[ILUG] 64 Bit OS

[John P. Looney]

On Fri, Feb 25, 2000 at 06:16:03AM -0600, Stephen_Reilly at dell.com mentioned:
> 	both, ehm Dell are involved in the drive to port Redhat effectively.
> I therefore would like to know everything possible about it from external
> sources before I talk to these people again. I can't really speak for Dell
> on the topic though, you know, yada yada. Although last word from Redhat,

 Ah right. Does this mean we'll get another hardware-vendor version of the
OS (Like SGI, VA Linux, ABIT are all doing) ? 

> 	"Red Hat Linux is already available for the Alpha (yours truly is
> 	using such a machine :), so there shouldn't really be much porting
> 	beyond "recompile". Whether architecture-specific changes are
> 	necessary is another thing, but 64-bitness issues have generally been
> 	worked out by now."
> Hmmm, working or optimised ?

 On 64 bit, I don't see much difference. The "Big wins" for 64 bitness
are:

 1) Break backware compatibility - everything has to be recompiled (or
"thunked", like Win16 apps on Win32 OSes). So the chip can be greatly
simplified - you start out with a new architechture.
 2) You get 64 bit filedescriptors, sizes, time_t is 64 bit etc.
 3) Address space is huge (none of this any greater than 1GB is a hack
        type stuff).

 Going from 8->16 bit rocked. Everyone used numbers bigger than 256. Going
from 16->32 made a big difference, for stuff like memory, and the fact
that a lot of things were bigger than 64k. However, how often do you use
numbers bigger than 4,000,000,000 in code ? And, considering that, how
much "optimisation" would you require, once it's *there* ? Smeg all. As
long as all the bits of the OS that could use 64bit for speed, like
memcpy() & the like are optimised, you are sorted. That's been done for
about 18 months now, on the Alpha port.

Kate