Printable View
why does vmware and kcpuload say I have 1 prosesor when kpowersave(kde3)
kssyguard(kde4) cpuinfo harddrake say 2
I have a hyperthreding p4
and why wouldn't the 2 cores be balanced, most of the time 1 coar is
flat out and the other one is dooing nothing then they switch.
chris gareau wrote:[color=blue]
> why does vmware and kcpuload say I have 1 prosesor when kpowersave(kde3)
> kssyguard(kde4) cpuinfo harddrake say 2
>
> I have a hyperthreding p4
>
> and why wouldn't the 2 cores be balanced, most of the time 1 coar is
> flat out and the other one is dooing nothing then they switch.[/color]
Perhaps because a hyperthreading P4 has only one cpu, and the
second processing stream (the hyperthread) only comes into play
when the kernel recognizes machine code suitable for separating
into two processing streams and interleaving?
IAMA hyperthread expert, or even knowledgable in that area, but
the above is my impression. If I am wrong, someone will surely
correct me.
Cheers!
jim b.
--
UNIX is not user unfriendly; it merely
expects users to be computer-friendly.
Jim Beard wrote:[color=blue]
> chris gareau wrote:[color=green]
>> why does vmware and kcpuload say I have 1 prosesor when
>> kpowersave(kde3) kssyguard(kde4) cpuinfo harddrake say 2
>>
>> I have a hyperthreding p4
>>
>> and why wouldn't the 2 cores be balanced, most of the time 1 coar is
>> flat out and the other one is dooing nothing then they switch.[/color]
>
> Perhaps because a hyperthreading P4 has only one cpu, and the
> second processing stream (the hyperthread) only comes into play
> when the kernel recognizes machine code suitable for separating
> into two processing streams and interleaving?
>
> IAMA hyperthread expert, or even knowledgable in that area, but
> the above is my impression. If I am wrong, someone will surely
> correct me.
>
> Cheers!
>
> jim b.
>[/color]
all other versions of mandriva I have used on this computer said there
are 2 cpus, the ht shows up as 2 prosesors
On Friday 24 October 2008 02:55, someone identifying as *Jim Beard* wrote
in /alt.os.linux.mandriva:/
[color=blue]
> chris gareau wrote:[color=green]
>> why does vmware and kcpuload say I have 1 prosesor when kpowersave(kde3)
>> kssyguard(kde4) cpuinfo harddrake say 2
>>
>> I have a hyperthreding p4
>>
>> and why wouldn't the 2 cores be balanced, most of the time 1 coar is
>> flat out and the other one is dooing nothing then they switch.[/color]
>
> Perhaps because a hyperthreading P4 has only one cpu, and the
> second processing stream (the hyperthread) only comes into play
> when the kernel recognizes machine code suitable for separating
> into two processing streams and interleaving?[/color]
No, that is not correct. If the processor supports hyperthreading and the
kernel supports SMP, both hyperthreads will be seen as physical CPUs by the
system.
The additional SMT kernel support allows for the kernel to better balance
the load across hyperthreads and physical processors in a multi-socket
system, but with only a single socket, SMT support is not really required.
What *is* required however is that you have just enough ACPI support -
through the /acpi=ht/ or /acpi=on/ kernel boot parameters - to enable the
hyperthreading - /acpi=off/ will disable hyperthreading support at the
kernel level - and of course, the BIOS must be set-up to allow
hyperthreading as well.
[color=blue]
> IAMA hyperthread expert, or even knowledgable in that area, but
> the above is my impression. If I am wrong, someone will surely
> correct me.[/color]
Quod erat demonstrandum. :p
--
*Aragorn*
(registered GNU/Linux user #223157)
On Friday 24 October 2008 01:12, someone identifying as *chris gareau* wrote
in /alt.os.linux.mandriva:/
[color=blue]
> why does vmware and kcpuload say I have 1 prosesor when kpowersave(kde3)
> kssyguard(kde4) cpuinfo harddrake say 2
>
> I have a hyperthreding p4[/color]
A hyperthreaded Pentium 4 is not a dualcore processor. Hyperthreading only
makes the processor appear as two processors to the operating system, but
it is not the same thing as a dualcore processor.
