AidenShaw said:
Not hype? Then why does OSX 10.4 "64-bit support" not extend to GUI applications? Why only terminal apps? Any why was Apple able to ship a 10.4 update that completely disabled 64-bit support without anyone noticing until after it had shipped?
That is little too much of a spin... it is much more then just hype. It exists and it can be used.
Apple hasn't yet provided 64 bit support (pointer/long) via the Carbon, Cocoa, Core Foundation, etc. frameworks because it is a non-trivial task (all of them are heavily intertwined so it cannot be done piecemeal, Carbon is likely the biggest problem area). As a result of this Apple chose to provide 64 bit support (pointer/long) only via libSystem which is a much smaller set of APIs and a generally flat set of APIs that can be used independently of the higher-level APIs. This allowed them to get 64 bit address support out much more quickly and in a form that developers can leverage for processes that really do require 64 bit addressing (a vast major of graphical application do NOT need 64 bit addressing, in fact IMHO I would say almost none really need it... they could be implemented as a 32b GUI with 64b process back-end if developers wanted to take the time).
Note 64b general math for use in an application is fully supported as of 10.2.8 and related tool chain. Also many of the system libraries and frameworks use 64b general math internally when on the G5 to reap benefit when processing 64b wide integers.
Anyway Apple had implied (before the Intel switch) that they would be expanding 64 bit addressing to the whole of the user space APIs while keeping the kernel 32 bit. Basically one kernel using 32b with 32b and 64b versions of user mode libraries/frameworks.
Now the Intel switch disrupts Apples original 64b support plans because, as you and others have noted, x86/64 operational mode has benefits outside of just 64b math and addressing that make putting the CPU into that mode a worthwhile thing for all of the operating system. Also x86/64 work differently then PowerPC when operating in a dual addressing mode environment.
I personally expect Apple to come forward with x86/64 plan mid to late this year that aligns (likely lags to some extent) the switch of PowerMacs to a x86/64 capable processor. The release vehicle for this will likely be 10.5.
Anyway as you note not many Mac OS X application use 64b support that currently exists (or at least also provide 32b versions) so if Apple takes a temporary step back from 64b during the Intel transition it won't affect things much.
I rather Apple take their time... think out how they want to attack this issue... get buy in from developers (ala WWDC)... instead of rushing out a way of getting 64b support for MacIntels. Of course the sooner us developer know what they are thinking the better (I bet WWDC 2005).
