Re: [OT] multicore/cpu history

[Grant Edwards]

On 2018-03-25, Steven D'Aprano  wrote:

> Not really. With multiple CPUs, you have the option of running two 
> distinct OSes in isolation, not merely virtual machines but actual 
> distinct machines in the same box.

Not on any of the multi-CPU motherboards I ever worked with.  The CPUs
shared SDRAM and used the same physical address space.  They both saw
the same PCI/ISA buses and all other peripherals. There was no way you
could run two different OSes without some sort of hypervisor -- there
was no practical difference between them and a modern multi-core CPU.

-- 
Grant Edwards   grant.b.edwardsYow! All of life is a blur
  at   of Republicans and meat!
  gmail.com

-- 
https://mail.python.org/mailman/listinfo/python-list

Re: [OT] multicore/cpu history

[Peter J. Holzer]

On 2018-03-25 22:52:59 +, Steven D'Aprano wrote:
> On Sun, 25 Mar 2018 23:29:07 +0200, Peter J. Holzer wrote:
> >> >> By the way, multiple CPU machines are different from CPUs with
> >> >> multiple cores:
> >> >>
> >> >> http://smallbusiness.chron.com/multiple-cpu-vs-multicore-33195.html
> >> > 
> >> > Yeah, it was always "multiple CPUs", not "multiple cores" when I was
> >> > growing up.
> > 
> > Yes, but the difference is only an implementation detail.
> 
> Not really. With multiple CPUs, you have the option of running two 
> distinct OSes in isolation, not merely virtual machines but actual 
> distinct machines in the same box.

Not in general, no. There may be hardware architectures which allow this
(if I remember correctly, hardware partitioning on HP and IBM unix
machines in the early noughties worked like this), but on a typical PC
motherboard this wouldn't work: There is a lot of shared hardware
outside of the CPUs, and two OSes running on different processors would
have to be aware of each other to avoid stepping on each other's toes.
And if they can do that, they can also do it on two cores of the same
CPU.

In a normal SMP system, there is no real difference between having 2
8-core processors and 1 16-core processor from the OS's point of view.
The scheduler cares about it because caches and NUMA may make migrating
a process from one core to another more expensive depending on where
that other core is physically, but otherwise a core is processor.

hp

-- 
   _  | Peter J. Holzer| we build much bigger, better disasters now
|_|_) || because we have much more sophisticated
| |   | h...@hjp.at | management tools.
__/   | http://www.hjp.at/ | -- Ross Anderson 


signature.asc
Description: PGP signature
-- 
https://mail.python.org/mailman/listinfo/python-list

Re: [OT] multicore/cpu history

[Gene Heskett]

On Monday 26 March 2018 12:12:46 Dennis Lee Bieber wrote:

> On Mon, 26 Mar 2018 11:30:54 -0400, Gene Heskett
> 
>
> declaimed the following:
> >On Monday 26 March 2018 10:06:36 Dennis Lee Bieber wrote:
>
>   
>
> >>As I recall, the bootloader on the Raspberry Pi runs on the
> >> graphics processor, and it sets up the memory image for Linux
> >> before passing control to the ARM processor.
> >
> >Does that come with docs on how to change kernels in case the one you
> > are running has to be rebooted 2-10 times to get the keyboard/mouse
> > ducks all in a row so it doesn't randomly throw away events?
>
>   If Broadcom hasn't changed things, the bootloader is a black-box blob
> provided by Broadcom to the R-Pi foundation. The Linux kernel likely
> falls under regular Linux image management (the bootloader doesn't
> contain the Linux image, just runs on the graphics processor to load
> Linux). I don't think R-Pi uses U-Boot (whereas the BeagleBone Black
> is shoving everything into U-Boot -- making most of the BBB text books
> out-of-date as the kernel no longer loads device tree overlays,
> they've been pre-loaded by U-Boot)

Which to me, isn't a lot of help. But, thats broadcom...

Too bad I can't put a bounty on them. It seems to me that any outfit with 
more lawyers than engineers ought to fall over from top heavy and foot 
damage eventually, but they seem to be the exception to that rule.

OTOH, its even harder to get any usable info out of Pine to facilitate 
using an rtai kernel on a product of theirs called a rock64.  Which can 
build that kernel in under an hour. And it does use u-boot, or claims 
to.  Here we have another credit card sized boy wonder computer thats 
probably 10x or more faster than a top of the line pi, with a usb3 port, 
and the info as to how to make it work seems locked in a safe behind 
non-disses they can't even admit to.

Hell of a way to run a train.  Thanks Dennis.

> --
>   Wulfraed Dennis Lee Bieber AF6VN
>   wlfr...@ix.netcom.comHTTP://wlfraed.home.netcom.com/



-- 
Cheers, Gene Heskett
--
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page 
-- 
https://mail.python.org/mailman/listinfo/python-list

Re: [OT] multicore/cpu history

[Gene Heskett]

On Monday 26 March 2018 10:06:36 Dennis Lee Bieber wrote:

> On Mon, 26 Mar 2018 10:02:15 + (UTC), Steven D'Aprano
>
>  declaimed the following:
> >Hardware people can probably tell you what it is that CPUs do that
> > FPUs and GPUs don't do. Or specialised Bitcoin mining chips.
> > Whatever it is that they don't do, but CPUs do, is probably a good
> > dividing line between "CPU" and "auxiliary chip".
>
>   And then... to confuse matters...
>
>   As I recall, the bootloader on the Raspberry Pi runs on the graphics
> processor, and it sets up the memory image for Linux before passing
> control to the ARM processor.