See, with hyperthreading, the operating system believes the processor is in
fact two processors and will thus delegate two threads at the same time to
these two processors. In reality however, the processor simply has extra
registers to trap those two threads, and then one thread will become
executed, while the second thread is simply put on hold until the first
thread needs to wait for some I/O or memory operation, in which case this
first thread will then be put on hold and the second thread will be
executed in the meantime.
So in other words, it's not true parallel processing as is the case with
multicore or multisocket machines. The kernel thinks so, but the hardware
is just faking it.
[color=blue]
> and why wouldn't the 2 cores be balanced, most of the time 1 coar is
> flat out and the other one is dooing nothing then they switch.[/color]
Because they are not cores, they are thread registers. The processor takes
two instructions at the same time - just as if you had two processors or
two processor cores, which would then also allow the kernel to delegate two
instructions to the processors - but these threads are not executed in
parallel. Only one of them is executed at the time, and the processor will
only switch to the second thread when the first thread needs to wait for
some memory read/write or I/O.
In addition to all of the above, I do not know whether the version of VMWare
that comes with Mandriva actually even supports multiprocessing for guest
virtual machines. This in itself is also an important factor in this
equation.
--
*Aragorn*
(registered GNU/Linux user #223157)
On Fri, 24 Oct 2008 16:56:06 +0200, Aragorn wrote:
[color=blue]
> So in other words, it's not true parallel processing as is the case with
> multicore or multisocket machines.[/color]
Those aren't doing true parallel processing either. That's why the
kernel says SMP, for serial multi-processing.
On Friday 24 October 2008 19:07, someone identifying as *Mark Madsen* wrote
in /alt.os.linux.mandriva:/
[color=blue]
> On Fri, 24 Oct 2008 16:56:06 +0200, Aragorn wrote:
>[color=green]
>> So in other words, it's not true parallel processing as is the case with
>> multicore or multisocket machines.[/color]
>
> Those aren't doing true parallel processing either. That's why the
> kernel says SMP, for serial multi-processing.[/color]
I'm afraid you are mistaken. SMP is an acronym for Synchronous
Multi-Processing, and it is by all means parallel.
Of course, the kernel needs to correctly schedule this parallel processing
and this will cause some overhead as opposed to the idea that a system with
two processors would be twice as fast as a system with one processor, but
this overhead is fairly minimal, and especially so with the Linux kernel.
--
*Aragorn*
(registered GNU/Linux user #223157)
On Fri, 24 Oct 2008 19:32:16 +0200, Aragorn wrote:
[color=blue]
> On Friday 24 October 2008 19:07, someone identifying as *Mark Madsen*
> wrote in /alt.os.linux.mandriva:/
>[color=green]
>> On Fri, 24 Oct 2008 16:56:06 +0200, Aragorn wrote:
>>[color=darkred]
>>> So in other words, it's not true parallel processing as is the case
>>> with multicore or multisocket machines.[/color]
>>
>> Those aren't doing true parallel processing either. That's why the
>> kernel says SMP, for serial multi-processing.[/color]
>
> I'm afraid you are mistaken. SMP is an acronym for Synchronous
> Multi-Processing, and it is by all means parallel.[/color]
That "synchronous" is a back formation to make it sound better. Look
back through the literature on multi-processing and you'll see.
[color=blue]
> Of course, the kernel needs to correctly schedule this parallel
> processing and this will cause some overhead as opposed to the idea that
> a system with two processors would be twice as fast as a system with one
> processor, but this overhead is fairly minimal, and especially so with
> the Linux kernel.[/color]
True parallel processing requires completely different hardware and
different ways of writing the code. There are no true parallel
processing systems available to the consumer.
