VMS on ARM?

[Bill Gunshannon]

Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi? :-)

bill
(Yes, I could really use that about now!! :-)

[Robert A. Brooks]

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi?  :-)
bill
(Yes, I could really use that about now!!  :-)

Uh, we're somewhat busy now.

We don't want Reagan to go running off into the woods quite yet.

[Chris]

Post by Robert A. Brooks

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi? :-)
bill
(Yes, I could really use that about now!! :-)

Uh, we're somewhat busy now.
We don't want Reagan to go running off into the woods quite yet.

Lol, but would have thought the most likely next arch would be
Power. Would fit better as market perception of professionalism.

Chris

[Robert A. Brooks]

Post by Chris
Lol, but would have thought the most likely next arch would be
Power. Would fit better as market perception of professionalism.

Big endian -- not going to happen.

[John Reagan]

Post by Robert A. Brooks

Post by Chris
Lol, but would have thought the most likely next arch would be
Power. Would fit better as market perception of professionalism.

Big endian -- not going to happen.
--
-- Rob

Power, like Alpha, is bi-endian. It can be either big or little.

[John Reagan]

Post by Robert A. Brooks

Post by Chris
Lol, but would have thought the most likely next arch would be
Power. Would fit better as market perception of professionalism.

Big endian -- not going to happen.
--
-- Rob

Power, like Alpha, is bi-endian. It can be either big or little.

I should have added Itanium as well. The user-mask has an endian bit.

And yes, LLVM has both a ppc64be and ppc64le target.

[Dave Froble]

Post by Chris

Post by Robert A. Brooks

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi? :-)
bill
(Yes, I could really use that about now!! :-)

Uh, we're somewhat busy now.
We don't want Reagan to go running off into the woods quite yet.

Lol, but would have thought the most likely next arch would be
Power. Would fit better as market perception of professionalism.
Chris

Don't n0ow much about ARM and Raspberry Pi, bot would not either of them
be a good thing, allowing VMs on small, cheap, HW? Maybe even take back
some of the embedded market?

[Arne Vajhøj]

Post by Dave Froble
Don't n0ow much about ARM and Raspberry Pi, bot would not either of them
be a good thing, allowing VMs on small, cheap, HW?  Maybe even take back
some of the embedded market?

x86-64 support will allow VMS to run on 500 dollars HW. And if you don't
need it 24x7 then maybe a cloud instance @ 5 cents per hour may be
cheaper.

Sure a Raspberry Pi may only cost 50 dollars, but
would it matter for the typical VMS user?

I assume that the VMS licenses with ongoing support is
going to cost a lot more than the HW.

Arne

[Chris]

Post by Arne Vajhøj

Post by Dave Froble
Don't n0ow much about ARM and Raspberry Pi, bot would not either of
them be a good thing, allowing VMs on small, cheap, HW? Maybe even
take back some of the embedded market?

x86-64 support will allow VMS to run on 500 dollars HW. And if you don't
cheaper.
Sure a Raspberry Pi may only cost 50 dollars, but
would it matter for the typical VMS user?
I assume that the VMS licenses with ongoing support is
going to cost a lot more than the HW.
Arne

Rpi hasn't anything like the io bandwidth needed for a server os
anyway, nor enough memory.

Put VMS on Power however and you have an immediate security benefit,
in that x86 malware just won't run on it. Maybe one day eh ?...

Chris

[Scott Dorsey]

Post by Chris
Rpi hasn't anything like the io bandwidth needed for a server os
anyway, nor enough memory.

It has more bandwidth and memory than the 11/780 that my college had.
You can stick 50 people running vi on the thing and get decent performance,
which is more than the 780.

Post by Chris
Put VMS on Power however and you have an immediate security benefit,
in that x86 malware just won't run on it. Maybe one day eh ?...

We already have that benefit with Itanium. It is a mixed blessing, because
legitimate code won't run on it either....
--scott

[Chris]

Post by Scott Dorsey

Post by Chris
Rpi hasn't anything like the io bandwidth needed for a server os
anyway, nor enough memory.

It has more bandwidth and memory than the 11/780 that my college had.
You can stick 50 people running vi on the thing and get decent performance,
which is more than the 780.

Post by Chris
Put VMS on Power however and you have an immediate security benefit,
in that x86 malware just won't run on it. Maybe one day eh ?...

We already have that benefit with Itanium. It is a mixed blessing, because
legitimate code won't run on it either....
--scott

Rofl, um yes. I guess if you design a cpu that is so complex no one can
copy it and few understand it, no surprise that there are few able to
write good compilers for it.

Hoisted on their own protectionist and greedy petard springs to mind...

Chris

[^P]

Post by Chris

Post by Scott Dorsey

Post by Chris
Rpi hasn't anything like the io bandwidth needed for a server os
anyway, nor enough memory.

It has more bandwidth and memory than the 11/780 that my college had.
You can stick 50 people running vi on the thing and get decent performance,
which is more than the 780.

Post by Chris
Put VMS on Power however and you have an immediate security benefit,
in that x86 malware just won't run on it. Maybe one day eh ?...

We already have that benefit with Itanium. It is a mixed blessing, because
legitimate code won't run on it either....
--scott

Rofl, um yes. I guess if you design a cpu that is so complex no one can
copy it and few understand it, no surprise that there are few able to
write good compilers for it.
Hoisted on their own protectionist and greedy petard springs to mind...
Chris

There Russian / East German clones of the VAX

^P

[Phillip Helbig (undress to reply)]

Post by ^P
There Russian / East German clones of the VAX

Which is why some VAX chips had written them, in Russian, "VAX: When you
care enough to steal the best". :-)

[Arne Vajhøj]

Post by Chris

Post by Robert A. Brooks

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi?  :-)
bill
(Yes, I could really use that about now!!  :-)

Uh, we're somewhat busy now.
We don't want Reagan to go running off into the woods quite yet.

Lol, but would have thought the most likely next arch would be
Power. Would fit better as market perception of professionalism.

Power is yet another expensive niche CPU.

That has already been tried.

Arne

[Scott Dorsey]

Post by Chris
Lol, but would have thought the most likely next arch would be
Power. Would fit better as market perception of professionalism.

PPC and the higher end ARM designs are all good architectures. PPC has
IBM behind it and ARM has Apple behind it. I think the long-term future
of one of them is much better than that of the x86... but which of the
two? In a decade we should know and then it might be worth considering
a VMS port.
--scott

[John Dallman]

Post by Scott Dorsey
PPC and the higher end ARM designs are all good architectures. PPC
has IBM behind it and ARM has Apple behind it. I think the long-term
future of one of them is much better than that of the x86... but
which of the two? In a decade we should know and then it might be
worth considering a VMS port.

My money's on ARM, because there's more than Apple behind it. ARM
Holdings, who define the architecture and produce the reference core
designs, are doing cores and variants for more niches than Apple. Fujitsu
are building an ARM-based supercomputer. The CHERI project are doing
ARM-based capability architecture for stronger security. The PowerPC
world is less dynamic.

