[OT??] {32 vs 64 bit} &/or {single vs multi core} ???

{also posted to a Linux group}

In the "real world", is there a difference for an 
average/typical user.
I'm contemplating the move from Windows to Linux.
Coward that I am, I'm considering separate hardware as testbed.
I spend most of my time in text editing and web surfing - 
doubt any advantage there.
My other interest include Scilab, Scioslab, and gnuplot. I 
would run and display display results of *LARGE* fft's of 
_stored_ data. "Real time" minor importance.

Are there subtle advantages to wider data path and/or 
multiple cores?
Are there advantages to 64 bit implementations of Linux?

Thanks in advance.

On Fri, 25 Nov 2011 20:02:41 -0600, Richard Owlett wrote:

> {also posted to a Linux group}
> 
> In the "real world", is there a difference for an average/typical user.
> I'm contemplating the move from Windows to Linux. Coward that I am, I'm
> considering separate hardware as testbed. I spend most of my time in
> text editing and web surfing - doubt any advantage there.
> My other interest include Scilab, Scioslab, and gnuplot. I would run and
> display display results of *LARGE* fft's of _stored_ data. "Real time"
> minor importance.
> 
> Are there subtle advantages to wider data path and/or multiple cores?
> Are there advantages to 64 bit implementations of Linux?

Scilab uses both cores of my two-core computer.  Before they upgraded it 
went quite perceptibly slower -- the speedup wasn't 2:1 by any means, but 
it was significant (and nice).

-- 
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
Why am I not happy that they have found common ground?

Tim Wescott, Communications, Control, Circuits & Software
http://www.wescottdesign.com

Tim Wescott wrote:
> On Fri, 25 Nov 2011 20:02:41 -0600, Richard Owlett wrote:
>
>> {also posted to a Linux group}
>>
>> In the "real world", is there a difference for an average/typical user.
>> I'm contemplating the move from Windows to Linux. Coward that I am, I'm
>> considering separate hardware as testbed. I spend most of my time in
>> text editing and web surfing - doubt any advantage there.
>> My other interest include Scilab, Scioslab, and gnuplot. I would run and
>> display display results of *LARGE* fft's of _stored_ data. "Real time"
>> minor importance.
>>
>> Are there subtle advantages to wider data path and/or multiple cores?
>> Are there advantages to 64 bit implementations of Linux?
>
> Scilab uses both cores of my two-core computer.  Before they upgraded it
> went quite perceptibly slower -- the speedup wasn't 2:1 by any means, but
> it was significant (and nice).
>

Is that with Scilab 4.x or 5.x?
I had stayed away from  5.x due to it's early problems and 
the size of it's download.
I curently have have Scilab 4.? and Scioslab 4.?? on my 
machine although I've not done any FFT's for months.

Richard Owlett <rowlett@pcnetinc.com> wrote:
> {also posted to a Linux group}

> In the "real world", is there a difference for an 
> average/typical user.

For x86, 64 bit gets you more registers, but also larger
addresses (pointers).  If you aren't using more than about 2GB
(the most that some systems will give you in 32 bit mode) then
there is a good chance that 32 bit is the better choice.

Well, you can run 32 bit programs in most 64 bit OSs.

> I'm contemplating the move from Windows to Linux.
> Coward that I am, I'm considering separate hardware as testbed.
> I spend most of my time in text editing and web surfing - 
> doubt any advantage there.

You could also run one with VMware inside the other one.

> My other interest include Scilab, Scioslab, and gnuplot. I 
> would run and display display results of *LARGE* fft's of 
> _stored_ data. "Real time" minor importance.

Some software companies might charge more for 64 bit versions.

> Are there subtle advantages to wider data path and/or 
> multiple cores?

Multiple cores should work in either 32 or 64 bit.

> Are there advantages to 64 bit implementations of Linux?

-- glen

On 26 Nov, 03:02, Richard Owlett <rowl...@pcnetinc.com> wrote:
> {also posted to a Linux group}
>
> In the "real world", is there a difference for an
> average/typical user.

Somebody once said that 'moving from single-core
to duo-core PC left one core to deal with viruses
and the other to do work.' I never really got to
know if he meant 'search for viruses' or 'let
viruses do their mischief'...

> I'm contemplating the move from Windows to Linux.
> Coward that I am, I'm considering separate hardware as testbed.

Very wise.

> I spend most of my time in text editing and web surfing -
> doubt any advantage there.
> My other interest include Scilab, Scioslab, and gnuplot. I
> would run and display display results of *LARGE* fft's of
> _stored_ data. "Real time" minor importance.
>
> Are there subtle advantages to wider data path and/or
> multiple cores?