Multi-processing is not parallel processing. They are different beasts.
On Friday 24 October 2008 20:21, someone identifying as *Mark Madsen* wrote
in /alt.os.linux.mandriva:/
[color=blue]
> On Fri, 24 Oct 2008 19:32:16 +0200, Aragorn wrote:
>[color=green]
>> On Friday 24 October 2008 19:07, someone identifying as *Mark Madsen*
>> wrote in /alt.os.linux.mandriva:/
>>[color=darkred]
>>> On Fri, 24 Oct 2008 16:56:06 +0200, Aragorn wrote:
>>>
>>>> So in other words, it's not true parallel processing as is the case
>>>> with multicore or multisocket machines.
>>>
>>> Those aren't doing true parallel processing either. That's why the
>>> kernel says SMP, for serial multi-processing.[/color]
>>
>> I'm afraid you are mistaken. SMP is an acronym for Synchronous
>> Multi-Processing, and it is by all means parallel.[/color]
>
> That "synchronous" is a back formation to make it sound better. Look
> back through the literature on multi-processing and you'll see.[/color]
[url]http://en.wikipedia.org/wiki/Symmetric_multiprocessing[/url]
[color=blue][color=green]
>> Of course, the kernel needs to correctly schedule this parallel
>> processing and this will cause some overhead as opposed to the idea that
>> a system with two processors would be twice as fast as a system with one
>> processor, but this overhead is fairly minimal, and especially so with
>> the Linux kernel.[/color]
>
> True parallel processing requires completely different hardware and
> different ways of writing the code. There are no true parallel
> processing systems available to the consumer.[/color]
It all depends on your definition of parallel processing. You're probably
thinking of mainframes, clusters and supercomputers, but I wrote my reply
to the OP based upon his misinterpretation of the concept of parallelism
with regard to hyperthreading.
[color=blue]
> Multi-processing is not parallel processing. They are different beasts.[/color]
Multi-processing *is* a form of parallel processing. Electronically, the
individual processors operate in parallel. What I will grant you is that
the rest of the system does not - although there is a slightly higher
degree of parallelism with AMD's multisocket machines since they use a
ccNUMA design, contrary to Intel's current batch of (x86(-64)) multisocket
machines - but this falls beyond the scope of what the OP was alluding to,
and what I was replying to.
Within that particular scope, multicore and multisocket machines can be
considered parallel processing, and multithreading is fake parallel
processing, in that the operating system thinks that it's parallel but in
reality the processor serializes it.
--
*Aragorn*
(registered GNU/Linux user #223157)
On Fri, 24 Oct 2008 20:41:55 +0200, Aragorn wrote:
[color=blue][color=green]
>> Multi-processing is not parallel processing. They are different
>> beasts.[/color]
>
> Multi-processing *is* a form of parallel processing.[/color]
Parallel processing is a form of multi-processing. A special form.
X is a subset of Y and Y is a subset of X are different statements.
Redefining "parallel processing" to mean "multi-processing" is a
marketing sleight of hand. And false.
Aragorn wrote:[color=blue]
> On Friday 24 October 2008 01:12, someone identifying as *chris gareau* wrote
> in /alt.os.linux.mandriva:/
>[color=green]
>> why does vmware and kcpuload say I have 1 prosesor when kpowersave(kde3)
>> kssyguard(kde4) cpuinfo harddrake say 2
>>
>> I have a hyperthreding p4[/color]
>
> A hyperthreaded Pentium 4 is not a dualcore processor. Hyperthreading only
> makes the processor appear as two processors to the operating system, but
> it is not the same thing as a dualcore processor.
