|
I haven't jumped into this (or anything else for a while) but Paul's
response compels me to chime in. On Unix platforms at least, Csound is a
wonderful low-latency high performance real-time DSP environment that
easily rivals MAX/MSP on the Mac in terms of ressources. I have not done
tests, but i would wager that Csound on a 150MHz SGI Indy has at least
equal crunching power and audio throughput as MSP on a 300Mhz G3. Its
not the machine, its the nature of the code and the operating system.
Csound should definitely NOT be tuned for file generation to the
exclusion of real-time performance. A clean 16-bit signal is fine in
almost all performance situations. Making sample throughput slower goes
against the grain of every developement effort in pro-audio. Its a sure
way to make Csound irrelevant.
If double-precision CAN be acheived, then, by all means
Paul wrote:
>
> Mike Berry wrote:
> >
> > In the float/double question, everybody is making it a real-time vs.
> > file-time comparison, which I don't think is totally fair. CSound,
> > however you use it now, was designed as a file-time program.
>
> I'm afraid I have to turn this on its head and say that csound, however
> it was designed, is now used both real-time and file-time. Why should
> the intended limited scope of a tool outweigh the actual uses people
> find for it? I don't think it should. So I disagree with your statement
> that:
>
> > .... Any improvements should be judged first on whether they improve
> > file generation (e.g. sound quality) and then on how they impact compile
> > times.
>
> I think we should rather, whenever possible, take both into account. If
> there is a way to provide options so the user can push csound further in
> either direction, then that is what I'm in favor of. A compile-time flag
> would do fine for me (I could just have two binaries: a speedier one and
> a "better" one).
>
> This is assuming that there is significant speed loss in the actual
> performance of the proposed double-precision version, which several
> folks have pointed out might not be the case (or at least might not be
> that bad). I'm now quite curious to see what actually happens in
> practice. It would be pretty funny if all this talk were shown to be
> completely irrelevant.
>
> --PW
--
________________________________________________________
Jean Piche
Universite de Montreal
http://mistral.ere.umontreal.ca/~pichej
http://www.musique.umontreal.ca/electro/CEC/
Received: from wallace.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa18323;
10 Jun 99 17:11 BST
Received: from [144.173.6.14] (helo=exeter.ac.uk)
by wallace.maths.bath.ac.uk with esmtp (Exim 2.12 #1)
id 10s7Qm-0006Jc-00
for jpff@maths.bath.ac.uk; Thu, 10 Jun 1999 17:11:52 +0100
Received: from noether [144.173.8.10] by hermes via SMTP (RAA08736); Thu, 10 Jun 1999 17:08:30 +0100 (BST)
Received: from exeter.ac.uk by maths.ex.ac.uk; Thu, 10 Jun 1999 17:08:18 +0100
Received: from mercury.acs.unt.edu [129.120.220.1] by hermes via ESMTP (RAA09709); Thu, 10 Jun 1999 17:08:10 +0100 (BST)
Received: from venus.acs.unt.edu (venus.acs.unt.edu [129.120.220.72])
by Mercury.unix.acs.cc.unt.edu (8.8.8/8.8.8) with ESMTP id LAA12676;
Thu, 10 Jun 1999 11:08:00 -0500 (CDT)
Received: from jove.acs.unt.edu (1Cust63.tnt5.grapevine.tx.da.uu.net [208.253.166.63])
by venus.acs.unt.edu (8.8.8/8.8.8) with ESMTP id LAA03890;
Thu, 10 Jun 1999 11:07:51 -0500 (CDT)
Message-ID: <375FE334.DDDD9099@jove.acs.unt.edu>
Date: Thu, 10 Jun 1999 11:09:26 -0500
From: Michael Thompson
X-Mailer: Mozilla 4.5 (Macintosh; I; PPC)
X-Accept-Language: en
MIME-Version: 1.0
To: Jean Piche
CC: The CSound mailinglist
Subject: Re: Precision vs Implementation
References: <375C46D1.FE17F605@nmol.com> <375F5A96.FED2954F@ulster.net> <375FAA7E.ACA1CD6B@umontreal.ca>
Content-Type: text/plain; charset=us-ascii; x-mac-type="54455854"; x-mac-creator="4D4F5353"
Content-Transfer-Encoding: 7bit
Sender: owner-csound-outgoing@maths.ex.ac.uk
Precedence: bulk
Jean Piche wrote:
> I haven't jumped into this (or anything else for a while) but Paul's
> response compels me to chime in. On Unix platforms at least, Csound is a
> wonderful low-latency high performance real-time DSP environment that
> easily rivals MAX/MSP on the Mac in terms of ressources. I have not done
> tests, but i would wager that Csound on a 150MHz SGI Indy has at least
> equal crunching power and audio throughput as MSP on a 300Mhz G3. Its
> not the machine, its the nature of the code and the operating system.