Multiple cores is a mess. One really don't get the benefit
until one parallelizes the problem, which is a mess in
itself. FFTs and linear algebra are problems that can be
parallelized relatively easily, but then the question is
how well this has been done in the octave/scilab programs
you use. This is the kind of thing that likely has to be
fine tuned very accurately, or they introduce enough
overhead on the parallelization to squander the benefit
of parallelization.

> Are there advantages to 64 bit implementations of Linux?

Naively, 64-bit systems might offer larger data sets
beyond the 2GB limit of 32-bit systems. However, that
might not be the case in practice. I don't know if
one can take for granted that 64-bit systems actually
use 64-bit (or at least > 32 bit) address pointers.
It's something about the economy of cost/benefit;
you will need to have a *lot* more than 2GB RAM
for 64-bit pointers to increase memory capacity, and
using pointer sizes between 32 and 64 bits will be
an engineering challenge.

So do your research on exactly what a 64 bit system
offers, before you commit.

Rune

On Sat, 26 Nov 2011 03:12:41 -0800, Rune Allnor wrote:
>> Are there advantages to 64 bit implementations of Linux?
> 
> Naively, 64-bit systems might offer larger data sets beyond the 2GB
> limit of 32-bit systems. However, that might not be the case in
> practice. I don't know if one can take for granted that 64-bit systems
> actually use 64-bit (or at least > 32 bit) address pointers.
> It's something about the economy of cost/benefit;
> you will need to have a *lot* more than 2GB RAM for 64-bit pointers to
> increase memory capacity, and using pointer sizes between 32 and 64 bits
> will be an engineering challenge.

64-bit costs from the extra d-cache space consumed by double-size 
pointers, and that's about the only down-side.  If your code is dominated 
by array accesses rather than pointer-dense data structures, then there 
is little difference in that regard.

On the other hand, the 64-bit instruction set has 16 general-purpose 
registers instead of eight, and with it significantly more efficient ABI 
(calling convention) schemes, so most code is actually a bit faster.  As 
a freebie, all of your "int64_t/long long" arithmetic, such as timestamps 
and file offsets now happens in-line rather than though runtime support 
extended precision arithmetic libraries.  From a DSP point of view, 64-
bit accumulators are just about exactly right for a mix of 16 and 32-bit 
data and coefficients... ;-)

> So do your research on exactly what a 64 bit system offers, before you
> commit.

Or you could just give it a spin and see how you go: it's not as though 
64-bit Linux is any more expensive than 32-bit.  FWIW I've been running 
my FreeBSD systems 64 bit for five or six years (essentially since it was 
an option) with no problems, my MacOS systems for a couple of years 
(since that was possible) and I get to play with a couple of 64-bit Win7 
systems now and then too.  Only the latter have issues, of the occasional 
missing or broken device-driver nature.

Cheers,

-- 
Andrew

Richard Owlett wrote:

> {also posted to a Linux group}
> 
> In the "real world", is there a difference for an
> average/typical user.

It all depends on the software. I use bibble5 (raw converter) on 
64 bit Linux running on a i7-920 (4 physical and 8 logical cores).
It runs like hell because it uses all cores. This would probably work
the same on a 32 bit system since, I don't think that bibble5 needs
moe than 2 GB memory. 

On another 64 bit Linux system (i7-2600, 4 physical and 8 logical 
cores) I compile code with 'make -j 8' which leads to a speed up 
of factor 4 or more because 'make' runs 8 threads in parallel.
This works the same on a 32 bit OS.

> I'm contemplating the move from Windows to Linux.
> Coward that I am, I'm considering separate hardware as testbed.
> I spend most of my time in text editing and web surfing -
> doubt any advantage there.

OpenOffice and Firefox exist.

> My other interest include Scilab, Scioslab, and gnuplot. 

Don't know the 'Sci' stuff, but gnuplot works fine.

> I would run and display display results of *LARGE* fft's of
> _stored_ data. "Real time" minor importance.

You need a 64 bit OS if your application needs more than 2 GB,
or AFAIK if you have more than 3.5 GB main memory.

> Are there subtle advantages to wider data path and/or
> multiple cores?

Multicore is an advantage if one runs an application with heavy
processing in the background and needs a responsive system in 
the foreground.

> Are there advantages to 64 bit implementations of Linux?

You can use more than 3,5 GB main memory (same as with Windows).

bye
Andreas
-- 
Andreas H�nnebeck | email: acmh@gmx.de
----- privat ---- | www  : http://www.huennebeck-online.de
Fax/Anrufbeantworter: 0721/151-284301
GPG-Key: http://www.huennebeck-online.de/public_keys/andreas.asc
PGP-Key: http://www.huennebeck-online.de/public_keys/pgp_andreas.asc