>
> See, with hyperthreading, the operating system believes the processor is in
> fact two processors and will thus delegate two threads at the same time to
> these two processors. In reality however, the processor simply has extra
> registers to trap those two threads, and then one thread will become
> executed, while the second thread is simply put on hold until the first
> thread needs to wait for some I/O or memory operation, in which case this
> first thread will then be put on hold and the second thread will be
> executed in the meantime.
>
> So in other words, it's not true parallel processing as is the case with
> multicore or multisocket machines. The kernel thinks so, but the hardware
> is just faking it.
>[color=green]
>> and why wouldn't the 2 cores be balanced, most of the time 1 coar is
>> flat out and the other one is dooing nothing then they switch.[/color]
>
> Because they are not cores, they are thread registers. The processor takes
> two instructions at the same time - just as if you had two processors or
> two processor cores, which would then also allow the kernel to delegate two
> instructions to the processors - but these threads are not executed in
> parallel. Only one of them is executed at the time, and the processor will
> only switch to the second thread when the first thread needs to wait for
> some memory read/write or I/O.
>
> In addition to all of the above, I do not know whether the version of VMWare
> that comes with Mandriva actually even supports multiprocessing for guest
> virtual machines. This in itself is also an important factor in this
> equation.
>[/color]
I am running vmware server from vmware not from mandriva
On Saturday 25 October 2008 00:54, someone identifying as *chris gareau*
wrote in /alt.os.linux.mandriva:/
[color=blue]
> Aragorn wrote:
>[color=green]
>> On Friday 24 October 2008 01:12, someone identifying as *chris gareau*
>> wrote in /alt.os.linux.mandriva:/
>>[color=darkred]
>>> and why wouldn't the 2 cores be balanced, most of the time 1 coar is
>>> flat out and the other one is dooing nothing then they switch.[/color]
>>
>> Because they are not cores, they are thread registers. The processor
>> takes two instructions at the same time - just as if you had two
>> processors or two processor cores, which would then also allow the kernel
>> to delegate two instructions to the processors - but these threads are
>> not executed in parallel. Only one of them is executed at the time, and
>> the processor will only switch to the second thread when the first thread
>> needs to wait for some memory read/write or I/O.
>>
>> In addition to all of the above, I do not know whether the version of
>> VMWare that comes with Mandriva actually even supports multiprocessing
>> for guest virtual machines. This in itself is also an important factor
>> in this equation.[/color]
>
> I am running vmware server from vmware not from mandriva[/color]
Still, there is VMWare Player and there is VMWare Workstation - the Server
versions are a different critter altogether - and whether they support
multiprocessing for guest operating systems is something you should verify
from reading the documentation.
However, what I wrote higher up still applies. Hyperthreading isn't real
SMP. It only /looks/ like SMP to the operating system, but at the hardware
level, it's just a more specialized form of multitasking within the same
processor.
In fact, Intel's performance claims regarding hyperthreading - on a
processor that has two hyperthreads, because they now have processors with
four hyperthreads on the way - is that with hyperthreading enabled, the
processor's performance is increased with about 35% max, which is quite
below the performance characteristics of dualcore processors or multisocket
machines.
--
*Aragorn*
(registered GNU/Linux user #223157)
Aragorn wrote:[color=blue]
> On Saturday 25 October 2008 00:54, someone identifying as *chris gareau*
> wrote in /alt.os.linux.mandriva:/
>[color=green]
>> Aragorn wrote:
>>[color=darkred]
>>> On Friday 24 October 2008 01:12, someone identifying as *chris gareau*
>>> wrote in /alt.os.linux.mandriva:/
>>>
>>>> and why wouldn't the 2 cores be balanced, most of the time 1 coar is
>>>> flat out and the other one is dooing nothing then they switch.
>>> Because they are not cores, they are thread registers. The processor
>>> takes two instructions at the same time - just as if you had two
>>> processors or two processor cores, which would then also allow the kernel
>>> to delegate two instructions to the processors - but these threads are
>>> not executed in parallel. Only one of them is executed at the time, and
>>> the processor will only switch to the second thread when the first thread
>>> needs to wait for some memory read/write or I/O.