Well, as much as I love my SGI when its running csound it is no match for Max/MSP
in real-time performance on a G3 (you need ram for max to run well, but also on my
O2 the more ram the better the performance -- not only csound but system
performance). I have done some simple tests and you can do a ton of simple
processes(fm, filters and soundfile i/o) with low latency and it doesnt have to do
orc/sco preprocessing. Yes, csound is more flexible in some aspects but not in the
real-time world. A G3 can easily do fft/ifft phase vocoder implementations or fft
based 512 band eq processes (even on a 120mhz 8500, well-- not easy on this
machine). I can not do this using csound on my O2 (even with the code compiled n32
with all the right speed flags) its not that the O2 cant handle it, its more to do
with csound itself. Under jMax, my 180 mhz R5000 O2 really runs well, maybe even
better than a Max/MSP on a G3. Csound code is far from cross-platform optimization
not to mention target-platform, except maybe the Direct X version.
just my $0.02 US...
Michael
Received: from wallace.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa19377;
11 Jun 99 1:55 BST
Received: from [144.173.6.14] (helo=exeter.ac.uk)
by wallace.maths.bath.ac.uk with esmtp (Exim 2.12 #1)
id 10sFbc-0006YS-00
for jpff@maths.bath.ac.uk; Fri, 11 Jun 1999 01:55:36 +0100
Received: from noether [144.173.8.10] by hermes via SMTP (BAA08056); Fri, 11 Jun 1999 01:53:22 +0100 (BST)
Received: from exeter.ac.uk by maths.ex.ac.uk; Fri, 11 Jun 1999 01:53:09 +0100
Received: from adsl-209-78-185-158.dsl.lsan03.pacbell.net [209.78.185.158] by hermes via ESMTP (BAA10551); Fri, 11 Jun 1999 01:53:07 +0100 (BST)
Received: from screech.weirdnoise.com (localhost [127.0.0.1])
by screech.weirdnoise.com (8.8.7/8.8.7) with ESMTP id RAA05780
for ; Thu, 10 Jun 1999 17:53:07 -0700
Message-Id: <199906110053.RAA05780@screech.weirdnoise.com>
X-Mailer: exmh version 2.0.2
To: The CSound mailinglist
Subject: double/float times
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Thu, 10 Jun 1999 17:53:07 -0700
From: Ed Hall
Sender: owner-csound-outgoing@maths.ex.ac.uk
Precedence: bulk
This was done in a great deal of haste, but given some of the claims
made here I just had to make a small test. In essence, I start with the
wrapper script I used to allow Csound to compile on 64-bit machines
and modified it to convert float to double throughout Csound. It took
all of fifteen minutes to produce two binaries for Csound 3.54, one
with floats, the other with doubles. For all I know, the latter is
broken 100 different ways, but at least it ran the "Xanadu" orc and sco
that have been used for benchmarking. (See:
http://members.tripod.com/~slinkP/pw_linux/csbench.html
for the benchmark.)
Here are the results:
float: 95.72 seconds
double: 107.47 seconds
So the double version is about 12% slower. Details:
Dell D333XP 333MHz: Pentium II w/128MB RAM runing RedHat 5.2 Linux.
Csound 3.54 Compiled with EGCS 1.1.2, options "-O3". 16-bit IRCAM output.
I've listened to the outputs of both versions a couple of times, and
there are very slight differences which appear to have to do with
oscillator phasing (the beats between voices toward the end of the
piece sound just slightly different). This may be an actual difference
due to 32- vs. 64-bit floating point, or may be some breakage due to
the rather blunt way I did the conversion. Aesthetically, they were
equivalent--there certainly wasn't anything that would make me prefer
one over the other.
This is really more time than I ever intended to spend on this issue.
Someday I'll benchmark more scores and more platforms, but you'll have
to wait...
-Ed
Received: from shaun.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa19420;
11 Jun 99 2:14 BST
Received: from [144.173.6.14] (helo=exeter.ac.uk)
by shaun.maths.bath.ac.uk with esmtp (Exim 2.12 #1)
id 10sFty-0004oS-00
for jpff@maths.bath.ac.uk; Fri, 11 Jun 1999 02:14:34 +0100
Received: from noether [144.173.8.10] by hermes via SMTP (CAA02864); Fri, 11 Jun 1999 02:14:27 +0100 (BST)
Received: from exeter.ac.uk by maths.ex.ac.uk; Fri, 11 Jun 1999 02:14:14 +0100
Received: from renoir.op.net [209.152.193.4] by hermes via ESMTP (CAA02208); Fri, 11 Jun 1999 02:14:13 +0100 (BST)
Received: from someip.ppp.op.net (d-bm3-0c.ppp.op.net [209.152.194.76]) by renoir.op.net (o1/$Revision: 1.18 $) with ESMTP id VAA02903; Thu, 10 Jun 1999 21:12:12 -0400 (EDT)
Message-Id: <199906110112.VAA02903@renoir.op.net>
To: Jean Piche
Cc: csound@maths.ex.ac.uk
Subject: Re: Precision vs Implementation
In-reply-to: Your message of "Thu, 10 Jun 1999 08:07:26 EDT."
<375FAA7E.ACA1CD6B@umontreal.ca>
Date: Thu, 10 Jun 1999 21:21:12 -0400
From: Paul Barton-Davis
Sender: owner-csound-outgoing@maths.ex.ac.uk
Precedence: bulk
>easily rivals MAX/MSP on the Mac in terms of ressources. I have not done
>tests, but i would wager that Csound on a 150MHz SGI Indy has at least
>equal crunching power and audio throughput as MSP on a 300Mhz G3. Its
>not the machine, its the nature of the code and the operating system.
actually, it *is* the machine, as has been so clearly demonstrated in
the discussions about the "oscillates" benchmark. The MIPS CPU does
way better than a Pentium of the same clock rating (and I think we can
assume than the G3), even with the *same* code and a vaguely similar
(IRIX/Linux) operating system.
--p
Received: from wallace.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa19459;
11 Jun 99 2:36 BST
Received: from [144.173.6.14] (helo=exeter.ac.uk)
by wallace.maths.bath.ac.uk with esmtp (Exim 2.12 #1)
id 10sGFZ-0006Z9-00
for jpff@maths.bath.ac.uk; Fri, 11 Jun 1999 02:36:53 +0100
Received: from noether [144.173.8.10] by hermes via SMTP (CAA15003); Fri, 11 Jun 1999 02:36:43 +0100 (BST)
Received: from exeter.ac.uk by maths.ex.ac.uk; Fri, 11 Jun 1999 02:36:32 +0100
Received: from root@deliverator.io.com [199.170.88.17] by hermes via ESMTP (CAA18060); Fri, 11 Jun 1999 02:36:31 +0100 (BST)
Received: from fnord.io.com (IDENT:jti@fnord.io.com [199.170.88.12])
by deliverator.io.com (8.9.3/8.9.3) with ESMTP id UAA25098
for ; Thu, 10 Jun 1999 20:34:32 -0500
Received: from localhost (jti@localhost)
by fnord.io.com (8.8.7/8.8.7) with ESMTP id UAA09757
for ; Thu, 10 Jun 1999 20:34:32 -0500
X-Authentication-Warning: fnord.io.com: jti owned process doing -bs
Date: Thu, 10 Jun 1999 20:34:32 -0500 (CDT)
From: "Jeremiah T. Isaacs"
To: csound@maths.ex.ac.uk
Subject: getting started, weird problem...
Message-ID:
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-csound-outgoing@maths.ex.ac.uk
Precedence: bulk
with the following orc and sco, i only hear instrument 1. if i comment
out instrument 1 in the .sco, i hear only instrument 2.
please tell me im making some strange mistake.
; arg.sco
f1 0 1024 10 10 9 8 7 6 5 4 3 2 1
f2 0 1024 10 10 9 8 7 6 5 4 3 2 1
i1 0 4 10000 6.00
i2 4 5 10000 8.00
e
; arg.orc
sr=44100
kr=44100
ksmps=1
nchnls=2
instr 1
kenv1 linseg p4/3,p3/4,p4,p3/4,p4/2,p3/4,p4/2,p3/4,0
asig1 oscil kenv1, cpspch(p5), 1
asig2 oscil kenv1, cpspch(p5/2), 1
asig3 oscil kenv1*0.2, cpspch(p5*2), 1
outs asig1*asig3, asig2*asig3
endin
instr 2
kenv1 linseg 0,p3/4,p4,p3/4,p4/2,p3/4,p4/2,p3/4,0
kenv2 linseg 0,p3/3,p4,p3/3,p4,p3/3,0
asig1 oscil kenv1, cpspch(p5), 2
asig2 oscil kenv2, cpspch(p5/2), 2
asig3 oscil kenv1, cpspch(p5*2), 2
asig4 oscil kenv2, cpspch(p5*3), 2
outs asig2+asig4, asig1+asig3
endin
;;;;;
;and the output from csound, which indicates no errors...
;;;;;
Csound Version 3.51 (Jan 26 1999)
WARNING: floats encoding information cannot
be contained in the header...
orchname: moog:Desktop Folder:csoundinsh:mine:arg.orc
scorename: moog:Desktop Folder:csoundinsh:mine:arg.sco
sorting score ...
... done
orch compiler:
35 lines read
instr 1
instr 2
MIT Csound: 3.51 (Jan 26 1999)
(Mills/PPC: 3.5.1)
orch now loaded
audio buffered in 8192 sample-frame blocks
SFDIR undefined. using current directory
writing 65536-byte blks of floats to arg.aif (AIFF)
SECTION 1:
ftable 1:
ftable 2:
new alloc for instr 1:
B 0.000 .. 4.000 T 4.000 TT 4.000 M:19043136.019462674.0
new alloc for instr 2:
B 4.000 .. 9.000 T 9.000 TT 9.000 M: 10000.5 19708.7
end of score. overall amps:19043136.019462674.0
0 errors in performance
49 65536-byte soundblks of floats written to arg.aif (AIFF)
**Total Rendering time was: 1.66666 secs
;;;;;;;;
just getting started really. whats my damage...?
------------------------------
jti@io.com
http://www.io.com/~jti/
http://www.sonictherapy.com/fluoroscopic/fk.html
------------------------------------------------
--
Received: from wallace.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa19699;
11 Jun 99 4:04 BST
Received: from [144.173.6.14] (helo=exeter.ac.uk)
by wallace.maths.bath.ac.uk with esmtp (Exim 2.12 #1)
id 10sHce-0006bJ-00
for jpff@maths.bath.ac.uk; Fri, 11 Jun 1999 04:04:48 +0100
Received: from noether [144.173.8.10] by hermes via SMTP (EAA12083); Fri, 11 Jun 1999 04:03:37 +0100 (BST)
Received: from exeter.ac.uk by maths.ex.ac.uk; Fri, 11 Jun 1999 04:03:24 +0100
Received: from mtiwmhc06.worldnet.att.net [204.127.131.41] by hermes via ESMTP (EAA14295); Fri, 11 Jun 1999 04:03:23 +0100 (BST)
Received: from worldnet.att.net ([12.76.98.101])
by mtiwmhc06.worldnet.att.net (InterMail v03.02.07 118 124)
with ESMTP id <19990611030022.JARZ19793@worldnet.att.net>;
Fri, 11 Jun 1999 03:00:22 +0000
Message-ID: <37607BF3.97094A81@worldnet.att.net>
Date: Thu, 10 Jun 1999 23:01:09 -0400
From: Kirsh Family
Reply-To: kirsh@worldnet.att.net
Organization: Kirsh Family
X-Mailer: Mozilla 4.5 (Macintosh; U; PPC)
X-Accept-Language: en-US
MIME-Version: 1.0
To: "Jeremiah T. Isaacs"
CC: csound@maths.ex.ac.uk
Subject: Re: getting started, weird problem...