Also, ARM64 code is surprisingly compact. At work, I have the same code
built for a lot of platforms at work: ARM64 Windows DLLs are smaller than
Windows x86-64 ones, and Android ARM64 shared libraries are smaller than
Linux x86-64 ones (which are, in turn, smaller than AIX PowerPC ones).
This matters, since cache is the chief influence on performance these
days, and smaller code takes up less cache space.

John

[Scott Dorsey]

Post by John Dallman

Post by Scott Dorsey
PPC and the higher end ARM designs are all good architectures. PPC
has IBM behind it and ARM has Apple behind it. I think the long-term
future of one of them is much better than that of the x86... but
which of the two? In a decade we should know and then it might be
worth considering a VMS port.

My money's on ARM, because there's more than Apple behind it. ARM
Holdings, who define the architecture and produce the reference core
designs, are doing cores and variants for more niches than Apple. Fujitsu
are building an ARM-based supercomputer. The CHERI project are doing
ARM-based capability architecture for stronger security. The PowerPC
world is less dynamic.

I agree and I think being able to inherit the effort on good compilers and
development tools that come out of the embedded systems world is going to
be a huge win for ARM on the desktop. But I wouldn't actually put money on
it because things change.

I'm the guy who predicted the IBM PC would be a flop because it was overpriced,
had weird nonstandard disk drives, and unless you paid extra you couldn't get
even a barely usable single-threaded OS. But, I was wrong. Too many of those
experiences makes me reluctant to put money on long-term predictions.

Post by John Dallman
Also, ARM64 code is surprisingly compact. At work, I have the same code
built for a lot of platforms at work: ARM64 Windows DLLs are smaller than
Windows x86-64 ones, and Android ARM64 shared libraries are smaller than
Linux x86-64 ones (which are, in turn, smaller than AIX PowerPC ones).
This matters, since cache is the chief influence on performance these
days, and smaller code takes up less cache space.

Yes. ARM gets you DEC Alpha levels of compactness, the sort of compactness
that Itanium promised but could not even come close to delivering.
--scott

[Simon Clubley]

Post by John Dallman
My money's on ARM, because there's more than Apple behind it. ARM
Holdings, who define the architecture and produce the reference core
designs, are doing cores and variants for more niches than Apple. Fujitsu
are building an ARM-based supercomputer. The CHERI project are doing
ARM-based capability architecture for stronger security. The PowerPC
world is less dynamic.

That depends on what the ARM ecosystem looks like once NVIDIA get through
with it.

Simon.

[V***@SendSpamHere.ORG]

Post by Robert A. Brooks

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi?  :-)
bill
(Yes, I could really use that about now!!  :-)

Uh, we're somewhat busy now.
We don't want Reagan to go running off into the woods quite yet.

ROFL!

[Arne Vajhøj]

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi?  :-)

How many licenses do you think they could sell for that?

Arne

[Bill Gunshannon]

Post by Arne Vajhøj

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi?  :-)

How many licenses do you think they could sell for that?

How many licenses do you think they could sell for x86-64?

Your answer would be no more accurate than mine.

I seriously think there could be a market for an rPi version of VMS.

But only time will tell.

bill

[Arne Vajhøj]

Post by Bill Gunshannon

Post by Arne Vajhøj

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi?  :-)

How many licenses do you think they could sell for that?

How many licenses do you think they could sell for x86-64?
Your answer would be no more accurate than mine.
I seriously think there could be a market for an rPi version of VMS.

We know that companies like Microsoft, Redhat, SUSE, Canonical
etc. can make money selling OS and/or OS support for the
x86-64 market.

Does the same size companies exist for rPi?

If it does then sure rPi is just as good a bet as x86-64.

But if not then ...

Arne

[Scott Dorsey]

Post by Arne Vajhøj
We know that companies like Microsoft, Redhat, SUSE, Canonical
etc. can make money selling OS and/or OS support for the
x86-64 market.
Does the same size companies exist for rPi?
If it does then sure rPi is just as good a bet as x86-64.

The rPi is a low end ARM machine, but there are plenty of high end ARM
machines out there with plenty of support. And now with the Apple move
to ARM for desktop systems, it will be very interesting to see how ARM
goes.

The rPi may be more or less representative of the typical ARM system today
since they have previously found their greatest market in embedded systems,
but don't think it will be representative of the typical ARM system in two
years. ARM is in the server world with systems like the Gigabyte machines,
and they'll be farther in soon.

Is that enough to port VMS to ARM? No, but it's worth thinking about because
if things keep going it will be enough in a decade.
--scott

[Arne Vajhøj]

Post by Scott Dorsey

Post by Arne Vajhøj
We know that companies like Microsoft, Redhat, SUSE, Canonical
etc. can make money selling OS and/or OS support for the
x86-64 market.
Does the same size companies exist for rPi?
If it does then sure rPi is just as good a bet as x86-64.

The rPi is a low end ARM machine, but there are plenty of high end ARM
machines out there with plenty of support. And now with the Apple move
to ARM for desktop systems, it will be very interesting to see how ARM
goes.
The rPi may be more or less representative of the typical ARM system today
since they have previously found their greatest market in embedded systems,
but don't think it will be representative of the typical ARM system in two
years. ARM is in the server world with systems like the Gigabyte machines,
and they'll be farther in soon.
Is that enough to port VMS to ARM? No, but it's worth thinking about because
if things keep going it will be enough in a decade.

That could happen.

If server volume switch from x86-64 to ARM, then VSI would
need to follow the industry.

Arne

[Bill Gunshannon]

Post by Arne Vajhøj

Post by Scott Dorsey

Post by Arne Vajhøj
We know that companies like Microsoft, Redhat, SUSE, Canonical
etc. can make money selling OS and/or OS support for the
x86-64 market.
Does the same size companies exist for rPi?
If it does then sure rPi is just as good a bet as x86-64.

The rPi is a low end ARM machine, but there are plenty of high end ARM
machines out there with plenty of support.  And now with the Apple move
to ARM for desktop systems, it will be very interesting to see how ARM
goes.
The rPi may be more or less representative of the typical ARM system today
since they have previously found their greatest market in embedded systems,
but don't think it will be representative of the typical ARM system in two
years.  ARM is in the server world with systems like the Gigabyte
machines,
and they'll be farther in soon.
Is that enough to port VMS to ARM?  No, but it's worth thinking about
because
if things keep going it will be enough in a decade.

That could happen.
If server volume switch from x86-64 to ARM, then VSI would
need to follow the industry.

Was DEC following the industry when they created Alpha and
put VMS on it? Too many followers, not enough leaders.

bill

[Arne Vajhøj]

Post by Bill Gunshannon

Post by Arne Vajhøj

Post by Scott Dorsey
Is that enough to port VMS to ARM?  No, but it's worth thinking about
because
if things keep going it will be enough in a decade.

That could happen.
If server volume switch from x86-64 to ARM, then VSI would
need to follow the industry.

Was DEC following the industry when they created Alpha and
put VMS on it? Too many followers, not enough leaders.

DEC tried to lead with Alpha and failed.

HP and to some extent Intel trie dto lead with Itanium and failed.

It is often a good thing to lead instead of follow.

But modern CPU's is a crazy business.

It cost billions to design a CPU. It cost billions
to build fab to manufacture them. And it cost cents
in raw materials to produce one.

