maddog: Letter to InfoWorld: Linux 64-bit since 1995!

[Jon maddog Hall]

tedroche at tedroche.com said:
> Bravo!

Thank you, Ted.

> I believe NT was available for the Alpha chip.  Imagine there is also a story
> there as well.

NT was originally supposed to run on the MIPS, Alpha and Intel chips, all
capable of little endian format.  MIPS was more or less still-born, and I do
not belive NT ever was officially released for it.

NT was available for the Alpha, and isn't any more.  Microsoft dropped support
for it several years ago, leaving a lot of Alpha NT customers high and dry.
Of course the customers blamed Digital, not Microsoft.

ken at jots.org said:
> So, regarding x86-64, yeah, MS is pretty late to the party.  And since that's
> what runs the huge majority of desktops, that's really where it's at.

Sorry, Ken, have to disagree with you.  There is such a thing as a "server
system".

There are two big differences that go into making a "64-bit" operating system:

	o use of 64-bit registers as the "native" registers
	o ability to access large amounts of virtual memory

There are four minor differences that *usually* "come along" with those:

	o ability to do "double-word" floating point operations as one operation
	o ability to access large amounts of real memory
	o ability to handle large data busses
	o ability to do "atomic" operations over large data structures

Note that the "double-word" floating point operation is especially useful in
"C", since all single-word floating point operations are expanded to
double-word, then truncated back to single-word again.  If you are doing lots
of floating-point work (scientific and engineering), this means it typically
speeds up significantly given the same number of clock-cycles per operation.

Microsoft operating systems eventually did the first "big difference", and
the minor differences, but they (until now) have not been able to access large
amounts of virtual memory.

There are whole classes of problems that really benefit from having large
address spaces.  Data bases, for instance can have much larger tables, with
fewer levels of indices.  Hashing algorithms can work much better, mmap comes
into its own, along with very large amounts of virtual memory being able to be
locked into real memories.  Movie rendering becomes more interesting when you
can map the entire film into one virtual address space.  Simulations can be
done easier when you don't have to worry about artificial "edge processing"
caused by lack of address space.

If Microsoft had only touted their operating systems for desktop, I might
agree with you, but they have been saying that they are a server operating
system too, and I feel that their inability to produce a TRUE 64 bit operating
system until now is pathetic, particularly when Digital offered them the code
to make NT truly 64-bit back in 1992.  If Microsoft had accepted the code, then
the Alpha would have been the only processor to support those very large
address spaces, and perhaps its fate would have been a little bit different.
As it was, the Alpha was basically a very fast, incompatible Intel chip to most
developers.

The lack of real 64 bit support in MS products (particularly in their "server"
products) I believe has hurt the computer industry.  While we had a few
applications (most of the database vendors, some CAD apps) re-write their code
to take advantage of the larger address space, a lot of applications held
back until MS said they were coming out with it.

The article in Infoworld that I responded to was trumpeting the "arrival" of
this address space.  I simply pointed out that (once again) MS was late to the
table.

Apparently the Infoworld people agreed with me.  In fact they wrote me a letter
telling me how much their entire staff appreciated my letter.

md
-- 
Jon "maddog" Hall
Executive Director           Linux International(R)
email: maddog at li.org         80 Amherst St. 
Voice: +1.603.672.4557       Amherst, N.H. 03031-3032 U.S.A.
WWW: http://www.li.org

Board Member: Uniforum Association, USENIX Association

(R)Linux is a registered trademark of Linus Torvalds in several countries.
(R)Linux International is a registered trademark in the USA used pursuant
   to a license from Linux Mark Institute, authorized licensor of Linus
   Torvalds, owner of the Linux trademark on a worldwide basis
(R)UNIX is a registered trademark of The Open Group in the USA and other
   countries.