Author: Robert Hyatt
Date: 14:17:03 11/17/98
Go up one level in this thread
On November 17, 1998 at 16:56:43, Eugene Nalimov wrote: >Sorry, with 8-bytes IEEE doubles counter will cease to >increment when it'll reach (roughly) 2**53 == 10**16, not >10**308. Of course it's still better than 4*10**9 with >32-bit unsigned integer. > >Eugene I think bruce was talking about a "long long" or "__int64" data type which goes up to 2^63-1... > >On November 17, 1998 at 15:10:48, Dann Corbit wrote: > >>On November 17, 1998 at 13:03:38, Bruce Moreland wrote: >>[snip] >>>An acceptable solution is to use a 64-bit number to count nodes. On current >>>machines, this will delay the advent of this problem for something on the order >>>of a million years. >>A double would also be peachy. On some machines, floating point is as fast or >>faster than integer math. Unless you added them in batches, at some point a >>double would not increment any more (that pesky DBL_EPSILON). But with an IEEE >>8 byte double, you should be able to hold about 10^308, which should hold down >>the fort for a while.
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.