References:
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: owner-csound-outgoing@maths.ex.ac.uk
Precedence: bulk
I admit I didn't take time to run the given orc/sco myself, but from
looking at the Csound output, I think I can guess what's wrong. There
are only two notes in the score, and look at the amplitude each one
reaches, as given in the output:
new alloc for instr 1:
B 0.000 .. 4.000 T 4.000 TT 4.000 M:19043136.019462674.0
new alloc for instr 2:
B 4.000 .. 9.000 T 9.000 TT 9.000 M: 10000.5 19708.7
end of score. overall amps:19043136.019462674.0
The amplitude of the first note is ~1900x the second one. My guess is
that after the output floats are rescaled to 16-bit ints (I'm assuming
you're using that handy feature of the Mac version), the second note is
being pushed down into the noise (~65dB below the 1st note). When you
comment out the first note, then the rescaling brings the 2nd note up to
16-bit full scale, so you can easily hear it. So, if you want to hear
them both, you need to do something to reduce the difference in the amplitudes.
David Kirsh, kirsh@att.net
Received: from shaun.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa19715;
11 Jun 99 4:15 BST
Received: from [144.173.6.14] (helo=exeter.ac.uk)
by shaun.maths.bath.ac.uk with esmtp (Exim 2.12 #1)
id 10sHmV-0004qx-00
for jpff@maths.bath.ac.uk; Fri, 11 Jun 1999 04:14:59 +0100
Received: from noether [144.173.8.10] by hermes via SMTP (EAA04868); Fri, 11 Jun 1999 04:14:50 +0100 (BST)
Received: from exeter.ac.uk by maths.ex.ac.uk; Fri, 11 Jun 1999 04:14:37 +0100
Received: from out5.ibm.net [165.87.194.243] by hermes via ESMTP (EAA00741); Fri, 11 Jun 1999 04:14:36 +0100 (BST)
Received: from ibm.net (slip166-72-73-90.dc.us.ibm.net [166.72.73.90]) by out5.ibm.net (8.8.5/8.6.9) with ESMTP id DAA59178; Fri, 11 Jun 1999 03:12:33 GMT
Message-ID: <37607FBD.1F8788B@ibm.net>
Date: Thu, 10 Jun 1999 23:17:17 -0400
From: "Job M. van Zuijlen"
Reply-To: zuijlen@ibm.net
Organization: electona
X-Mailer: Mozilla 4.5 [en] (Win98; U)
X-Accept-Language: en
MIME-Version: 1.0
To: Paul
CC: Csound
Subject: Re: Higher numerical precission in Csound
References: <001001beb06a$3da380a0$79d496c0@Realizer.ngt.sungard.com> <375AFED1.8523842A@ibm.net> <375F5545.102AD54B@ulster.net>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: owner-csound-outgoing@maths.ex.ac.uk
Precedence: bulk
One way to look at this is to view one bit as serial and 16+ bit as
parallel. There are a few USB (Universal Serial Bus) audio devices on
the market now and I also read about Ethernet used as multichannel audio
cable. So you may see more and more of those devices become available.
There were already CD players with 1-bit converters several years ago.
