Author: Eugene Nalimov
Date: 12:25:07 06/15/99
Go up one level in this thread
On June 15, 1999 at 15:03:43, KarinsDad wrote: >On June 15, 1999 at 13:38:16, leonid wrote: > >>On June 15, 1999 at 10:15:18, David Blackman wrote: >> >>>On June 15, 1999 at 07:06:38, leonid wrote: >>> >>>>Hi! >>>> >>>>In this year, presumably, 64 bits Intel chip should be produced. It >>>>expected to be more that attractive - 128 registers each 64 bits wide. >>>>Do you know something about it? When it will hit the market? >>>> >>>>Leonid. >>> >>>The architecture has been disclosed. You can get it from somewhere on the intel >>>website (over 1MB PDF). It looks pretty weird. At the same clock speed, i >>>suspect it will be faster than anything else. But it might turn out to be hard >>>to make at fast clock speeds. The public documents i've seen give no hints about >>>how they will implement it, when it will hit the market, etc. >>> >>>Past attempts to make similar architectures were huge flops. (There were quite a >>>few in the late 1980s.) But Intel has more resources, and technology has >>>improved a lot, so who knows? >>> >>>I suspect most of it's advanced features won't be a big help to chess programs >>>unless the programmers are extremely clever. >> >>I think otherwise. I have the impression that this chip could be of >>incredible help in the chess programming. For this I see 2 reasons: >> >>1) 128 registers and no more meager, around ten, registers of present CPU. >> Many of those new registers can be used for the variables of quick access. >> >>2) Each register is 64 bits wide. This goes in miraculous coincidence with >> the chess board that is composed of 64 squares. >> >>And at the end, one small additional advantage. Very often in the game we >>must save the chess board position far later recall. Now this will be done >>at double speed. More I think about the 64 bits computer, more I am eager >>to reach it for first tryal. >> >>Leonid. > >You may be correct, but there are some other considerations: > >1) The compilers will not be able to take advantage of the new registers/64 bits >for 6-12 months until after the chip is released. Porting the Alpha compilers >will not work (different registers, etc.). > >2) The motherboard manufacturers will be behind as well. Also, you should be >EXTREMELY careful about which motherboard you use with such a chip. > >Therefore, chess programmers will not be able to take real advantage of such as >chip for at least a year and possibly a year and a half until after it is >released. > >KarinsDad :) My prediction is that you'll be able to use 64-bit MSVC immediately after first IA-64 bit (Merced) will be released. At least I can use it long before that :-). So, you'll be able to recompile program like Crafty with only minor modifications. I agree that second generation IA-64 compilers will be better. But even first generation compilers will be able to produce good code - chess programs are very easy for optimizing compilers (no virtual calls, large functions, predictable access to structures, etc.) And IA-64 contains a lot of concepts that will help chess programs immediately - e.g. predications, speculations, safe prefetching, large register file (so you can load entire board representation there), etc. Eugene
This page took 0 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.