Tuesday, June 07, 2005

Apple on Intel

From what I saw on the streaming keynote address, it appears that the Jobs and the presenters where very careful not mention what sort of Intel chip is going to be used. Apple would be very stupid not to be going with 64bit chip from the get go. Their Universal Binary Programming Guidelines currently only speaks about IA-32.

Looking at the guidelines, it appears that even if an application is properly written and is just a recompile away from becoming a universal binary, there are some very subtle issues which might not be readily apparent until some heavy testing is done. For example, on the PowerPC, when a float is cast into an int, but the value is too large to be represented, the int gets assigned INTMAX, while on an x86, it gets assigned INTMIN. Also, an integer divide by zero is fatal on the x86, while on the PowerPC, zero is returned.

Hopefully, Rosetta (the dynamic PowerPC to Intel translator) will work well since I'm sure it's going to be a long time before a lot of applications are rebuilt as universal binaries. I almost think that come 2007, it might still be smarter to buy a G5 rather than an Apple Intel box due to software availability.

I'll probably buy a G5 within the next year or two, but I'm definitely not going to make any big software purchases (especially multimedia type apps) in the near future unless absolutely necessary since I'm not going to want to upgrade again once the software gets recompiled as universal binaries.

No comments:

Post a Comment