Does that come with docs on how to change kernels in case the one you are 
running has to be rebooted 2-10 times to get the keyboard/mouse ducks 
all in a row so it doesn't randomly throw away events?
>
> --
>   Wulfraed Dennis Lee Bieber AF6VN
>   wlfr...@ix.netcom.comHTTP://wlfraed.home.netcom.com/



-- 
Cheers, Gene Heskett
--
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page 
-- 
https://mail.python.org/mailman/listinfo/python-list

Re: [OT] multicore/cpu history

[Steven D'Aprano]

On Mon, 26 Mar 2018 10:03:43 +1100, Chris Angelico wrote:

> At what point does it change from being two CPUs to being one CPU and
> one auxiliary processing unit?

When someone writes an OS that will run on the "auxiliary processing 
unit" alone, then it's probably time to start calling it a CPU :-)


> Back in the 80s and early 90s, the
> auxiliary was most likely to be a floating-point unit; today, it'd be a
> graphics chip. But it could just as easily be a Lisp chip.

Or a Forth chip.

Hardware people can probably tell you what it is that CPUs do that FPUs 
and GPUs don't do. Or specialised Bitcoin mining chips. Whatever it is 
that they don't do, but CPUs do, is probably a good dividing line between 
"CPU" and "auxiliary chip".


-- 
Steve

-- 
https://mail.python.org/mailman/listinfo/python-list

Re: [OT] multicore/cpu history

[Chris Angelico]

On Mon, Mar 26, 2018 at 9:52 AM, Steven D'Aprano
 wrote:
> On Sun, 25 Mar 2018 23:29:07 +0200, Peter J. Holzer wrote:
>
> [...]
>>> >> By the way, multiple CPU machines are different from CPUs with
>>> >> multiple cores:
>>> >>
>>> >> http://smallbusiness.chron.com/multiple-cpu-vs-multicore-33195.html
>>> >
>>> > Yeah, it was always "multiple CPUs", not "multiple cores" when I was
>>> > growing up.
>>
>> Yes, but the difference is only an implementation detail.
>
> Not really. With multiple CPUs, you have the option of running two
> distinct OSes in isolation, not merely virtual machines but actual
> distinct machines in the same box. And the CPUs don't necessarily need to
> be the same type, see for example the hybrid Apple Mac/Lisp Machine
> released in the late 1980s or early 90s.

At what point does it change from being two CPUs to being one CPU and
one auxiliary processing unit? Back in the 80s and early 90s, the
auxiliary was most likely to be a floating-point unit; today, it'd be
a graphics chip. But it could just as easily be a Lisp chip.

ChrisA
-- 
https://mail.python.org/mailman/listinfo/python-list

Re: [OT] multicore/cpu history

[Steven D'Aprano]

On Sun, 25 Mar 2018 23:29:07 +0200, Peter J. Holzer wrote:

[...]
>> >> By the way, multiple CPU machines are different from CPUs with
>> >> multiple cores:
>> >>
>> >> http://smallbusiness.chron.com/multiple-cpu-vs-multicore-33195.html
>> > 
>> > Yeah, it was always "multiple CPUs", not "multiple cores" when I was
>> > growing up.
> 
> Yes, but the difference is only an implementation detail.

Not really. With multiple CPUs, you have the option of running two 
distinct OSes in isolation, not merely virtual machines but actual 
distinct machines in the same box. And the CPUs don't necessarily need to 
be the same type, see for example the hybrid Apple Mac/Lisp Machine 
released in the late 1980s or early 90s.

Of course, today, virtualisation is in many practical senses virtually as 
good (pun intended) as actual distinct machines with distinct CPUs, and 
the cost of adding multiple cores is typically cheaper than adding 
multiple CPUs, so there's little point unless you're working with serious 
supercomputer hardware where you need more cores than will fit on a 
single chip.



-- 
Steve

-- 
https://mail.python.org/mailman/listinfo/python-list

Re: [OT] multicore/cpu history

[Peter J. Holzer]

On 2018-02-19 12:39:51 +0100, Adriaan Renting wrote:
> >>> On 17-2-2018 at 22:02, in message
> ,
> Chris Angelico  wrote: 
> > On Sun, Feb 18, 2018 at 5:05 AM, Steven D'Aprano
> >  wrote:
> >> If you're talking about common desktop computers, I think you're
> >> forgetting how recent multicore machines actually are. I'm having
> >> difficulty finding when multicore machines first hit the market,
> >> but it seems to have been well into the 21st century -- perhaps as
> >> late as 2006 with the AMD Athelon 64 X2:
> > 
> > No, I'm talking about big iron. Has Python been running on multi-CPU
> > supercomputers earlier than that?
> > 
> >> By the way, multiple CPU machines are different from CPUs with
> >> multiple cores:
> >>
> >> http://smallbusiness.chron.com/multiple-cpu-vs-multicore-33195.html
> > 
> > Yeah, it was always "multiple CPUs", not "multiple cores" when I was
> > growing up.

