Re: Audio recording bitdepth
Re: Audio recording bitdepth
- Subject: Re: Audio recording bitdepth
- From: Bjorn Roche <email@hidden>
- Date: Wed, 9 Dec 2009 18:22:57 -0500
On Dec 9, 2009, at 2:24 PM, Paul Davis wrote:
On Wed, Dec 9, 2009 at 2:07 PM, Bjorn Roche <email@hidden> wrote:
At the risk of starting a flame war, here is my follow-up:
http://blog.bjornroche.com/2009/12/linearity-and-dynamic-range-in-int.html
although i appreciate the focus of the two posts is on audio quality
(as it should be), it would be instructive to follow up with a measure
of the cache cost of using a conditional to determine negativity in
the asymmetric method. cachegrind should do this relatively easily. we
ran some small not very representative tests when revisiting the JACK
ALSA backend conversion code and concluded that cost in branch-induced
cache misses was larger than we were willing to pay in exchange for
the extra sample value. although there are many cases where the user
may wish to put fidelity over cpu cost, at least being able to
estimate the likely cost (say, per channel of N bit samples at a given
SR) would aid people in making this choice wisely.
Thanks for your comments. Perhaps I will look into it. I do this a lot
and it's never come up for me in profiling, so it's not a big enough
deal for me to care for my own purposes.
I am a bit confused by your comments, though: I think in this case you
aren't going to deal with cache misses. The likely performance issue
is branch prediction failures. If that turns into a cache problem
somehow, it's beyond my knowledge. To my understanding, in the worst
case, an incorrect branch predication should flush the processor
pipeline. I don't really see that happening though, since each loop
contains independent data, so on a branch prediction failure it should
be able to continue with the next data even when the last data failed.
Of course, should and does are different, but, again, that's where the
profiler comes in.
bjorn
-----------------------------
Bjorn Roche
XO Wave
Digital Audio Production and Post-Production Software
http://www.xowave.com
http://blog.bjornroche.com
http://myspace.com/xowave
_______________________________________________
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