[nikki-k]

Hi!
With the release of the G5, and AMD entering the fray, 64-bit computing seems to be "all the rage" for those that cannot seem to be happy where they are, instead searching every near-by island, somehow thinking it "mnight be better than what I have now." Others just like the thrill of new stuff.

Some threads on the TDM Mac forum targeted the question of the possibility of a 64-bit edition of Pro Tools , and Mac stuff related tot he G5.
(for those unfamaliar with how 32-bit works, do some reading on OS's and memory use, addressing, etc. Diminishing ignorance is bliss).

From the Mac TDM forum: Kenn LeGault from Digidesign had this to say on 64-bit:

Hello,

It's worth asking: what is a "64-bit app" anyway? Is there such a thing on a G5/OSX system?

There are two things to consider when talking about "64-bitness" ...

First, there's the address bus and address registers. The G5 supports 64-bit addresses so, in theory, a "true" 64-bit application could address 2^64 bytes of memory (this is a very large number), rather than 2^32 bytes of memory (4 GB). However, the problem here is that OSX is not a "true 64-bit OS" since it doesn't let any single process (application) access more than 2^32 bytes of memory (4 GB). It's true that the G5 is "a bit better" than 32-bit systems in terms of addressing (pun intended):

On a G4, an individual application can access "only" 2^31 bytes of memory (2 GB). The G5 lets a single application address twice as much memory as the G4, or 2^32 (4 GB). It's true that the G5 allows you to put twice as much RAM in your computer as compared to a PC -- 8 GB rather than 4 GB -- but 8 GB is "only" 2^33 (again, a "bit" better than a PC).

For database applications and some specialized scientific applications, having a 64-bit logical address space is important (even if you don't have more than 8 GB of physical RAM available), since it simplifies how the code is written to access bazillions of records in a database (using a so-called "64-bit flat address space"). However, Pro Tools (and most other applications) don't really care about 64-bit addressing. It doesn't buy us anything. Two or four gigabytes of memory address space is more than enough for >95% of the applications out there.

The second important part of "64-bitness" is that there are 64-bit data registers and a 64-bit FPU (floating point unit) in the G5. But guess what? The G4 already has this feature! We've had 64-bit ("double-precision") math code in our plug-ins for quite some time (and many third party plug-ins use 64-bit internal math in their algorithms, too). So, our 64-bit "double-precision" code will continue to run on a G5 just as before, except for the fact that the G5 runs much faster, and has two FPUs. So, while the G5 chip is certainly more "powerful" than the G4 in terms of floating point operations (it can do more per second), we don't need to change our C++ code, and we don't even need to recompile our existing code. Our plug-ins will take advantage of the new G5 performance automagically.

So, don't get too hung up on the notion of "64-bit computing." If you're running on a "true" 64-bit OS (like Microsoft and perhaps Apple may ship someday), and if you're running a 64-bit database or scientific application, you'll see a difference. Otherwise, 64-bit addressing doesn't really help most applications.

Regards,
Kenn


--------------------
Kenn LeGault
Director of Software Engineering
Digidesign