Thirty plus years ago

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

User avatar
sje
Posts: 4675
Joined: Mon Mar 13, 2006 7:43 pm

Thirty plus years ago

Post by sje »

For those who missed the joys of hands-on eight bit programming of the mid 1970s, here's your chance to show your coding skills:

http://www.altairkit.com/

All for US$1,500.

I'd be seriously tempted if it were the 8800B model that had several more features and a much cleaner front panel.
nczempin

Re: Thirty plus years ago

Post by nczempin »

sje wrote:For those who missed the joys of hands-on eight bit programming of the mid 1970s, here's your chance to show your coding skills:

http://www.altairkit.com/

All for US$1,500.

I'd be seriously tempted if it were the 8800B model that had several more features and a much cleaner front panel.
Or you could get a Pico Edition XGamestation for a mere $59.95:

http://www.xgamestation.com/view_product.php?id=29

Granted, the 80 MHz leave something to be desired towards the retro-8-bit-feeling, but you really need that if you're going to generate NTSC/PAL in software, via a mere 4-bit DAC!

Closer in spirit to the Altair kit is the Apple I Replica:
http://www.applefritter.com/briel


8 bit is definitely fun! I think it was a high water mark for those creating games by themselves. Nowadays you need about 150 people for a commercial game, back then the lone coder was dominant.

Of course, in chess engines the lone coder is still dominant...
User avatar
sje
Posts: 4675
Joined: Mon Mar 13, 2006 7:43 pm

Re: Thirty plus years ago

Post by sje »

One attractive feature of the early microcomputers was that the main memory was usually as fast as the processor. It was like all references were to level one cache, a far cry from today's usual situation.

I built an IMSAI 8080 (2 MHz Intel 8080, 4 KB RAM, traditional front panel) back in 1976 and it was a lot of fun. I never got around to hooking up peripherals, so I didn't do any chess on it. There has been an attempt starting some six years ago to revive the machine; I had ordered one but later canceled as there was no real progress in making a delivery. Others are still waiting.

A little while after I built the IMSAI I got some part time access to some Zilog z80 word processor terminals that had 32 KB RAM and could be reprogrammed. This was the platform for my first complete chess program. It was weak, but could still beat a typical human player. It was my first and last chess program written entirely in assembly language.

Believe it or not, back in the late 1970s there was at least one Lisp interpreter available for the eight bit world, and it would have been possible to write a non trivial chess program in Lisp for a limited machine. Given that my z80 program could manage only some 35 nodes per second, I can only assume that playing a Lisp program on one of those machines would have required great patience.
User avatar
hgm
Posts: 27796
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Re: Thirty plus years ago

Post by hgm »

sje wrote:One attractive feature of the early microcomputers was that the main memory was usually as fast as the processor. It was like all references were to level one cache, a far cry from today's usual situation.
Not at all! Today's L1 sizes are often bigger than the address space of those 8-bit machines. On an Intel machine you have 32KB code and 32KB data. On an AMD even 2x64KB. The Z80 and 6502 were limited to 64KB memory, and most machines did have to go with less. x86 machine code is not very much larger than that of those old 8-bitters, perhaps only 50%. (Most simple instructions take 3 bytes, opcode + modRM + offset, rather than 2.) It should be easy to write engines that run entirely from the L1 cache. Except for the hash table, but on these 8-bitters you could not afford one anyway.
User avatar
sje
Posts: 4675
Joined: Mon Mar 13, 2006 7:43 pm

Re: Thirty plus years ago

Post by sje »

If today's machines had sufficient level one cache, then we wouldn't be reading about the nearly pathological fear of table lookups that likely access main memory.

--------

I found out the other day about the availability of 4 GB FB-DIMMs. That means my Mac Pro can be expanded to 32 GB RAM instead of "only" 16 GB, although I'd have to rob a bank to afford the parts.

What to do with 32 GB?

1) Keep the opening book in memory.

2) Jumbo transposition tables.

3) Tablebases in RAM.

4) Chatterbot to annoy ICS human opponents.
User avatar
sje
Posts: 4675
Joined: Mon Mar 13, 2006 7:43 pm

32 GB price/availability

Post by sje »

At US$6,579 for 32 GB, I think I'll wait a while for the price to come down.

See: http://eshop.macsales.com/item/Other%20 ... B4MPK32GB/