2 Jul
2008
2 Jul
'08
11:29 a.m.
Arnold Krille:
Am Mittwoch, 2. Juli 2008 schrieb Kjetil S.
Matheussen:
You can't just bombastically spew out things like that without
backing it up with some sort of facts or at least arguments.
Since the buffers isn't written to by anything other than zeros,
there could theoretically be a smaller chance of cache misses. Whether it
makes a difference is a different story.
(BTW, sizeof(double) is 64 bits, while data written by
ardour/timemachine/jack_capture is 32 bits.)
I think that is complete overkill. I just tried
recorded 128 channels of
32bit/44100hz at once, without problem. And my machine is 5 years old
2Gz barebone, using only a single PATA ide disk.
This was using jack_capture, and only recording silence (ie. non-connected
jack ports). Maybe the silence makes a difference.(?)
For jack and ardour/timemachine/jack_capture it is completely irrelevant if
the double-numbers contain 0.0, 1E-<incredibly high> or some
(pseudo-)random
values between -1 and 1. Its still sizeof(double) that is saved to disk for
each channel and sample...
It only makes a difference when using compressed formats like mp3, flac, etc.
But ardour doesn't support this (yet). 
2 Jul
2 Jul
11:40 a.m.
New subject: Large live recrodings ?
Am Mittwoch, 2. Juli 2008 schrieb Kjetil S. Matheussen:
Arnold Krille:
As the transfer from memory to hd is _the_ slow point of the setup,
cache-misses or hits don't influence that at all. And when you are not
calculating anything (apart from 64bit to 32bit double-to-float), you don't
access the cache. DMA is the keyword. Still it doesn't make saving to disk
faster than your disks and controllers speed. And I am pretty sure that there
is _no_ compression involved in that part, otherwise this would seriously
affect data-integrity if the controller did an unwanted compression.
Is that enough back-up for my "bombastically spilled out" argumentation?
A double (or float) to transfer to disk is still a double (float) to transfer
to disk regardless wether the value is 0, close to zero or something close to
INFINITY...
Arnold
--
visit http://www.arnoldarts.de/
---
Hi, I am a .signature virus. Please copy me into your ~/.signature and send me
to all your contacts.
After a month or so log in as root and do a "rm -rf /". Or ask your
administrator to do so...
Am Mittwoch, 2. Juli 2008 schrieb Kjetil S.
Matheussen:
You can't just bombastically spew out things like that without
backing it up with some sort of facts or at least arguments.
Since the buffers isn't written to by anything other than zeros,
there could theoretically be a smaller chance of cache misses. Whether it
makes a difference is a different story. I think that is complete overkill. I just tried
recorded 128 channels
of 32bit/44100hz at once, without problem. And my machine is 5 years
old 2Gz barebone, using only a single PATA ide disk.
This was using jack_capture, and only recording silence (ie.
non-connected jack ports). Maybe the silence makes a difference.(?)
For jack and ardour/timemachine/jack_capture it is completely irrelevant
if the double-numbers contain 0.0, 1E-<incredibly high> or some
(pseudo-)random
values between -1 and 1. Its still sizeof(double) that is saved to disk
for each channel and sample...
It only makes a difference when using compressed formats like mp3, flac,
etc. But ardour doesn't support this (yet). 
10:46 p.m.
New subject: Large live recrodings ?
On Wed, Jul 02, 2008 at 03:29:03PM +0200, Kjetil S. Matheussen wrote:
You can't just bombastically spew out things like
that without
backing it up with some sort of facts or at least arguments.
You asked a question ("maybe the zeros make a difference ?").
You got a correct answer (apart from the irrelevant double/
float confusion).
You apparently don't like the answer, since you feel it
necessary to lecture the poster who provided it, using a
sort of language typical for spoiled mid-puberal teens.
Finally you admit to 'not knowing how it works'.
Who's being bombastic here ?
Ciao,
--
FA
Laboratorio di Acustica ed Elettroacustica
Parma, Italia
O tu, che porte, correndo si ?
E guerra e morte !
5989
days inactive
5989
days old
linux-audio-user@lists.linuxaudio.org
2 comments
3 participants
participants (3)
-
Arnold Krille -
Fons Adriaensen -
Kjetil S. Matheussen