What does that mean? It means that volume is
almost everything. Those that sell the most
evolve the most and eventually kill their
competitors.

x86-64 did not kill Alpha, PA, SPARC and Power
(Power is not totally dead yet, but ...)
because it was a better design - it won because
Intel and AMD sold hundred of millions of CPU's
for PC's and it was impossible to compete
with that volume.

ARM is not a potential threat to x86-64 because of its design,
but because there is made hundreds of millions of them
for phones and tablets.

If one day (10 years, 15 years or whatever) ARM kills
x86-64 then VSI will not have to migrate.

Arne

[Bill Gunshannon]

Post by Arne Vajhøj

Post by Bill Gunshannon

Post by Arne Vajhøj

Post by Scott Dorsey
Is that enough to port VMS to ARM?  No, but it's worth thinking
about because
if things keep going it will be enough in a decade.

That could happen.
If server volume switch from x86-64 to ARM, then VSI would
need to follow the industry.

Was DEC following the industry when they created Alpha and
put VMS on it? Too many followers, not enough leaders.

DEC tried to lead with Alpha and failed.
HP and to some extent Intel trie dto lead with Itanium and failed.

Itanium was not DEC's. And Alpha, which was, outlasted Itanium by
a considerable amount of time. Alpha was the success, even with
everyone trying to kill it. Itanium was a failure which, sadly, VMS
chose to follow.

Post by Arne Vajhøj
It is often a good thing to lead instead of follow.
But modern CPU's is a crazy business.
It cost billions to design a CPU. It cost billions
to build fab to manufacture them. And it cost cents
in raw materials to produce one.
What does that mean? It means that volume is
almost everything. Those that sell the most
evolve the most and eventually kill their
competitors.
x86-64 did not kill Alpha, PA, SPARC and Power
(Power is not totally dead yet, but ...)

Don't think Sparc is dying yet. It has its niche. And Power
is far from dead.

Post by Arne Vajhøj
because it was a better design - it won because
Intel and AMD sold hundred of millions of CPU's
for PC's and it was impossible to compete
with that volume.
ARM is not a potential threat to x86-64 because of its design,
but because there is made hundreds of millions of them
for phones and tablets.
If one day (10 years, 15 years or whatever) ARM kills
x86-64 then VSI will not have to migrate.

Don't you mean "will have to migrate"? Any particular reason
why they can't support both if the use is there?

bill

[John Dallman]

Itanium was not DEC's. And Alpha, which was, outlasted Itanium by
a considerable amount of time. Alpha was the success, even with
everyone trying to kill it.

Alpha-based systems stopped being sold in 2007. Final Itanium chip
deliveries will be in the middle of 2021, with the end date for system
sales not yet publicised. I don't quite see how Alpha outlasted Itanium
in the commercial sense.

Itanium was a failure which, sadly, VMS chose to follow.

HP bought Compaq, who'd bought DEC. HP were doubling down on Itanium at
the time. They've lost that bet, but it took them plenty of years to
realise it.

Don't think Sparc is dying yet. It has its niche.

It is dying. Oracle have not admitted that public ally, but they got rid
of most of their SPARC chip design and Solaris OS people shortly after
releasing UltraSPARC M8. They have announced that there won't be a
successor to the M8 core and while they have promised that Solaris will
be runnable into the 2030s, they don't seem to know how this will be
accomplished.

John

[Bill Gunshannon]

Post by John Dallman

Itanium was not DEC's. And Alpha, which was, outlasted Itanium by
a considerable amount of time. Alpha was the success, even with
everyone trying to kill it.

Alpha-based systems stopped being sold in 2007. Final Itanium chip
deliveries will be in the middle of 2021, with the end date for system
sales not yet publicised. I don't quite see how Alpha outlasted Itanium
in the commercial sense.

Because Alpha started long before Itanium and is still in use as Itanium
reaches EOL. Just count up the years for each of them and compare.

Post by John Dallman

Itanium was a failure which, sadly, VMS chose to follow.

HP bought Compaq, who'd bought DEC. HP were doubling down on Itanium at
the time. They've lost that bet, but it took them plenty of years to
realise it.

Don't think Sparc is dying yet. It has its niche.

It is dying. Oracle have not admitted that public ally, but they got rid
of most of their SPARC chip design and Solaris OS people shortly after
releasing UltraSPARC M8. They have announced that there won't be a
successor to the M8 core and while they have promised that Solaris will
be runnable into the 2030s, they don't seem to know how this will be
accomplished.

Isn't there still a Japanese company still doing Sparc development?

bill

[John Dallman]

Post by Bill Gunshannon
Isn't there still a Japanese company still doing Sparc development?

It was Fujitsu. They've switched to ARM.

John

[Arne Vajhøj]

Post by Bill Gunshannon

Post by John Dallman

Itanium was not DEC's. And Alpha, which was, outlasted Itanium by
a considerable amount of time. Alpha was the success, even with
everyone trying to kill it.

Alpha-based systems stopped being sold in 2007. Final Itanium chip
deliveries will be in the middle of 2021, with the end date for system
sales not yet publicised. I don't quite see how Alpha outlasted Itanium
in the commercial sense.

Because Alpha started long before Itanium and is still in use as Itanium
reaches EOL.  Just count up the years for each of them and compare.

If you compare first delivery to EOL then Itanium has lived longer.

If you compare first delivery to last delivery then Itanium has lived
longer.

If you compare first delivery to last system shutdown, then we don't
know as both are still in usage.

If you compare first deliver to last usage for Alpha and
first delivery to EOL or last delivery of Itanium, then
you are being very misleading.

Arne

[Phillip Helbig (undress to reply)]

Post by Arne Vajhøj
If you compare first delivery to EOL then Itanium has lived longer.
If you compare first delivery to last delivery then Itanium has lived
longer.
If you compare first delivery to last system shutdown, then we don't
know as both are still in usage.
If you compare first deliver to last usage for Alpha and
first delivery to EOL or last delivery of Itanium, then
you are being very misleading.

Let's compare the number of years VMS users were happy with their
hardware. :-)

[Arne Vajhøj]

Post by Phillip Helbig (undress to reply)

Post by Arne Vajhøj
If you compare first delivery to EOL then Itanium has lived longer.
If you compare first delivery to last delivery then Itanium has lived
longer.
If you compare first delivery to last system shutdown, then we don't
know as both are still in usage.
If you compare first deliver to last usage for Alpha and
first delivery to EOL or last delivery of Itanium, then
you are being very misleading.

Let's compare the number of years VMS users were happy with their
hardware. :-)

Alpha may win over Itanium 25-0 in that regard.

It is my clear impression that people liked Alpha and
only bought Itanium because it was the the only
option available if one wanted a new VMS system.

But that does not change the commercial reality.

Arne

[Dave Froble]

Post by Arne Vajhøj

Post by Phillip Helbig (undress to reply)

Post by Arne Vajhøj
If you compare first delivery to EOL then Itanium has lived longer.
If you compare first delivery to last delivery then Itanium has lived
longer.
If you compare first delivery to last system shutdown, then we don't
know as both are still in usage.
If you compare first deliver to last usage for Alpha and
first delivery to EOL or last delivery of Itanium, then
you are being very misleading.

