Author: Eugene Nalimov
Date: 11:40:22 03/23/99
Go up one level in this thread
On March 23, 1999 at 14:25:26, Dann Corbit wrote: >On March 23, 1999 at 14:23:29, KarinsDad wrote: > >>On March 23, 1999 at 13:47:41, Eugene Nalimov wrote: >> >>>On March 23, 1999 at 13:32:17, Dann Corbit wrote: >>> >>>>Does anyone know of Crafty binaries for the Cray computers? >>>> >>>>I may have need of one. I would like to get the best possible binary from a >>>>commercial compiler if that is feasible. To make matters more difficult, I >>>>would like to get the source files I have tweaked compiled if possible (not >>>>necessary, but I make special tiny tweaks for the C.A.P. project). >>>> >>>>Also, about tablebase files for the Cray. Does anyone know if the standard >>>>files have the right endian ordering, or do they need to be recomputed? >>>> >>>>I do not know the model right now, but I know that the machine is two years old. >>> >>>Tablebases and compressed tablebases are endian-neutral. >>> >>>Eugene >> >>What does this mean? >> >>I thought that endian ordering was based on the capabilities of the chip and/or >>OS, not on the database. Wouldn't you need a different set of code to access the >>files properly on different systems? >They must be stored as character strings or something like that. You can't get >away with it if you use integers or floating point. It's pretty remarkable if >it works. Of course you can use binary representation, just be sure you don't extract more than byte at a time. Look at a code fragment from the probing code: #define R4(i) \ ((ptr[i] << 24) + \ (ptr[(i) + 1] << 16) +\ (ptr[(i) + 2] << 8) +\ (ptr[(i) + 3])) Obviously that will work on both little-endian and big-endian machines. You'll lost some performance, but in TB case that code works only during program initialization. Yes, formally speaking I'd have to say "TB probing code is endian- neutral", but TBs without probing code are quite useless. And I doubt there ever will be indepenently developed version of the probing code. 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.