[linux-audio-dev] Athlon64 .. performance boost ?

[Benno Senoner]

Taybin Rutkin wrote:

>>From my understanding, they are completely different chips with different instruction sets.  I seriously doubt the intel compiler will produce x86-64 instructions.  Not to mention the business perspective...
>
Yep Taybin is right, AMD64 and IA64 (Itanium) instruction sets are 
completely different.
Anyway Juan will soon retract his float vs int resamplers claim that int 
resampling is much faster:
it depends from the CPU you use. On older CPUs it is true, but Juan and 
I made some tests today while hanging out on #lad
and the Celeron P4 CPU took 4.5secs for integer resampling while it took 
4.5 for float resampling.
I think the integer advantage will soon be gone (and as you see on the 
P4 it becomes a disadvantage) with new generations of CPUs that are 
equally fast in handling ints or floats.
floats do have the advantage that you do not need to fiddle around with 
bitshifting, scaling, clipping etc.

PS: some infos about the unpleaseant surprises that await you when 
running a 64bit AMD CPU on Windows:

http://www.pcworld.com/news/article/0,aid,112749,pg,6,00.asp

In short you have 2 choices:
- stick with 32bit Windows and have your Ferrari CPU unable to switch 
gears past the 3rd out of 6 available.
Wait 2004 Q1 (I doubt billy boy will meet the deadline) for a crippled 
64bit version of Windows: basically you it will require most drivers to 
be ported to 64bit even trivial stuff like joystick drivers.
This means lots of compatibility issues which will convince most of the 
Windows people to stick to their 32bit Windows for long time.

PS2: I see a big problem with the "schism" of the 64bit instruction sets 
(between Intel Itanium and AMD64).

Basically it puts Microsoft in an uncomfortable position:
- increases the amount of development resources
- incompatibilities between Itanium and AMD 64:  if you bought MS Office 
for AMD64 it will not run on Itanium and viceversa.
Users will say "hey, it's always windows why is this incompatible ?"
AMD is risking too because users usually stick with the strongest one 
thus it could be that when Itanium arrives on the
desktop it will eclipse the AMD64 because of these Windows 
incompatibility problems.

Of course in the server domain the instruction set does not matter that 
much but in the desktop field it matters alot.

What will happen in your opinion ? Will Microsoft be tempted to stick to 
the Wintel paradigm or will they try to avoid it
in order to not risking loosing sales to the competition (aka Linux).
It's a complex equation I see no easy solution.

Perhaps Linux will be the salvation of AMD64 on the Desktop (if it can 
get a foothold in that domain) ?

Anyway we should be grateful to the kernel developers that made Linux a 
solid cross platform product that can run on
almost any CPU including Itanium and AMD 64.

cheers,
Benno


>
>Taybin
>
>-----Original Message-----
>From: Thomas Webb <magusofthedark at yahoo.com>
>Sent: Oct 16, 2003 11:51 AM
>To: 
>	The Linux Audio Developers' Mailing List <linux-audio-dev at music.columbia.edu>
>Subject: Re: [linux-audio-dev] Athlon64 .. performance boost ?
>
>I know the Intel C++ compiler isn't free software (you
>can get it free as in beer to compile open source
>apps, though), but it's avail for Itanium.. I don't
>know how well it works with AMD's stuff. It's my
>compiler of choice.
>
>  
>