truth, he did neither. The

That I've heard, although I certainly know it's not true. Assembly lines existed before that.

He just had the idea to do it with a car and to do it on such a massive scale as to make it afordable. Take it from being a work of art from a mechanical craftsman to just simply a product of the industrial revolution.

And he let people buy on credit, which was a major inovation at the time. Without that, people wouldn't have bought his cars no matter how many he made.

And then he had mechanics go around to junk yards and see which parts failed, and those that were still in good condition, he made cheaper and more fragile with the goal that the whole car would break all at once when it was ready to be replaced.

one original idea that Ford did have was the one he got while touring a Swift Meatpacking Company pork processing facility. Ford watched hogs being turned into lunchmeat in Swift's disassembly line and got the brainstorm that the process could be run in reverse.

[Steelman]

I have a copy of "Computer Chess II" by David Walsh & Boris Baczynskyj which has many of Belle's games annotated. As well as all the games from the 1983 World Championship in which Cray Blitz won. Maybe you could find an old copy. "Computer Chess" by David Walsh has many of Belle's games listed too but I like Computer Chess II the best.

For it's day Belle was the software/hardware combination to beat.
According to this book Belle was searching at 110,000 moves per second and had an opening book of 375,000 positions. Very impressive!

And I believe Ken Thompson had to pay a $800 fine to get Belle back from the State Dept. after it was inpounded. Ken must have been pissed about that.

It's too bad that Computergames.com only has 11 of Belle's games.
And I did not know that in 1990 Belle played a game against Deep Thought. Belle must have taken out and been "dusted off" for that one? Was this the last time Belle played? Does anyone have more info on this?

[Terry McCracken]

Cray Blitz - Belle was most definitely 1-0, that was the year we won our first WCCC.

Hi Bob,

I know the hardware for Cray Blitz was a Cray-1 supercomputer in 1982 geting 20.000 positions per second.

Was the hardware any different in 1983?

best regards

Murat :.

Yes. In mid-Summer, 1983, Cray built the prototype for the first dual-CPU machine, called the Cray-XMP2. We used that prototype in the WCCC that year and developed the first parallel version of Cray Blitz for that machine. We were searching something like 20K nodes per second by then, which was about 4x faster than Belle. Note that the 20K is an estimate. When I get to the office, I have an old file folder with various test results for most versions, so I can flip through that to see if I can find a more accurate NPS. Unfortunately most of it is on the old Texas Instruments thermal paper, which is not known for remaining readable for long periods of time.

Bob, I have an old computer textbook and it claimed that Belle could do 30K per sec.

Was that when you were down to very few pieces in an endgame, or basic ending, probably no Queens or Rooks?

Curious Regards,
Terry

Actually your old book is wrong. As was I. The original belle (1977) did about 1K, the next generation (1978) did 5K with hardware move generation and evaluation, software search. The 1980 version did 160K. And CB was doing 20K so we were 8x slower, but we were significantly more efficient. Belle could only use a very simple move ordering in the hardware, no killers, no hash move, just MVV/LVA ordered captures, then the rest of the moves. It was doing 6 plies in hardware, 2 in software. we were actually outsiearching belle by one ply in 1983 at the WCCC.

So sorry for that 5K error. That was the previous version. The belle in the picture at the top of this thread was the last version, and is at the Smithsonian. That was the 160K 8-ply version.

Thanks Bob! I'll have to find that book as I haven't read it in nearly 20 years. It was printed in '83.

Now that I think back it may have said 130K!!

[mike860]

This is the best thread I've seen on this forum in some time.

[bob]

Cray Blitz - Belle was most definitely 1-0, that was the year we won our first WCCC.

Hi Bob,

I know the hardware for Cray Blitz was a Cray-1 supercomputer in 1982 geting 20.000 positions per second.

Was the hardware any different in 1983?

best regards

Murat :.

Yes. In mid-Summer, 1983, Cray built the prototype for the first dual-CPU machine, called the Cray-XMP2. We used that prototype in the WCCC that year and developed the first parallel version of Cray Blitz for that machine. We were searching something like 20K nodes per second by then, which was about 4x faster than Belle. Note that the 20K is an estimate. When I get to the office, I have an old file folder with various test results for most versions, so I can flip through that to see if I can find a more accurate NPS. Unfortunately most of it is on the old Texas Instruments thermal paper, which is not known for remaining readable for long periods of time.

Bob, I have an old computer textbook and it claimed that Belle could do 30K per sec.

Was that when you were down to very few pieces in an endgame, or basic ending, probably no Queens or Rooks?