Let's compare the number of years VMS users were happy with their
hardware. :-)

Alpha may win over Itanium 25-0 in that regard.
It is my clear impression that people liked Alpha and
only bought Itanium because it was the the only
option available if one wanted a new VMS system.
But that does not change the commercial reality.
Arne

Well, if you want to go there, it's possible that more people liked VAX
than all the rest put together.

[Scott Dorsey]

Post by Dave Froble
Well, if you want to go there, it's possible that more people liked VAX
than all the rest put together.

People liked vaxen. People loved writing code for the vax.

But compilers... compilers didn't like vaxen so much.
--scott

[Dave Froble]

Post by Scott Dorsey

Post by Dave Froble
Well, if you want to go there, it's possible that more people liked VAX
than all the rest put together.

People liked vaxen. People loved writing code for the vax.
But compilers... compilers didn't like vaxen so much.
--scott

Don't know that I'd agree with that.

I always considered VAX/VMS as a robust development environment. Yeah,
those using VAX C had some issues. Basic, Fortran, Pascal, Cobol, and
such seemed to work rather well.

[Phillip Helbig (undress to reply)]

Post by Dave Froble

Post by Scott Dorsey
People liked vaxen. People loved writing code for the vax.
But compilers... compilers didn't like vaxen so much.
--scott

Don't know that I'd agree with that.
I always considered VAX/VMS as a robust development environment. Yeah,
those using VAX C had some issues. Basic, Fortran, Pascal, Cobol, and
such seemed to work rather well.

I'm most familiar with Fortran. VAX Fortran was the gold standard for
compilers.

I'm currently reading the latest book by Jim Peebles, a cosmologist who
was awarded the Nobel Prize in physics last year. He mentions in the
book that the only programming languages he knows is Fortran. :-) He
was one of the first to do so-called N-body simulations, which are used
to study how structure forms via the action of gravity.

[Scott Dorsey]

Post by Dave Froble

Post by Scott Dorsey

Post by Dave Froble
Well, if you want to go there, it's possible that more people liked VAX
than all the rest put together.

People liked vaxen. People loved writing code for the vax.
But compilers... compilers didn't like vaxen so much.

Don't know that I'd agree with that.
I always considered VAX/VMS as a robust development environment. Yeah,
those using VAX C had some issues. Basic, Fortran, Pascal, Cobol, and
such seemed to work rather well.

Yes, but I could -always- write macro code that ran faster and was tighter
than what the compiler produced. There were always tricks that the compiler
didn't know.

Whereas with the MIPS and later Alpha I was shocked to discover that the
compiler could write better code than I could. The processor was designed
to make the compiler easy, rather than to make manual assembler coding easy.
--scott

[Phillip Helbig (undress to reply)]

Post by Scott Dorsey
Whereas with the MIPS and later Alpha I was shocked to discover that the
compiler could write better code than I could. The processor was designed
to make the compiler easy, rather than to make manual assembler coding easy.

RISC: Relegate Interesting Stuff to the Compiler

[Dave Froble]

Post by Scott Dorsey

Post by Dave Froble

Post by Scott Dorsey

Post by Dave Froble
Well, if you want to go there, it's possible that more people liked VAX
than all the rest put together.

People liked vaxen. People loved writing code for the vax.
But compilers... compilers didn't like vaxen so much.

Don't know that I'd agree with that.
I always considered VAX/VMS as a robust development environment. Yeah,
those using VAX C had some issues. Basic, Fortran, Pascal, Cobol, and
such seemed to work rather well.

Yes, but I could -always- write macro code that ran faster and was tighter
than what the compiler produced. There were always tricks that the compiler
didn't know.

Tricks can sometimes turn around and bite you.

If performance is king for you, then perhaps not all compilers wowuld
fit your needs. Raw performance was never at the top of my list of
"must have". Flexibility, reliability, capability wee much more
important to me.

Post by Scott Dorsey
Whereas with the MIPS and later Alpha I was shocked to discover that the
compiler could write better code than I could. The processor was designed
to make the compiler easy, rather than to make manual assembler coding easy.

That is a dubious claim. Yes, compilers can produce some good code.
But if you tried hard enough you just might out-do them. Of course,
sometimes the effort isn't worth doing. Good compilers are of course a
good thing.

I'm assuming that your MIPS and Alpha code was running under some Unix,
or Linux. Let me ask, what compiler(s) were available to you, and how
good were they on average?

[Simon Clubley]

Post by Dave Froble

Post by Scott Dorsey
Whereas with the MIPS and later Alpha I was shocked to discover that the
compiler could write better code than I could. The processor was designed
to make the compiler easy, rather than to make manual assembler coding easy.

That is a dubious claim. Yes, compilers can produce some good code.

There's nothing dubious about it. Scott is correct and it is a well-known
fact generally.

Post by Dave Froble
But if you tried hard enough you just might out-do them. Of course,
sometimes the effort isn't worth doing. Good compilers are of course a
good thing.
I'm assuming that your MIPS and Alpha code was running under some Unix,
or Linux. Let me ask, what compiler(s) were available to you, and how
good were they on average?

Current compilers have lots of optimisation options and passes and are
sometimes helped by the processor designs Scott talks about. Just look
at the various LLVM optimisation passes for example.

We are no longer in the VAX C compiler era.

Simon.

[John Reagan]

Post by Dave Froble
Well, if you want to go there, it's possible that more people liked VAX
than all the rest put together.

People liked vaxen. People loved writing code for the vax.
But compilers... compilers didn't like vaxen so much.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

I think most of the VAX compilers do OK. BLISS and Pascal in particular do a very good job at using the rich set of operands. COBOL/BASIC probably the least effective at that. And VAX COBOL essentially has zero optimization.

[Andreas Eder]

Post by Bill Gunshannon
Isn't there still a Japanese company still doing Sparc development?

Fujitsu?

'Andreas

[John Dallman]

On Mo 07 Sep 2020 at 09:48, Bill Gunshannon

Post by Bill Gunshannon
Isn't there still a Japanese company still doing Sparc
development?

Fujitsu?

Not any more. They announced in 2016 that they were switching to ARM, and
their first-generation supercomputer using it currently leads the
performance rankings.

John

[Scott Dorsey]

Post by Andreas Eder

Post by Bill Gunshannon
Isn't there still a Japanese company still doing Sparc development?

Fujitsu?

Not any more, they are doing ARM systems now. Once Oracle made it clear
that they didn't give a damn about the sparc, it became clear that Fujitsu
was going to lose their only customer for those chips. They got out fast.
--scott

[^P]

Post by Andreas Eder

Post by Bill Gunshannon
Isn't there still a Japanese company still doing Sparc development?

Fujitsu?

Not any more, they are doing ARM systems now. Once Oracle made it clear
that they didn't give a damn about the sparc, it became clear that Fujitsu
was going to lose their only customer for those chips. They got out fast.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

A bit like when Samsung was about to make Alpha chips

[Bill Gunshannon]

Post by ^P

Post by Andreas Eder

Post by Bill Gunshannon
Isn't there still a Japanese company still doing Sparc development?

