[Info-vax] OpenVMS I64 V8.1 "Evaluation Release"?

[Johnny Billquist]

My god, I can hardly believe my eyes when I read this.

Have none of you the slightest clue about the difference between virtual 
address and physical address?

Even today, there is no CPU that can address 64 bit physical address space.
Is that relevant? No. You've been able to address 64 bits of virtual 
address space in your programs the whole time, independent of the 
physical address capabilities. And independent of the actual physical 
ram existing on the machine.

To read an argument that it was pointless to have machines with 64-bit 
addressing since there were no existing machines that shipped with more 
than 2 GB of RAM anyway is about as stupid an argument as I've ever 
seen. That is just showing that you have not understood computers at all.

	Johnny

On 2012-03-21 02.19, Michael Kraemer wrote:
> Dennis Grevenstein schrieb:
>
>> As you yourself like to say, it can be important to be
>> compatible across the whole line from a workstation
>> to the bigger servers. If you just count workstations,
>> what about the Sun Ultra 80 (4GB), SGI Octane (8GB)
>> and several HP Cxxx or Cxxxx workstations with at least
>> a PA8000 CPU?
>
> I'm not only counting workstations,
> but those oodles of RAM, which make 64bit worthwhile,
> appeared in expensive servers first and then trickled
> down the product lines. When this trend reached workstations
> and entry servers, one could call 64bit "mainstream",
> at the earliest. This wasn't before beginning 200x,
> give or take a year. Before that, 64bit was a rather moot point,
> although CPUs offered 64bit addressing almost a decade before,
> that was my point.
>
> To give a more practical example,
> about two years ago I "inherited" a half dozen Alpha's
> (workstations/entry servers), of mid-1990s vintage.
> They had around 160MB RAM
> (upgrade from an original 64MB or so, I guess),
> a 1GB disk (the original one, now used as swap) and 4GB disk upgrade
> (now used as system disk). That's what was available
> and what mere mortals could afford at that time.
> So one couldn't care less whether the CPU
> addresses 64bits or just 32bit, it just doesn't make a difference.
> Nevertheless the machines did useful work, I guess.
>
>> One could have chosen one of the mature RISC platforms over
>> x86_64, but for some reason it was adopted really fast.
>
> You can run old x86 stuff, have 64bit in addition,
> and it came just when the time was right,
> i.e. problem sizes started to demand >4GB and
> such amounts of RAM became affordable.
>
>> If anything, we can state that Intel couldn't get the Itanium out
>> fast enough. All the other 32bit -> 64bit updates took many
>> years before 64bit software was ready.
>
>  From the practical point of view,
> there's little incentive for 64bit software to become ready
> if you can't stuff enough RAM into your boxes.
> Do a calloc( 4GB ) on a 1GB machine and watch the poor box
> grind to a halt.
> OTOH, it's not too difficult to ready a software for 64bits,
> provided one didn't mess int's/long's etc (like I did unfortunately :-(
> Run the compiler in 64bit mode and it'll tell you most of
> the critical constructs.
>
>> MIPS R4000 1991 - IRIX 6.2 1996
>> PA8000 1996 - HP-UX 11.00 1997
>> UltraSPARC I 1996 - Solaris 7 1998
>> Alpha 1992 - OpenVMS 7.1 1997
>>
>> When Intel introduced the 80386, people were happy with 4MB
>> of RAM,
>
> That was around 1987, and I doubt many PCs had that much RAM.
> Most were happy with 640kB, as they were told by their master.
>
>