Curious Regards,
Terry

Actually your old book is wrong. As was I. The original belle (1977) did about 1K, the next generation (1978) did 5K with hardware move generation and evaluation, software search. The 1980 version did 160K. And CB was doing 20K so we were 8x slower, but we were significantly more efficient. Belle could only use a very simple move ordering in the hardware, no killers, no hash move, just MVV/LVA ordered captures, then the rest of the moves. It was doing 6 plies in hardware, 2 in software. we were actually outsiearching belle by one ply in 1983 at the WCCC.

So sorry for that 5K error. That was the previous version. The belle in the picture at the top of this thread was the last version, and is at the Smithsonian. That was the 160K 8-ply version.

Thanks Bob! I'll have to find that book as I haven't read it in nearly 20 years. It was printed in '83.

Now that I think back it may have said 130K!!

Belle didn't count nodes, so they could not display node counts or NPS on the terminal, so Joe rigged up an integrating counter that essentially counted pulses per second. He hot-wired it to the chess hardware clock, as each major tick was one node. It was interesting to watch when he first hooked it up as nobody had gone that fast. But it didn't take long. In 84 we were doing 80K, by 85 we were doing 160K or so, and by 86 we had passed him although chiptest was just being built and took the NPS battle to a whole new level.

[sje]

Belle hardware had a piece count register that had bit fields assigned for each piece color and kind. To save a little on costs and make the design easier, the bit field for the queen count of either color was only two bits wide. This worked fine if and until the very rare event occurred that either side got a fourth queen anywhere in the search tree. That new fourth queen of a given color would appear on the board, but the material score for that side would drop by 27 pawns because of counter overflow.

[sje]

Belle could only use a very simple move ordering in the hardware, no killers, no hash move, just MVV/LVA ordered captures, then the rest of the moves.

I seem to recall that the final Belle hardware did have an integral transposition table; one MB of RAM. It was used both for move ordering and score adjustment along with repetition detection.

[sje]

And this reminds me of the old Multics operating system for the pdp10; it had a GMT time zone hour delta field that was only four bits wide. This worked fine until someone tried to install Multics at a Hawaiian site that was too close to the International Date Line.

And at university that will remain nameless, the local OS development group ran into problems because the major/minor OS version bit field was only 12 bits wide. The world would end if they tried to push things past version 63/63.

[Carey]

Belle didn't count nodes, so they could not display node counts or NPS on the terminal, so Joe rigged up an integrating counter that essentially counted pulses per second. He hot-wired it to the chess hardware clock, as each major tick was one node. It was interesting to watch when he first hooked it up as nobody had gone that fast.

Interesting bit of trivia.

Bob, I've mentioned this to you before and I'll probably keep doing it occasionally.... but you really should write a book or at least create a blog of some sort and write down all these bits of trivia and tidbits from 40 years of chess programming & tournaments.

But it didn't take long. In 84 we were doing 80K, by 85 we were doing 160K or so, and by 86 we had passed him although chiptest was just being built and took the NPS battle to a whole new level.

Brag brag brag...

How many nodes per $ did Belle get compared to your $$$$$ Crays??

If Ken had wanted to, he could have built Belle with his own money. You probably couldn't have even seen a Cray for the money you had. (I'm still amazed that Cray spent so much money over the years supporting computer chess.)

Sounds like Ken won.

Of course, Hsu beat Ken on the nodes per $. It cost Hsu darn near nothing for the first version of ChipTest.

[bob]

Belle didn't count nodes, so they could not display node counts or NPS on the terminal, so Joe rigged up an integrating counter that essentially counted pulses per second. He hot-wired it to the chess hardware clock, as each major tick was one node. It was interesting to watch when he first hooked it up as nobody had gone that fast.

Interesting bit of trivia.

Bob, I've mentioned this to you before and I'll probably keep doing it occasionally.... but you really should write a book or at least create a blog of some sort and write down all these bits of trivia and tidbits from 40 years of chess programming & tournaments.

But it didn't take long. In 84 we were doing 80K, by 85 we were doing 160K or so, and by 86 we had passed him although chiptest was just being built and took the NPS battle to a whole new level.

Brag brag brag...

How many nodes per $ did Belle get compared to your $$$$$ Crays??

Perhaps a better question is "how many nodes per hour of development time did each get?" Designing hardware burns a lot of hours. And then the design is fixed and changing it is a bit of a pain compared to software.

If Ken had wanted to, he could have built Belle with his own money. You probably couldn't have even seen a Cray for the money you had. (I'm still amazed that Cray spent so much money over the years supporting computer chess.)

Sounds like Ken won.

Of course, Hsu beat Ken on the nodes per $. It cost Hsu darn near nothing for the first version of ChipTest.