Fujitsu?

Not any more, they are doing ARM systems now. Once Oracle made it clear
that they didn't give a damn about the sparc, it became clear that Fujitsu
was going to lose their only customer for those chips. They got out fast.
--scott
--
"C'est un Nagra. C'est suisse, et tres, tres precis."

A bit like when Samsung was about to make Alpha chips

So, I guess the question now is, does this mean my SunBLADE 100 is
more valuable or less valuable? :-)

bill

[Scott Dorsey]

Post by John Dallman

Post by Bill Gunshannon
Don't think Sparc is dying yet. It has its niche.

It is dying. Oracle have not admitted that public ally, but they got rid
of most of their SPARC chip design and Solaris OS people shortly after
releasing UltraSPARC M8. They have announced that there won't be a
successor to the M8 core and while they have promised that Solaris will
be runnable into the 2030s, they don't seem to know how this will be
accomplished.

SPARC is dead. Solaris is heading toward being dead. Oracle bought Sun
because they wanted Java and they didn't know what the hell to do with
the rest of the company. Their attempt to use Oracle-style pricing structures
for Sun hardware and software pretty much killed both new sales and long
term service profits for hardware in place.

The hardware is great. The software is great. They deserved better than
they got.
--scott

[Dave Froble]

Post by Bill Gunshannon

Post by Arne Vajhøj

Post by Bill Gunshannon

Post by Arne Vajhøj

Post by Scott Dorsey
Is that enough to port VMS to ARM? No, but it's worth thinking about because
if things keep going it will be enough in a decade.

That could happen.
If server volume switch from x86-64 to ARM, then VSI would
need to follow the industry.

Was DEC following the industry when they created Alpha and
put VMS on it? Too many followers, not enough leaders.

DEC tried to lead with Alpha and failed.
HP and to some extent Intel trie dto lead with Itanium and failed.

Slight correction, Intel tried to corner the CPU market. AMD shot them
down.

Post by Bill Gunshannon
Itanium was not DEC's. And Alpha, which was, outlasted Itanium by
a considerable amount of time. Alpha was the success, even with
everyone trying to kill it. Itanium was a failure which, sadly, VMS
chose to follow.

Alpha production did not last the forecast 25 years, but usage has done so.

In many minds the itanic was not successful, even though for some time
it was sort of a commercial success.

Post by Bill Gunshannon

Post by Arne Vajhøj
It is often a good thing to lead instead of follow.

Successful leaders reap the profits. "Me-too"s pick up the crumbs.

Post by Bill Gunshannon

Post by Arne Vajhøj
But modern CPU's is a crazy business.
It cost billions to design a CPU. It cost billions
to build fab to manufacture them. And it cost cents
in raw materials to produce one.
What does that mean? It means that volume is
almost everything. Those that sell the most
evolve the most and eventually kill their
competitors.
x86-64 did not kill Alpha, PA, SPARC and Power
(Power is not totally dead yet, but ...)

Don't think Sparc is dying yet. It has its niche. And Power
is far from dead.

Post by Arne Vajhøj
because it was a better design - it won because
Intel and AMD sold hundred of millions of CPU's
for PC's and it was impossible to compete
with that volume.
ARM is not a potential threat to x86-64 because of its design,
but because there is made hundreds of millions of them
for phones and tablets.
If one day (10 years, 15 years or whatever) ARM kills
x86-64 then VSI will not have to migrate.

Don't you mean "will have to migrate"? Any particular reason
why they can't support both if the use is there?

No reason at all, if the product already exists.

[John Dallman]

Post by Dave Froble
In many minds the itanic was not successful, even though for some
time it was sort of a commercial success.

Its announced intentions were to replace x86 and PA-RISC. It did replace
PA-RISC, but got nowhere with replacing x86.

HP had originally come up with the design concept as a replacement for
PA-RISC, but felt it was too expensive to develop by themselves, which
was perfectly true, and enlisted Intel. Intel saw it as a way to escape
competition in the x86 world, which mostly stemmed from old
second-sourcing agreements with AMD. As the dominant CPU manufacturer,
they would have no need for second-source agreements, and could maintain
a monopoly on the architecture.

This all collapsed because of bad decisions about the actual architecture,
which meant that it could not reach the promised performance, and also
took much longer to develop than had been promised. Along the way, Intel
managed to annoy many of the companies they'd enlisted to develop Itanium
software, sufficiently badly to make them stop.

This meant that lots of the promised software didn't exist, so the
hardware didn't sell, so there was no motive to develop software, and the
whole business went into a death spiral. This was clear by 2005, and the
Itanium market after that consisted almost entirely of customers who were
locked into operating systems that ran on nothing else: HP-UX, VMS, and
NonStop.

HP got what they wanted out of it, aided by a great deal of Intel's money.


John

[Arne Vajhøj]

Post by Arne Vajhøj
It is often a good thing to lead instead of follow.

Successful leaders reap the profits.  "Me-too"s pick up the crumbs.

Post by Arne Vajhøj
But modern CPU's is a crazy business.

Often means what it means.

But if you look at the CPU market then the picture
is different.

SUN created SPARC, HP created PA, IBM created Power, DEC
created Alpha, HP and Intel created Itanium.

HP, Dell, IBM/Lenovo etc. sell x86-64 based servers. Apple,
Samsung, Huawei etc. sell ARM based phones/tablets.

Arne

[Arne Vajhøj]

Post by Arne Vajhøj

Post by Bill Gunshannon

Post by Arne Vajhøj

Post by Scott Dorsey
Is that enough to port VMS to ARM?  No, but it's worth thinking
about because
if things keep going it will be enough in a decade.

That could happen.
If server volume switch from x86-64 to ARM, then VSI would
need to follow the industry.

Was DEC following the industry when they created Alpha and
put VMS on it? Too many followers, not enough leaders.

DEC tried to lead with Alpha and failed.
HP and to some extent Intel trie dto lead with Itanium and failed.

Itanium was not DEC's.  And Alpha, which was, outlasted Itanium by
a considerable amount of time.  Alpha was the success, even with
everyone trying to kill it.  Itanium was a failure which, sadly, VMS
chose to follow.

Alpha was introduced in 1992 and was killed in 2001.

Itanium was introduced in 2001 and killed in 2019.

One can of course argue that Alpha was in better shape when
it was killed than Itanium were when it was killed, but
still not many years of Alpha.

Alpha was pretty cool technically, but commercially
it was not a huge success.

Post by Arne Vajhøj
It is often a good thing to lead instead of follow.
But modern CPU's is a crazy business.
It cost billions to design a CPU. It cost billions
to build fab to manufacture them. And it cost cents
in raw materials to produce one.
What does that mean? It means that volume is
almost everything. Those that sell the most
evolve the most and eventually kill their
competitors.
x86-64 did not kill Alpha, PA, SPARC and Power
(Power is not totally dead yet, but ...)

Don't think Sparc is dying yet. It has its niche.  And Power
is far from dead.

SPARC is pretty dead and Power is on life support.

