Re: high-performance iphone and 8.24: convert to float and back or fixed-point math?
Re: high-performance iphone and 8.24: convert to float and back or fixed-point math?
- Subject: Re: high-performance iphone and 8.24: convert to float and back or fixed-point math?
- From: Jens Alfke <email@hidden>
- Date: Fri, 13 Feb 2009 14:27:40 -0800
On Feb 13, 2009, at 10:17 AM, Joel Reymont wrote:
What is the optimal (high-performance way) of working with 8.24 data
on the iPhone?
Should I disable Thumb for the app, do the signal processing using
floats and convert to and from 8.24 as needed?
Someone mentioned in the archives that doing fixed-point math
directly will quickly out-weight the hit of converting an array of
samples to float.
Last but not least, are there optimized routines for converting 8.24
to float and back?
No offense, but why do you need to do this, if you're just streaming
audio over the network? You shouldn't need to do math on raw samples
to do this; just tell CoreAudio what format you want it to fill
buffers with, and send and receive packets out of the buffers.
Also, isn't 8.24 massive overkill for voice? That's what pro audio
mastering and SA-CDs use. You can get by quite well with 8-bit samples
and cut your data rate by a factor of 4.
—Jens _______________________________________________
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