Author: Robert Hyatt
Date: 05:37:00 05/19/99
Go up one level in this thread
On May 19, 1999 at 01:37:52, James B. Shearer wrote: >On May 18, 1999 at 07:42:16, Robert Hyatt wrote: > >>On May 17, 1999 at 22:37:30, James B. Shearer wrote: >> >>>On May 17, 1999 at 09:28:31, Robert Hyatt wrote: >>> >>>>On May 17, 1999 at 00:36:39, James B. Shearer wrote: >>>> >>>>>On May 14, 1999 at 09:52:14, Robert Hyatt wrote: >>>>> >>>>>>On May 14, 1999 at 01:38:11, Gregor Overney wrote: >>>>> >>>>> <snip> >>>>> >>>>>>> >>>>>>>2) You still need to write the correct algorithms to make this chip work. And >>>>>>>those algorithms are pretty complex (see evaluation functions etc.) >>>>>>> >>>>>> >>>>>>But it has _already_ been done. All that is left is to use the "new" fab >>>>>>process to increase density and clock speed.. DB's chess chips only ran at >>>>>>20-24 megahertz. running that up to 16x faster seems quite easy with todays >>>>>>silicon capabilities as that would still be a modest < 400mhz processor. >>>>> >>>>> This assumes: >>>>>1) Hsu's startup has the right to use the IBM deep blue code. >>>> >>>>He's already publicly stated that he is doing this, so I would assume that >>>>permission has already been granted? >>> >>> I have not seen any such statement. >>> >> >>Then pick up the current issue of IEEE Micor and you will. :) >> >> >> >>>> >>>>>2) The IBM deep blue code (written for the big endian power chips) can be >>>>>trivially ported to the (little endian) Intel chips used in PCs. >>>>> I would doubt both of these assumptions. >>>>> James B. Shearer >>>> >>>>(2) is a non-issue. IE 'crafty' is much more 'endian' aware than DB, yet it >>>>runs on big-endian and little-endian machines with no problems at all. The >>>>PCI interface could 'correct' the endian-order of the data without the chip >>>>ever knowing... >>> >>> (2) might be a non-issue if the deep blue code was carefully written to >>>be endian independent. Is crafty naturally endian independent or carefully >>>written (and tested) to be endian independent? There are often major problems >>>in porting code to other endian machines if this was not considered and designed >>>for from the start. >>> James B. Shearer >> >>crafty is endian-sensitive, and has been specifically written to work with >>big-endian _and_ little-endian architectures. In the case of DB processors, >>since they sit outside the PC processor, and behind a PCI interface, the PCI >>interface itself can easily handle the endian issue if they have one. > > Perhaps we are referring to different things. I was talking about the >program running on the host processor (RS/6000 for deep blue, IBM PC for Hsu's >new chip) which does the top plies of the search and feeds positions to the >chess chip. Writing this program from scratch would require substantial effort. no reason to 'write it from scratch'. It already exists. The only endian issue I see is between the host and the chess processor, and since the chess processor sits on the PCI bus, the interface on the circuit board can do the byte-swapping stuff needed to make the endian issue disappear... Which means that the 'stock' DB software could be re-used (the code supporting the chess processors, etc). >All of these start up costs require financing. I think Hsu will find it >difficult to raise sufficient funds since the revenue potential for a custom >chip based chess program is in my opinion insufficient to make this an >attractive investment. > James B. Shearer It sounds to me like a 'done deal'. IE he says "I have a start-up ..." not "I am testing the water to see if I can get a start-up funded and ..."
This page took 0.01 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.