Post by Arne Vajhøj
because it was a better design - it won because
Intel and AMD sold hundred of millions of CPU's
for PC's and it was impossible to compete
with that volume.
ARM is not a potential threat to x86-64 because of its design,
but because there is made hundreds of millions of them
for phones and tablets.
If one day (10 years, 15 years or whatever) ARM kills
x86-64 then VSI will not have to migrate.

Don't you mean "will have to migrate"?

Oops. Yes.

  Any particular reason
why they can't support both if the use is there?

If ARM kills x86-64 they have to migrate.

If ARM never becomes really huge in servers then they will
not migrate.

If we end up having both x86-64 and ARM doing fine for
servers, then they could chose to support both.

But it would be a hassle for everyone.

Arne

[Scott Dorsey]

Post by Bill Gunshannon
Was DEC following the industry when they created Alpha and
put VMS on it? Too many followers, not enough leaders.

No, and that's part of what killed DEC. Trying to lead in too many different
directions at once.

They were pushing Alpha and Vax at the same time they were pushing MIPS...
Divisions were competing with one another when they should have been competing
with Sun...
--scott

[Dave Froble]

Post by Scott Dorsey

Post by Arne Vajhøj
We know that companies like Microsoft, Redhat, SUSE, Canonical
etc. can make money selling OS and/or OS support for the
x86-64 market.
Does the same size companies exist for rPi?
If it does then sure rPi is just as good a bet as x86-64.

The rPi is a low end ARM machine, but there are plenty of high end ARM
machines out there with plenty of support. And now with the Apple move
to ARM for desktop systems, it will be very interesting to see how ARM
goes.
The rPi may be more or less representative of the typical ARM system today
since they have previously found their greatest market in embedded systems,
but don't think it will be representative of the typical ARM system in two
years. ARM is in the server world with systems like the Gigabyte machines,
and they'll be farther in soon.
Is that enough to port VMS to ARM? No, but it's worth thinking about because
if things keep going it will be enough in a decade.
--scott

From what you wrote, Raspburry Pi is just another ARM system. Is that
correct? If so, then perhaps a general port to ARM would indeed run on
Raspberry Pi.

I don't know anything about Raspberry Pi, but, that concept sounds
interesting.

Still, there is the x86 port to complete and make profitable first.

[John Dallman]

Post by Dave Froble
Still, there is the x86 port to complete and make profitable first.

That's definitely first.

Post by Dave Froble
From what you wrote, Raspburry Pi is just another ARM system. Is
that correct? If so, then perhaps a general port to ARM would
indeed run on Raspberry Pi.

The current ARM hardware market is quite peculiar, by the standards of
traditional IT.

Start by ignoring all the tiny 32-bit embedded and realtime systems, and
think about things that are recognizably computers.

There are "server" systems whose selling point is a lot of semi-fast
processor cores for their power budget. These are very useful for cloud
computing, but are annoyingly slow for doing software development on.

The fastest ARM processors, by a fair way, are made by Apple to their own
designs, and sold only inside Apple hardware. Currently, that's in
iPhones, iPads and iPods, but they will be sold in desktop and laptop
Macs within a year or so. Apple don't want you to run other operating
systems on these, unless you use virtualisation.

Then there are ARM-based laptops intended for running the ARM version of
Windows 10. They aren't as fast as Apple devices, and they have very
little market success so far.

There is a vast assortment of Android devices, mostly phones, but also
tablets, slow laptops, watches, entertainment devices, and so on.

Raspberry Pis are single-board computers, using similar hardware to
Android devices, but with expansion ports. They're also very cheap,
around US$50, which makes it sensible to use them as single-purpose
embedded systems, if you don't need huge volumes. There's a wide range of
OSes for them, but the usual one is their own Linux. They're intended as
machines for learning and tinkering, rather than industrial-scale
software production. The big limitation with them is storage, which you
have to add through the expansion ports. Their only built-in storage is a
micro-SD card, which is quite slow.

The top-end Pi is easily comparable with an early Alpha system, with four
64-bit cores at 1.5GHz, 8GB RAM and gigabit Ethernet. The issue with it
running a generic VMS port to ARM would be boot-loaders and device
drivers.

John

[Stephen Hoffman]

Post by John Dallman

Post by Dave Froble
Still, there is the x86 port to complete and make profitable first.

That's definitely first.

Ayup.

Post by John Dallman

Post by Dave Froble
From what you wrote, Raspburry Pi is just another ARM system. Is that
correct? If so, then perhaps a general port to ARM would indeed run on
Raspberry Pi.

The current ARM hardware market is quite peculiar, by the standards of
traditional IT.

In some ways, it's rather closer to the bad old days of every vendor
having their own architectures, though now variants and extensions of
Arm for many.

Post by John Dallman
Start by ignoring all the tiny 32-bit embedded and realtime systems,
and think about things that are recognizably computers.
There are "server" systems whose selling point is a lot of semi-fast
processor cores for their power budget. These are very useful for cloud
computing, but are annoyingly slow for doing software development on.

Arm SBSA and related are intended for server applications, and are a
dependency for mainline Linux support. SBSA makes the Arm processor
permutations more manageable.

The Arm server markets are presently heavily hobbyists with low-end Arm
servers, and the high-end hosting providers; Amazon, Microsoft, and
other large hosting vendors.

https://azure.microsoft.com/en-us/blog/enabling-cloud-workloads-through-innovations-in-silicon/

https://www.anandtech.com/show/15578/cloud-clash-amazon-graviton2-arm-against-intel-and-amd

https://www.opencompute.org/wiki/Server/ProjectOlympus
etc.

Post by John Dallman
Apple don't want you to run other operating systems on these, unless
you use virtualisation.

Call back after the Macs with the Apple silicon Arm processors are
available, maybe?

Whether the new Macs will support SBSA?

Wether Microsoft Windows 10 for ARM (sic) and Linux both get hauled
over to the new Macs?

Post by John Dallman
Then there are ARM-based laptops intended for running the ARM version
of Windows 10. They aren't as fast as Apple devices, and they have very
little market success so far.

Surface Pro X being a salient example. That's Surface Pro X with the
Microsoft SQ1Qualcomm Snapdragon Arm processor.

Not that the Fugaku Arm super that's at the top of the TOP500
supercomputers list is any slouch, nor the the Summit POWER9 super in
second.

Various of the Arm processor providers were rather surprised at how
fast the Apple Arm processors were and are, though. With better
thermals and more power available...

Errata...

Windows 10 IoT Core can be installed on Raspberry Pi.

Other sources of Arm for hobbyists: Pinebook Pro, RockPro64, etc.
https://www.pine64.org/pinebook-pro/
https://www.pine64.org/rockpro64/

[Scott Dorsey]

Post by Dave Froble
From what you wrote, Raspburry Pi is just another ARM system. Is that
correct? If so, then perhaps a general port to ARM would indeed run on
Raspberry Pi.

Yes, the Raspberry Pi is a mid-end ARM processor intended for embedded use
in cellphones, which has been put on a little board with a simple bootloader
in ROM and an SSD with a linux port on it.

It's maybe the performance of a Vax 8600 in terms of overall throughput,
with a lot more main memory, and it costs $50.

A general port to ARM is not likely to include drivers for the Raspberry Pi
peripherals, but it's not that hard to port code to run on the metal of the
thing.