Yes, but the difference is only an implementation detail. Once chips
became large enough that you could fit multiple cores on a single chip
it became cheaper (and faster) to do this than put each core on a
separate chip. Just like modern CPUs may contain a floating point unit,
a memory controller, SATA and ethernet controllers, etc. when previously
those would have been separate chips (or even separate cards).

> > And it was only ever in reference to the expensive
> > hardware that I could never even dream of working with. I was always
> > on the single-CPU home-grade systems.
> > 
> 
> Multicore became a thing with the Pentium 4 hyperthreading around ~2002
> for consumers, and
> multi cpu was a thing much longer, even with "consumer grade"
> hardware:
> 
> I remember running 2 Mendocino 300 MHz Celerons on a Pentium II Xeon
> motherboard to get a
> multi-cpu machine for running multiple virtual machines for testing
> purposes around 1998.

I I bought my first multi-processor computer (a dual pentium-90) in
1995. AFAIK the Pentium was the first Intel processor intended for SMP
(there were multiprocessor systems based on earlier intel processors but
they used custom chipsets and were very much not consumer-grade).

> This was not as Intel intended, but a quite cheap consumer grade
> hardware solution.

My dual pentium wasn't cheap, but the motherboard and the second
processor weren't that expensive. The 21" monitor OTOH ...
and did it have a tape drive? I don't remember, but I think it did.

hp

-- 
   _  | Peter J. Holzer| we build much bigger, better disasters now
|_|_) || because we have much more sophisticated
| |   | h...@hjp.at | management tools.
__/   | http://www.hjp.at/ | -- Ross Anderson 


signature.asc
Description: PGP signature
-- 
https://mail.python.org/mailman/listinfo/python-list

[OT] multicore/cpu history Re: Are the critiques in "All the things I hate about Python" valid?

[Adriaan Renting]

Adriaan Renting| Email: rent...@astron.nl
Software Engineer Radio Observatory
ASTRON | Phone: +31 521 595 100 (797 direct)
P.O. Box 2 | GSM:   +31 6 24 25 17 28
NL-7990 AA Dwingeloo   | FAX:   +31 521 595 101
The Netherlands| Web: http://www.astron.nl/~renting/



>>> On 17-2-2018 at 22:02, in message
,
Chris
Angelico  wrote: 
> On Sun, Feb 18, 2018 at 5:05 AM, Steven D'Aprano
>  wrote:
>> On Sat, 17 Feb 2018 15:25:15 +1100, Chris Angelico wrote:
>>

...

>>> Totally not true. The GIL does not stop other threads from
running.
>>> Also, Python has existed for multiple CPU systems pretty much since
its
>>> inception, I believe. (Summoning the D'Aprano for history lesson?)
>>
>> If you're talking about common desktop computers, I think you're
>> forgetting how recent multicore machines actually are. I'm having
>> difficulty finding when multicore machines first hit the market, but
it
>> seems to have been well into the 21st century -- perhaps as late as
2006
>> with the AMD Athelon 64 X2:
> 
> No, I'm talking about big iron. Has Python been running on multi-CPU
> supercomputers earlier than that?
> 
>> By the way, multiple CPU machines are different from CPUs with
multiple
>> cores:
>>
>> http://smallbusiness.chron.com/multiple-cpu-vs-multicore-33195.html
> 
> Yeah, it was always "multiple CPUs", not "multiple cores" when I was
> growing up. And it was only ever in reference to the expensive
> hardware that I could never even dream of working with. I was always
> on the single-CPU home-grade systems.
> 

Multicore became a thing with the Pentium 4 hyperthreading around ~2002
for consumers, and
multi cpu was a thing much longer, even with "consumer grade"
hardware:

I remember running 2 Mendocino 300 MHz Celerons on a Pentium II Xeon
motherboard to get a
multi-cpu machine for running multiple virtual machines for testing
purposes around 1998.
This was not as Intel intended, but a quite cheap consumer grade
hardware solution.

...
> 
> ChrisA

-- 
https://mail.python.org/mailman/listinfo/python-list

Re: [OT] multicore/cpu history Re: Are the critiques in "All the things I hate about Python" valid?

[Chris Angelico]

On Mon, Feb 19, 2018 at 10:39 PM, Adriaan Renting  wrote:
> I remember running 2 Mendocino 300 MHz Celerons on a Pentium II Xeon
> motherboard to get a
> multi-cpu machine for running multiple virtual machines for testing
> purposes around 1998.
> This was not as Intel intended, but a quite cheap consumer grade
> hardware solution.
>

Thanks! That's the sort of thing I was looking for. Out of curiosity,
what was the purpose of that rig and its dual CPUs? Were you running
the host system on one CPU and the guest on another?

ChrisA
-- 
https://mail.python.org/mailman/listinfo/python-list