>>>
>>> In addition to all of the above, I do not know whether the version of
>>> VMWare that comes with Mandriva actually even supports multiprocessing
>>> for guest virtual machines. This in itself is also an important factor
>>> in this equation.[/color]
>> I am running vmware server from vmware not from mandriva[/color]
>
> Still, there is VMWare Player and there is VMWare Workstation - the Server
> versions are a different critter altogether - and whether they support
> multiprocessing for guest operating systems is something you should verify
> from reading the documentation.
>
> However, what I wrote higher up still applies. Hyperthreading isn't real
> SMP. It only /looks/ like SMP to the operating system, but at the hardware
> level, it's just a more specialized form of multitasking within the same
> processor.
>
> In fact, Intel's performance claims regarding hyperthreading - on a
> processor that has two hyperthreads, because they now have processors with
> four hyperthreads on the way - is that with hyperthreading enabled, the
> processor's performance is increased with about 35% max, which is quite
> below the performance characteristics of dualcore processors or multisocket
> machines.
>[/color]
vmware server will emulate smp for the guest if I ask for it, I am
talking about the host
On 2008-10-24, Aragorn <aragorn@chatfactory.invalid> wrote:[color=blue]
> On Friday 24 October 2008 20:21, someone identifying as *Mark Madsen* wrote
> in /alt.os.linux.mandriva:/
>[color=green]
>> On Fri, 24 Oct 2008 19:32:16 +0200, Aragorn wrote:
>>[color=darkred]
>>> On Friday 24 October 2008 19:07, someone identifying as *Mark Madsen*
>>> wrote in /alt.os.linux.mandriva:/
>>>
>>>> On Fri, 24 Oct 2008 16:56:06 +0200, Aragorn wrote:
>>>>
>>>>> So in other words, it's not true parallel processing as is the case
>>>>> with multicore or multisocket machines.
>>>>
>>>> Those aren't doing true parallel processing either. That's why the
>>>> kernel says SMP, for serial multi-processing.
>>>
>>> I'm afraid you are mistaken. SMP is an acronym for Synchronous
>>> Multi-Processing, and it is by all means parallel.[/color]
>>
>> That "synchronous" is a back formation to make it sound better. Look
>> back through the literature on multi-processing and you'll see.[/color]
>
> [url]http://en.wikipedia.org/wiki/Symmetric_multiprocessing[/url][/color]
Aragorn, thank you for correcting what the 'S' in 'SMP'
stands for: Symmetric.
--
Robert Riches
[email]spamtrap42@verizon.net[/email]
(Yes, that is one of my email addresses.)
On Saturday 25 October 2008 05:53, someone identifying as *Robert Riches*
wrote in /alt.os.linux.mandriva:/
[color=blue]
> On 2008-10-24, Aragorn <aragorn@chatfactory.invalid> wrote:[color=green]
>> On Friday 24 October 2008 20:21, someone identifying as *Mark Madsen*
>> wrote in /alt.os.linux.mandriva:/
>>[color=darkred]
>>> On Fri, 24 Oct 2008 19:32:16 +0200, Aragorn wrote:
>>>
>>>> On Friday 24 October 2008 19:07, someone identifying as *Mark Madsen*
>>>> wrote in /alt.os.linux.mandriva:/
>>>>
>>>>> On Fri, 24 Oct 2008 16:56:06 +0200, Aragorn wrote:
>>>>>
>>>>>> So in other words, it's not true parallel processing as is the case
>>>>>> with multicore or multisocket machines.
>>>>>
>>>>> Those aren't doing true parallel processing either. That's why the
>>>>> kernel says SMP, for serial multi-processing.
>>>>
>>>> I'm afraid you are mistaken. SMP is an acronym for Synchronous
>>>> Multi-Processing, and it is by all means parallel.