Post by Dave Froble
I don't know anything about Raspberry Pi, but, that concept sounds
interesting.

Get one, it's fifty bucks, and it's a real computer. There are some other
competing devices like the Beaglebone Black as well, but they all use
cellphone ARM processors and they all cost about the price of a good Bordeaux.

Post by Dave Froble
Still, there is the x86 port to complete and make profitable first.

Yes, worrying about anything else with regard to VMS is futile. However, if
you want to build cheap embedded stuff the Pi is very cool.
--scott

[Craig A. Berry]

Post by Arne Vajhøj

Post by Bill Gunshannon

Post by Arne Vajhøj

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi?  :-)

How many licenses do you think they could sell for that?

How many licenses do you think they could sell for x86-64?
Your answer would be no more accurate than mine.
I seriously think there could be a market for an rPi version of VMS.

We know that companies like Microsoft, Redhat, SUSE, Canonical
etc. can make money selling OS and/or OS support for the
x86-64 market.
Does the same size companies exist for rPi?

I believe all of the companies you mention offer their flagship OSs on
ARM. As do Apple and Google. Dunno about Raspberry Pi specifically, but
ARM is already here and all the major OSs already support it (except
macOS, which is about to).

Amazon EC2 offers its own ARM-based servers:

<https://aws.amazon.com/ec2/graviton/>

Of course Intel has a big lead and it was obviously the right target for
VMS when the decision was made to port six years ago. How long it
remains the obvious choice . . . .

[John H. Reinhardt]

Post by Craig A. Berry
I believe all of the companies you mention offer their flagship OSs on
ARM.  As do Apple and Google. Dunno about Raspberry Pi specifically, but
ARM is already here and all the major OSs already support it (except
macOS, which is about to).

The Raspberry Pi has always been an ARM architecture so by default any O/S running on a Pi is ARM-based. This includes Raspberry Pi OS (formerly Raspbian) plus Ubuntu Core, Server and Ubuntu MATE plus some specialty O/S for embedded applications. <https://www.raspberrypi.org/downloads/>

Post by Craig A. Berry
<https://aws.amazon.com/ec2/graviton/>
Of course Intel has a big lead and it was obviously the right target for
VMS when the decision was made to port six years ago. How long it
remains the obvious choice . . . .

[Craig A. Berry]

Post by John H. Reinhardt

Post by Craig A. Berry
I believe all of the companies you mention offer their flagship OSs on
ARM.  As do Apple and Google. Dunno about Raspberry Pi specifically, but
ARM is already here and all the major OSs already support it (except
macOS, which is about to).

The Raspberry Pi has always been an ARM architecture so by default any
O/S running on a Pi is ARM-based.  This includes Raspberry Pi OS
(formerly Raspbian) plus Ubuntu Core, Server and Ubuntu MATE plus some
specialty O/S for embedded applications.
<https://www.raspberrypi.org/downloads/>

Yes, of course. I just meant that, for example, I don't know whether
Windows Server 20XX for ARM runs only in Azure, outside of Azure on some
specified ARM platform that does or does not include Raspberry Pi, etc.
Just about everyone does ARM these days in some fashion. Which may or
may not include Raspbery Pi. Obviously anything that does run on
Raspberry Pi necessarily runs on ARM.

[John H. Reinhardt]

Post by Craig A. Berry
I believe all of the companies you mention offer their flagship OSs on
ARM.  As do Apple and Google. Dunno about Raspberry Pi specifically, but
ARM is already here and all the major OSs already support it (except
macOS, which is about to).

The Raspberry Pi has always been an ARM architecture so by default any O/S running on a Pi is ARM-based.  This includes Raspberry Pi OS (formerly Raspbian) plus Ubuntu Core, Server and Ubuntu MATE plus some specialty O/S for embedded applications. <https://www.raspberrypi.org/downloads/>

Yes, of course.  I just meant that, for example, I don't know whether
Windows Server 20XX for ARM runs only in Azure, outside of Azure on some
specified ARM platform that does or does not include Raspberry Pi, etc.
Just about everyone does ARM these days in some fashion. Which may or
may not include Raspbery Pi. Obviously anything that does run on
Raspberry Pi necessarily runs on ARM.

Sorry. I see where you were going now.

[Dave Froble]

Post by Craig A. Berry

Post by John H. Reinhardt

Post by Craig A. Berry
I believe all of the companies you mention offer their flagship OSs on
ARM. As do Apple and Google. Dunno about Raspberry Pi specifically, but
ARM is already here and all the major OSs already support it (except
macOS, which is about to).

The Raspberry Pi has always been an ARM architecture so by default any
O/S running on a Pi is ARM-based. This includes Raspberry Pi OS
(formerly Raspbian) plus Ubuntu Core, Server and Ubuntu MATE plus some
specialty O/S for embedded applications.
<https://www.raspberrypi.org/downloads/>

Yes, of course. I just meant that, for example, I don't know whether
Windows Server 20XX for ARM runs only in Azure, outside of Azure on some
specified ARM platform that does or does not include Raspberry Pi, etc.
Just about everyone does ARM these days in some fashion. Which may or
may not include Raspbery Pi. Obviously anything that does run on
Raspberry Pi necessarily runs on ARM.

As we've seen multiple times with VMS, the details are in the device
drivers and such. The question about anything running ARM also being
able to run on Raspberry Pi will depend on the devices. Will special
device drivers and such be required?

[John H. Reinhardt]

Post by John H. Reinhardt

Post by Craig A. Berry
I believe all of the companies you mention offer their flagship OSs on
ARM.  As do Apple and Google. Dunno about Raspberry Pi specifically, but
ARM is already here and all the major OSs already support it (except
macOS, which is about to).

The Raspberry Pi has always been an ARM architecture so by default any
O/S running on a Pi is ARM-based.  This includes Raspberry Pi OS
(formerly Raspbian) plus Ubuntu Core, Server and Ubuntu MATE plus some
specialty O/S for embedded applications.
<https://www.raspberrypi.org/downloads/>

Yes, of course.  I just meant that, for example, I don't know whether
Windows Server 20XX for ARM runs only in Azure, outside of Azure on some
specified ARM platform that does or does not include Raspberry Pi, etc.
Just about everyone does ARM these days in some fashion. Which may or
may not include Raspbery Pi. Obviously anything that does run on
Raspberry Pi necessarily runs on ARM.

As we've seen multiple times with VMS, the details are in the device drivers and such.  The question about anything running ARM also being able to run on Raspberry Pi will depend on the devices.  Will special device drivers and such be required?

Yep. The Raspberry Pi is a single board computer, tightly integrated with certain chips for device support. The drivers required would be fairly different from those needed for an server class ARM system with PCIe (or whatever) plug in boards.

[Arne Vajhøj]

Post by Craig A. Berry

Post by Arne Vajhøj

Post by Bill Gunshannon

Post by Arne Vajhøj

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi?  :-)

How many licenses do you think they could sell for that?

How many licenses do you think they could sell for x86-64?
Your answer would be no more accurate than mine.
I seriously think there could be a market for an rPi version of VMS.

