Wish we had (a 64-bit) Prop II... Maybe Prop III?

Rayman

I've started writing this chess program from scratch and it would be very convenient to have 64 bits!·

Then, every bit can represent a square on the board.· This makes programming a lot easier and execution a lot faster...

Also, with many more faster cogs looking at moves, the strength could go way up.

(side note:· I wonder if my new Imagecraft C compiler has a 64-bit data type...)

Post Edited (Rayman) : 7/7/2008 9:38:31 AM GMT

ImageCraft

re: (side note: I wonder if my new Imagecraft C compiler has a 64-bit data type...)

We do not -- yet, but it would be something we will be adding at some point.

MarkS

I thought the Prop II was going to still be 32-bit, but with 64 I/O pins?

Mike Green

Yes. ImageCraft's compiler for some other architectures has double precision integer arithmetic and they plan to add it to the Propeller compiler at some point.

Philldapill

I've noticed some confusion about the Prop II and it being 64 or 32-bit. So which is it? I can't wait.

hippy

I've not seen anything which suggests 64-bit so not sure where that impression crept in from.

Having two 32-bit DIRA and DIRB etc wouldn't really make sense for 64-bit and it's Cog architecture rather then Hub RAM which would change if we went to 64-bit.

Rayman

What, no 64-bit? Never mind...

stamped

Is there an official ETA for the Prop II?

Graham Stabler

Rayman said...


........ and execution a lot faster...

It's chess [noparse]:)[/noparse]

Ale

There where Chess programs for the C64 and the TK-85 (last one had only 16 kbytes RAM and Z80)... so I do not see a problem with the word size. The lack of real pointers may hinder table access quite a bit though (did someone look at my FFT for the propeller ?

Leon

stamped said...
Is there an official ETA for the Prop II?

No.

Leon

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Amateur radio callsign: G1HSM
Suzuki SV1000S motorcycle

Rayman

Ale said...
There where Chess programs for the C64 and the TK-85 (last one had only 16 kbytes RAM and Z80)... so I do not see a problem with the word size.

I've already ported someone else's Chess code to the Prop and it fits fine with plenty of room to spare.· But, now I want to make my own chess code...

Because there's enough room, I want to use Bitfields to speed the calculation.· It's a lot easier and faster to use Bitfields when you have a 64-bit native type...·


·

Ale

Since... 92 or 93 I wanted to write my own chess program, but my lack of understanding of the inner workings forbid it to me :-/ maybe you can be kind enough to explain it in some detail .

Well, maybe next revision of C will have long long support. Anyways using 32 bits to "emulate" 64 requires loads of code. Especially due to the lack of pointers... Maybe you can do some small functions that do what you want, set/clear and test I assume and you put them into the LMM kernel that the C provides, I assume they have provided some easy way to call a function that directly in the kernel space is, or is not possible, so you do

rayman_and(&var1, &var2);

instead of

var1 = var1 & var2;

But no idea if is possible

hippy

No lack of pointers in Spin nor PASM, just a bit more awkward to use in PASM through having no indexed addressing. 64-bit math with Spin not having unsigned comparisons and operators is more a problem.

VAR
  long var1[noparse][[/noparse] 1 ]
  long var2[noparse][[/noparse] 1 ]

PUB Main
  rayman_and( @var1, @var2 )

PRI rayman_and( ptr1, ptr2 )
  long[noparse][[/noparse] ptr1     ] &= long[noparse][[/noparse] ptr2     ]
  long[noparse][[/noparse] ptr1 + 4 ] &= long[noparse][[/noparse] ptr2 + 4 ]

Or do you mean C doesn't have pointers ? Surely it must ???

Post Edited (hippy) : 7/7/2008 7:08:52 PM GMT

Rayman

Ale said...
Since... 92 or 93 I wanted to write my own chess program, but my lack of understanding of the inner workings forbid it to me :-/ maybe you can be kind enough to explain it in some detail .

Here's what I'm going to use as a guide:
http://www.gamedev.net/reference/programming/features/chess1/

KeithE

Check out talkchess.com sometime. A fair number of chess programmers hang out there.

I think that it would be fun to write a standalone chess program for the prop. Maybe you could even build your own autosensory chessboard, and use the prop to handle the RF for piece detection (see Ken Thompson's work), graphics display (maybe multiple types of displays), and even speech synthesis. But don't set your performance goals too high initially. e.g. you're probably not going to have much of a hashtable, so your branching factor is going to be high in comparison with the state of the art. I would plan on trying to compete with the Excaliber products as a personal goal. You might even lobomize a cheap one for fun. I think that fun would come from the level of integration you could achieve, and having something portable that could run from batteries and have a decent battery life.

vampyre

I feel bad for Chip. If and when the prop II is released, everyone is going to expect it to levitate and dance.
My guess is that it will simply be two props 'glued' together. in theory that would make it more powerful, and easier (faster) to design.

Rsadeika

@vampyre, I have to agree with you on this. Chip is no doubt cringing every time he reads some of these threads, the expectations for the Prop II are reaching the level of, the new chip can do it all. I am starting to see two groups here, one group wants the new chip to be a base for there own personal computer, while the other group is still in the embedded processor·camp. It will be interesting to see how it really develops after the release of the Prop II.

Ray

Graham Stabler

Whether propII can levitate, dance or otherwise I am sure it will surprise us.

Graham

Ken Peterson

We got a taste of what the specs might be like in this forum. If it's anything like that, it'll levitate enough for me, and I'm sure with a little code we can make it dance! [noparse]:)[/noparse]

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔

ErNa

What would we do, if we had? More structurized work than we do today? The transputer T9000 came to late, for they were not able to make the virtual routing unit. So the whole project was stopped. Without this feature, the T9000 would have worked anyway. The PropI has a infinite number of cogs the moment, someone creates a virtual network. The PropII only shifts the borders, but limits will exist anyway.

waltc

The T9000 without its virtual routing would have been just another processor in a field saturated with all sorts of exotics at the time like MIPS, 860, 960, Intel's Neural net chips, Dec's Alpha, etc.

As far as the Propeller goes I'm firmly in the embedded camp. Its a nice multiprocessor chip thats easily accessible to us mere mortals unlike the other ones like Cell or Xmos. Hell its easier to learn than any of the ARM variants out there with their 800 page data books.

FWIW