>>>
>>> That "synchronous" is a back formation to make it sound better. Look
>>> back through the literature on multi-processing and you'll see.[/color]
>>
>> [url]http://en.wikipedia.org/wiki/Symmetric_multiprocessing[/url][/color]
>
> Aragorn, thank you for correcting what the 'S' in 'SMP'
> stands for: Symmetric.[/color]
Well, I actually already knew that, but somehow I must have had a brain fart
when I said "synchronous". While that was not entirely correct, it did
however come closer to the truth than Mark's "serial". ;-)
--
*Aragorn*
(registered GNU/Linux user #223157)
On Saturday 25 October 2008 05:19, someone identifying as *chris gareau*
wrote in /alt.os.linux.mandriva:/
[color=blue]
> Aragorn wrote:
>[color=green]
>> On Saturday 25 October 2008 00:54, someone identifying as *chris gareau*
>> wrote in /alt.os.linux.mandriva:/
>>[color=darkred]
>>> Aragorn wrote:
>>>
>>>> On Friday 24 October 2008 01:12, someone identifying as *chris gareau*
>>>> wrote in /alt.os.linux.mandriva:/
>>>>
>>>>> and why wouldn't the 2 cores be balanced, most of the time 1 coar is
>>>>> flat out and the other one is dooing nothing then they switch.
>>>> Because they are not cores, they are thread registers. The processor
>>>> takes two instructions at the same time - just as if you had two
>>>> processors or two processor cores, which would then also allow the
>>>> kernel to delegate two instructions to the processors - but these
>>>> threads are not executed in parallel. Only one of them is executed at
>>>> the time, and the processor will only switch to the second thread when
>>>> the first thread needs to wait for some memory read/write or I/O.
>>>>
>>>> In addition to all of the above, I do not know whether the version of
>>>> VMWare that comes with Mandriva actually even supports multiprocessing
>>>> for guest virtual machines. This in itself is also an important factor
>>>> in this equation.
>>>
>>> I am running vmware server from vmware not from mandriva[/color]
>>
>> Still, there is VMWare Player and there is VMWare Workstation - the
>> Server versions are a different critter altogether - and whether they
>> support multiprocessing for guest operating systems is something you
>> should verify from reading the documentation.
>>
>> However, what I wrote higher up still applies. Hyperthreading isn't real
>> SMP. It only /looks/ like SMP to the operating system, but at the
>> hardware level, it's just a more specialized form of multitasking within
>> the same processor.
>>
>> In fact, Intel's performance claims regarding hyperthreading - on a
>> processor that has two hyperthreads, because they now have processors
>> with four hyperthreads on the way - is that with hyperthreading enabled,
>> the processor's performance is increased with about 35% max, which is
>> quite below the performance characteristics of dualcore processors or
>> multisocket machines.
>>[/color]
> vmware server will emulate smp for the guest if I ask for it, I am
> talking about the host[/color]
Well, I have explained to you why the process scheduling is not balanced
properly on your two virtual processors, i.e. they are virtual, not real,
and their thread execution scheduling depends on how fast the other thread
can be processed.
Hyperthreads are not multiple cores.
--
*Aragorn*
(registered GNU/Linux user #223157)
Jim Beard wrote:
[color=blue]
> chris gareau wrote:[color=green]
>> why does vmware and kcpuload say I have 1 prosesor when kpowersave(kde3)
>> kssyguard(kde4) cpuinfo harddrake say 2
>>
>> I have a hyperthreding p4
>>
>> and why wouldn't the 2 cores be balanced, most of the time 1 coar is
>> flat out and the other one is dooing nothing then they switch.[/color]
>
> Perhaps because a hyperthreading P4 has only one cpu, and the
> second processing stream (the hyperthread) only comes into play
> when the kernel recognizes machine code suitable for separating
> into two processing streams and interleaving?