We know that companies like Microsoft, Redhat, SUSE, Canonical
etc. can make money selling OS and/or OS support for the
x86-64 market.
Does the same size companies exist for rPi?

I believe all of the companies you mention offer their flagship OSs on
ARM.  As do Apple and Google. Dunno about Raspberry Pi specifically, but
ARM is already here and all the major OSs already support it (except
macOS, which is about to).

They may.

But it is not where they are making their money.

Post by Craig A. Berry
Of course Intel has a big lead and it was obviously the right target for
VMS when the decision was made to port six years ago. How long it
remains the obvious choice . . . .

The only constant is change.

:-)

Arne

[Chris Scheers]

Post by Bill Gunshannon

Post by Arne Vajhøj

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi? :-)

How many licenses do you think they could sell for that?

How many licenses do you think they could sell for x86-64?
Your answer would be no more accurate than mine.
I seriously think there could be a market for an rPi version of VMS.
But only time will tell.

Actually, there is an rPi version of VMS.

Load SIMH and run VAX/VMS. <grin>

Seriously, it actually works well. I have even heard rumors of multiple
rPis being clustered.

For an rPi, SIMH will probably be the only VMS option for a while.

[Bill Gunshannon]

Post by Chris Scheers

Post by Bill Gunshannon

Post by Arne Vajhøj

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi?  :-)

How many licenses do you think they could sell for that?

How many licenses do you think they could sell for x86-64?
Your answer would be no more accurate than mine.
I seriously think there could be a market for an rPi version of VMS.
But only time will tell.

Actually, there is an rPi version of VMS.
Load SIMH and run VAX/VMS.  <grin>

Where you going to get the licenses? <frown>

Post by Chris Scheers
Seriously, it actually works well.  I have even heard rumors of multiple
rPis being clustered.

Unless you want to run something like Java.

Post by Chris Scheers
For an rPi, SIMH will probably be the only VMS option for a while.

Probably true, but not something to cheer about.

bill

[John H. Reinhardt]

Post by Chris Scheers

Post by Bill Gunshannon

Post by Arne Vajhøj

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi?  :-)

How many licenses do you think they could sell for that?

How many licenses do you think they could sell for x86-64?
Your answer would be no more accurate than mine.
I seriously think there could be a market for an rPi version of VMS.
But only time will tell.

Actually, there is an rPi version of VMS.
Load SIMH and run VAX/VMS.  <grin>

Where you going to get the licenses?  <frown>

Until 1-Jan-2022 the HPE Hobbyist ones work quite well.

[Joey Fish]

Post by Bill Gunshannon

Seriously, it actually works well. I have even heard rumors of multiple
rPis being clustered.

Unless you want to run something like Java.

For that, could you include something bigger in the cluster?
A native Itanium/Alpha/Vax install of VMS? An x86 VMS, either native or emulated?
(Or do all computers in the cluster need to be the same architecture?)

[Paul Hardy]

Post by Joey Fish

Post by Bill Gunshannon

Seriously, it actually works well. I have even heard rumors of multiple
rPis being clustered.

Unless you want to run something like Java.

For that, could you include something bigger in the cluster?
A native Itanium/Alpha/Vax install of VMS? An x86 VMS, either native or emulated?
(Or do all computers in the cluster need to be the same architecture?)

I run (sometimes) a hobbyist cluster which includes both Intel PC and
Raspberry Pi hardware, and both VAX and Alpha flavours of VMS. I did have
to intentionally slow down part of the satellite boot process on SimH on
one machine as it was happening faster than VMS could cope!

VMS on a SimH VAX on a Pi 4 is respectably fast - 10.4 VUPs, which for £33
isn’t bad!

[Simon Clubley]

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi? :-)
bill
(Yes, I could really use that about now!! :-)

VMS on ARM isn't going to happen until VSI can either use something
comparable to x86-64's PCID on ARM to efficiently solve the 4-modes on a
2-mode architecture problem or folds the 4 KESU modes into a KU model.

And as per the discussions when this last came up, the address space
attribute on ARM doesn't meet the criteria here.

Simon.

[Dave Froble]

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi? :-)
bill
(Yes, I could really use that about now!! :-)

So, got to wonder what the purchase of ARM by Nvidia will do to ARM usage?

[Simon Clubley]

Post by Dave Froble

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi? :-)
bill
(Yes, I could really use that about now!! :-)

So, got to wonder what the purchase of ARM by Nvidia will do to ARM usage?

Short-term: probably no change.

Medium-term: I'm very uneasy about what Nvidia will do to ARM.

Simon.

[Dave Froble]

Post by Simon Clubley

Post by Dave Froble

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi? :-)
bill
(Yes, I could really use that about now!! :-)

So, got to wonder what the purchase of ARM by Nvidia will do to ARM usage?

Short-term: probably no change.
Medium-term: I'm very uneasy about what Nvidia will do to ARM.
Simon.

Perhaps it's not what Nvidia will do, but the perception of Nvidia's
competitors.

Makes me wonder, why would Nvidia need to own ARM, when ARM makes it's
designs available to everyone? Perhaps some might feel that they no
longer have equal access?

[Simon Clubley]

Post by Dave Froble
Perhaps it's not what Nvidia will do, but the perception of Nvidia's
competitors.
Makes me wonder, why would Nvidia need to own ARM, when ARM makes it's
designs available to everyone? Perhaps some might feel that they no
longer have equal access?

There are some concerns in that area, but we will need to see what
Nvidia actually do.

ARM are unique in what they offer in that they don't merely do
desktop to data centre type systems but they truly have a viable
design for everything from tiny little MCUs (the Cortex-M0) that
can be used in tiny gadgets all the way to full data centre type
systems.

I hope that doesn't change.

Simon.

[Kerry Main]

-----Original Message-----
via Info-vax
Sent: September 14, 2020 10:08 AM
Subject: Re: [Info-vax] VMS on ARM?

Post by Simon Clubley

Post by Dave Froble

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi? :-)
bill
(Yes, I could really use that about now!! :-)

So, got to wonder what the purchase of ARM by Nvidia will do to ARM

usage?

Post by Simon Clubley
Short-term: probably no change.
Medium-term: I'm very uneasy about what Nvidia will do to ARM.
Simon.

Perhaps it's not what Nvidia will do, but the perception of Nvidia's
competitors.
Makes me wonder, why would Nvidia need to own ARM, when ARM makes
it's designs available to everyone? Perhaps some might feel that they no
longer have equal access?

More on Nvidia and ARM -

Sept 13, 2020:
NVIDIA to Acquire Arm for $40 Billion, Creating World's Premier Computing
Company for the Age of AI

<https://nvidianews.nvidia.com/news/nvidia-to-acquire-arm-for-40-billion-cre
ating-worlds-premier-computing-company-for-the-age-of-ai>


Regards,

Kerry Main
Kerry dot main at starkgaming dot com

[Alexander Schreiber]

Post by Bill Gunshannon
Just curious, but, is there any chance there is a Skunk Werks
project for VMS on the Raspberry Pi? :-)

Well, OpenVMS VAX runs just fine on SIMH on ARM, although admittedly
that isn't exactly usable for production workloads.

SCNR,
Alex.