[Fwd: Re: XCode 2.4 and 32/64-bit universal binaries]
[Fwd: Re: XCode 2.4 and 32/64-bit universal binaries]
- Subject: [Fwd: Re: XCode 2.4 and 32/64-bit universal binaries]
- From: "Angus F. Hewlett" <email@hidden>
- Date: Sun, 20 Aug 2006 18:25:59 +0100
Herbie Robinson wrote:
Okay, so, what about a 32-bit app on a 64 bit CPU vs. the same app
compiled for that target 64-bit CPU? Will the latter work faster or no?
The experience with RISC machines (that I am aware of) is that 64 bit
applications often run a little slower because the cache footprint
becomes larger and there are more cache misses. If your code is
written in a highly transportable fashion (i.e., all of your integers
are declared using the types in stdint.h or the Apple equivalents),
the hit should be minimal. With the Intel architecture, the 64 bit
mode has twice as many hardware registers and is reported to run faster.
Yep, also the FPU instruction set (key for audio) has been completely
redone for 64-bit x86 and is much cleaner -- goodbye (finally) to the
horror that was the x87 register-stack FPU instruction set.
I don't think anybody is talking about more than 10% differences,
BTW. The main reason you want to think about doing a 64 bit
application is if you need the address space (which seems unlikely for
audio applications).
You'd be surprised.. a quad-processor system has enough power to run a
lot of large plug-ins, you can easily fill up 3-4GB of RAM with large
multisample sets. Whether such large instruments should even be running
in the same process space as the host app to begin with is a matter of
debate, though.
Best regards,
Angus.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Coreaudio-api mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden