Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A random thought about bitboards

Author: Robert Hyatt

Date: 20:11:38 10/24/01

Go up one level in this thread


On October 24, 2001 at 21:52:29, Tom Kerrigan wrote:

>On October 24, 2001 at 17:48:26, Tony Werten wrote:
>
>>Could be. I'm quite sure though that's what they said about 32bits when the
>>got
>>16 bits, and about 64 when they got 32.
>>
>>People don't know what they want until there has been a big advertisement
>>campagne explaining why they really need it. After that it goes fast.
>
>There was a super-big need for 32 bit addressing.
>
>Very few <= 16 bit chips have address spaces as wide as their datapaths.
>Examples: the 8 bit Z80 had a 16 bit address space and the 16 bit 8086 had a 20
>bit address space. These are pretty hacky architectures because they need
>separate "sub-datapaths" to do addressing. Programming these chips is also hacky
>because you need segments and near/far pointers and crap like that.
>
>Most 32 bit chips are 32 bits all the way through and people get along with them
>just fine, with no hacks or awkward programming tricks to increase address
>space.
>
>Basically, there was a big, obvious need for ~32 bit addressing, but never a
>big, obvious need for > 32 bit addressing.
>
>This is backed up by the fact that 64 bit CPUs are readily available and few
>people use them.
>
>There are definitely cases where you want > 32 bit addressing, but not many.
>Remeber, 64 bits gives you 4 _billion_ times as much address space as 32 bits.
>
>I don't expect to see 128-bit CPUs (or at least a need for them) within the
>next, say, 30 years.
>
>-Tom


Note that 64 bit processors don't necessarily have 64 bit address spaces.
The Cray family is one example with 64 bit values, but a 32 bit address
space.  The MIPS guys have not yet implemented a full 64 bit address space
for obvious reasons...

In fact, I don't know of any 64 bit processors that really have a full 64
bit address space, due to the size of it and the unnecessary waste involved
in gating addresses that are impossible to generate.  I don't remember what
the MIPS guys (for example) have done recently, but early chips had something
like a 43 bit address space or some such oddball number...



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.