• Open Menu Close Menu
  • Apple
  • Shopping Bag
  • Apple
  • Mac
  • iPad
  • iPhone
  • Watch
  • TV
  • Music
  • Support
  • Search apple.com
  • Shopping Bag

Lists

Open Menu Close Menu
  • Terms and Conditions
  • Lists hosted on this site
  • Email the Postmaster
  • Tips for posting to public mailing lists
Re: 64bit processing - Why?
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: 64bit processing - Why?


  • Subject: Re: 64bit processing - Why?
  • From: Ian Kemmish <email@hidden>
  • Date: Fri, 31 Jul 2009 07:43:59 +0100


Rats, I'm breaking my own convictions never to post to threads like this.



On 30 Jul 2009, at 22:34, Justin Carlson <email@hidden>,
talking about promoting intermediate floating results, wrote:



While this is common on many architectures, it is not (correct me if I am wrong along the way) standard in AU plugin developers' most common deployment architecture: Intel, 32 bit, OS X. For this runtime, SSE instructions+registers are *default*.

Actually, the 8087 and its descendants were about the only architecture to have done this. Everyone else followed the VAX route of not promoting floats, and eventually what the VAX did became (with a few tweaks) the IEE standard.


The reason most modern Intel OSs use the scalar variants of SSE for scalar floating point in preference to the FPU is that the former are IEEE compliant, whereas the latter are not. Why do people prefer to use the floating point unit which offers less intermediate precision? Because IEEE compliant units allow to to better model and predict the propagation of errors. For example, with an 8087 FPU, an optimising compiler would frequently produce different answers to one which didn't optimise, simply because it homed variables less often... (and yes, I have seen this materially affect the reproducibility of customer bug reports). Not good!


As far as I can see, this hu-u-u-uge argument boils down to two things:

1) If you want to end up with 24 bits, then having a few guard bits to pass between AUs would be nice.

I think we can all agree with that, to some extent.

2) I want more than the 8 guard bits which 32 bit integer samples would give me.

I don't think we can all agree with that. How many AUs in a chain would you need for 8 guard bits to become a necessity?


- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Ian Kemmish 18 Durham Close, Biggleswade, Beds SG18 8HZ
email@hidden Tel: +44 1767 601361 Mob: +44 7952 854387
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



_______________________________________________ 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
  • Prev by Date: Re: 64bit processing - Why?
  • Next by Date: just how stable is kMIDIPropertyUniqueID ?
  • Previous by thread: Re: 64bit processing - Why?
  • Next by thread: just how stable is kMIDIPropertyUniqueID ?
  • Index(es):
    • Date
    • Thread