Job van Zuijlen
Paul wrote:
>
>
> And then there's the _real_ lone nuts in the wilderness, who claim that
> PCM was a mistake from the get-go, and we should be converting
> everything to some sort of 1-bit, extremely-high-sampling-rate
> technology that Sony is now using for archival masters; they claim that
> this blows the pants off of every other format currently in use (in part
> by eliminating the anti-aliasing filters). Who knows. I don't even
> understand how this system works. Too bad it's completely incompatible
> with all existing software and hardware... I wonder if those guys are
> right, and I wonder if we'll ever get to find out; or if we're now stuck
> with PCM regardless.
>
Received: from shaun.maths.bath.ac.uk by omphalos.maths.Bath.AC.UK id aa20482;
11 Jun 99 10:46 BST
Received: from [144.173.6.14] (helo=exeter.ac.uk)
by shaun.maths.bath.ac.uk with esmtp (Exim 2.12 #1)
id 10sNsy-00052q-00
for jpff@maths.bath.ac.uk; Fri, 11 Jun 1999 10:46:04 +0100
Received: from noether [144.173.8.10] by hermes via SMTP (KAA13322); Fri, 11 Jun 1999 10:43:35 +0100 (BST)
Received: from exeter.ac.uk by maths.ex.ac.uk; Fri, 11 Jun 1999 10:43:26 +0100
Received: from root@lix.intercom.es [194.179.21.2] by hermes via ESMTP (KAA18092); Fri, 11 Jun 1999 10:43:22 +0100 (BST)
Received: from intercom.es (iv3-253.intercom.es [195.76.131.253]) by lix.intercom.es (8.7.3/8.6.12) with ESMTP id LAA06460; Fri, 11 Jun 1999 11:46:27 +0100
Message-ID: <3760D983.97D502F8@intercom.es>
Date: Fri, 11 Jun 1999 11:40:19 +0200
From: Josep M Comajuncosas
X-Mailer: Mozilla 4.05 [en] (Win95; I)
MIME-Version: 1.0
To: "Jeremiah T. Isaacs"
CC: csound@maths.ex.ac.uk
Subject: Re: getting started, weird problem...
References:
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: quoted-printable
X-MIME-Autoconverted: from 8bit to quoted-printable by exeter.ac.uk id KAA18092
Sender: owner-csound-outgoing@maths.ex.ac.uk
Precedence: bulk
Well in instr 1 you=B4re multiplying two enveloped oscillators of maximum
ampliture =3D 10000 each, so you=B4ll get a peak amplitude of 1000000, fa=
r
beyond the limit of about 32000 for a 16 bit output. You=B4ll only hear a
click at the beginning and at the end of the note because all the signal =
is
clipped.
In instr 2 you perform an addition, so the output amplitude can go up to
20000 which is ok for 16bit files.
If you want to hear how ring modulation sounds (instr 1) I suggest you to
use normalized amplitudes (p4 =3D 1 in the score) and multiply the result=
by
10000 or whatever you want.
outs 10000*asig1*asig3, 10000*(asig2*asig3)
With normalized amplitudes you won=B4t have to be bothered by overflow.
Hope this helps,
Josep M Comajuncosas
Jeremiah T. Isaacs wrote:
> with the following orc and sco, i only hear instrument 1. if i comment
> out instrument 1 in the .sco, i hear only instrument 2.
>
> please tell me im making some strange mistake.
>
> i1 0 4 10000 6.00
> i2 4 5 10000 8.00
> instr 1
> asig1 oscil kenv1, cpspch(p5), 1
> asig2 oscil kenv1, cpspch(p5/2), 1
> asig3 oscil kenv1*0.2, cpspch(p5*2), 1
>
> outs asig1*asig3, asig2*asig3
> instr 2
> asig1 oscil kenv1, cpspch(p5), 2
> asig2 oscil kenv2, cpspch(p5/2), 2
>
> asig3 oscil kenv1, cpspch(p5*2), 2
> asig4 oscil kenv2, cpspch(p5*3), 2
>
> outs asig2+asig4, asig1+asig3
> --
--
Josep M Comajuncosas
C/ Circumval.lacio 75 08790 Gelida - Penedes
Catalunya - SPAIN
home phone : 93 7792243 / 00 34 3 7792243
Csound page at http://members.tripod.com/csound/
|