>
> IAMA hyperthread expert, or even knowledgable in that area, but
> the above is my impression. If I am wrong, someone will surely
> correct me.
>
> Cheers!
>
> jim b.
>[/color]
A hyper-threaded cpu like the P4 has some duplicated resources and some
shared resources. Certain instruction blocks can be divided up in such
a way as to take advantage of the duplicated resources. Instead of having
part of the cpu internals sitting idle while an instruction is being processed
those internal resources that are idle can, sometimes, be used to work on
another instruction at the same time. When instruction ordering lends itself to
this situation, you get a performance boost
Here's a link to HT that might be useful:
[url]http://software.intel.com/en-us/articles/introduction-to-hyper-threading-technology[/url]
Chris has a single, hyper-threaded cpu, to call it 2 cpu's is semantics. /proc/cpuinfo
is reporting the 2 threads. At one time there was a plan to add several logical threads
to a cpu, so, for example you might have a p4 with 4 threads. Still a single cpu but with
four logical processors. It didn't pan out. Intel moved to multicore instead - a much much
better choice. mutlicore puts more than one distinct and complete processor in a package
(except for shared cache).
Eric
On Fri, 24 Oct 2008 16:56:06 +0200
Aragorn <aragorn@chatfactory.invalid> wrote:
[...][color=blue]
> So in other words, it's not true parallel processing as is the case with
> multicore or multisocket machines. The kernel thinks so, but the hardware
> is just faking it.[/color]
[...]
Yeah, hyperthreading sounds like sex with a prostitute. You think the moans
mean she is having a good time, but she's just faking it. ;)
Based on this very reason, thanks to Aragorn for simplifying it so well, I have
never been interested in having a hyperthreading CPU. Give me real multi-core
or multiple CPUs instead.
Gene (e-mail: usenet0 \a\t eracc \d\o\t com)
blog.eracc.com latest article: [url]http://preview.tinyurl.com/64cbxc[/url]
'Mandriva Linux Used to Save a XP Professional PC'
--
Mandriva Linux release 2008.1 (Official) for i586
Got Rute? [url]http://www.anrdoezrs.net/email-2546588-42121?isbn=0130333514[/url]
ERA Computers & Consulting - [url]http://www.eracc.com/[/url]
Preloaded PCs - eComStation, Linux, FreeBSD, OpenServer & UnixWare
On Sat, 25 Oct 2008 16:29:21 +0200, Aragorn wrote:
[color=blue][color=green]
>> Aragorn, thank you for correcting what the 'S' in 'SMP' stands for:
>> Symmetric.[/color]
>
> Well, I actually already knew that, but somehow I must have had a brain
> fart when I said "synchronous". While that was not entirely correct, it
> did however come closer to the truth than Mark's "serial". ;-)[/color]
Revisionist terminology tells you nothing about history. It was
originally referred to as serial multi-processing to make clear the
contrast with parallel processing. The fact that there are several other
words starting with S is an artefact of the language.
Anyway, you guys can work out your own history lessons, or believe
whatever you like of marketing speak. It still won't turn any kind of
SMP (regardless of what you believe the S stands for) into parallel
processing.
Mark Madsen wrote:[color=blue]
> Anyway, you guys can work out your own history lessons, or believe
> whatever you like of marketing speak. It still won't turn any kind of
> SMP (regardless of what you believe the S stands for) into parallel
> processing.[/color]
Mark,
I have an AMD Athlon(tm) 64-bit X2 Dual Core Processor 5000+. The
thing has two cores ("two cpus" if you will), and one can be
executing one process while the other is executing another. This
IS parallel processing, simultaneous execution of two or more
processes -- whether you like the exact meaning of the English
language words or not.
There are some forms of "parallel processing" that my machine will
not do. But I think all following this thread and responding are
aware of that.
No cheers for pettifloggery.
jim b.
--
UNIX is not user unfriendly; it merely
expects users to be computer-friendly.