hdsp and midi problems

hdsp and midi problems

[Marcus Andersson]

Hi,

I have written a simple application that continuously reads one byte
from midi in on my hdsp multiface and prints the result to the console.
This is done with a plain loop without any calls to poll() or select().
When I run this loop as fast as I can, with 100% processor load, I loose
lots of bytes sent over midi. When I run the loop slower, I loose one
byte occasionally. Can this behaviour be explained? What should I do to
be sure to not loose any bytes? Is this a bug?

Marcus


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

Re: hdsp and midi problems

[Marcus Andersson]

More information,

when I run the program with 100% load using my VIA motherboard midi, I
loose no bytes as far as I can see.

I use alsa 0.9.4

Marcus

Marcus Andersson wrote:
> 
> Hi,
> 
> I have written a simple application that continuously reads one byte
> from midi in on my hdsp multiface and prints the result to the console.
> This is done with a plain loop without any calls to poll() or select().
> When I run this loop as fast as I can, with 100% processor load, I loose
> lots of bytes sent over midi. When I run the loop slower, I loose one
> byte occasionally. Can this behaviour be explained? What should I do to
> be sure to not loose any bytes? Is this a bug?
> 
> Marcus
> 
> -------------------------------------------------------
> This SF.Net email is sponsored by: INetU
> Attention Web Developers & Consultants: Become An INetU Hosting Partner.
> Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
> INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php
> _______________________________________________
> Alsa-devel mailing list
> Alsa-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/alsa-devel


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

RE: hdsp and midi problems

[Mark Knecht]

>
>
> More information,
>
> when I run the program with 100% load using my VIA motherboard midi, I
> loose no bytes as far as I can see.
>
> I use alsa 0.9.4
>
> Marcus

Marcus,
   The HDSP 9652 MIDI interface has (for me under Linux) always dropped lots
of notes. Mostly I find it interesting that it drops note off information
and doesn't seem to drop note-on. I cannot be sure how well it handles
controller information.

   The same machine booted into Windows runs perfectly, so I'm pretty sure
it's not a hardware issue.

Mark




-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

Re: hdsp and midi problems

[Marcus Andersson]

Mark Knecht wrote:
> 
> Marcus,
>    The HDSP 9652 MIDI interface has (for me under Linux) always dropped lots
> of notes. Mostly I find it interesting that it drops note off information
> and doesn't seem to drop note-on.

That is probably just a conincident depending on the read
pattern/processor load of your applications. Now it's just a matter of
convincing someone to debug it. I might even do it myself. If the ALSA
CVS would just compile. I have never experienced a project that cares so
little about code checked in being compilable. Frustrating.

>    The same machine booted into Windows runs perfectly, so I'm pretty sure
> it's not a hardware issue.

Mine works as well.

Marcus


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

Re: hdsp and midi problems

[Takashi Iwai]

At Mon, 23 Jun 2003 17:53:07 +0200,
Marcus Andersson wrote:
> 
> Mark Knecht wrote:
> > 
> > Marcus,
> >    The HDSP 9652 MIDI interface has (for me under Linux) always dropped lots
> > of notes. Mostly I find it interesting that it drops note off information
> > and doesn't seem to drop note-on.
> 
> That is probably just a conincident depending on the read
> pattern/processor load of your applications. Now it's just a matter of
> convincing someone to debug it. I might even do it myself. If the ALSA
> CVS would just compile. I have never experienced a project that cares so
> little about code checked in being compilable. Frustrating.

hmm, i have no problem at all for compiling.  what problem do you
have?

except for known compiler-check problem in the configure script, i've
not heard much about it.  at least, not reported to alsa-devel ML.
(a few days ago, there was a missing file, but it was already fixed.)


ciao,

Takashi


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

Re: hdsp and midi problems

[Marcus Andersson]

Takashi Iwai wrote:
> 
> hmm, i have no problem at all for compiling.  what problem do you
> have?

True, I don't know what the current CVS status is.

> except for known compiler-check problem in the configure script, i've
> not heard much about it.  at least, not reported to alsa-devel ML.
> (a few days ago, there was a missing file, but it was already fixed.)

Then this probably hit me the last time I tried to update. The problem
with alsa cvs is that, according to my experience, more often than not
it will not compile. This makes me think twice before I update my tree.
Developers should make an effort and check their code an extra time,
because a mistake affects so many.

Marcus


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

Re: hdsp and midi problems

[Takashi Iwai]

At Mon, 23 Jun 2003 18:46:35 +0200,
Marcus Andersson wrote:
> 
> Takashi Iwai wrote:
> > 
> > hmm, i have no problem at all for compiling.  what problem do you
> > have?
> 
> True, I don't know what the current CVS status is.
> 
> > except for known compiler-check problem in the configure script, i've
> > not heard much about it.  at least, not reported to alsa-devel ML.
> > (a few days ago, there was a missing file, but it was already fixed.)
> 
> Then this probably hit me the last time I tried to update. The problem
> with alsa cvs is that, according to my experience, more often than not
> it will not compile. This makes me think twice before I update my tree.
> Developers should make an effort and check their code an extra time,
> because a mistake affects so many.

yeah, that's true.

however, the most case in which the compilation breaks is a missing
check-in of some files (just like above).  and it's difficult to fix,
because forgetfulness is a nature of human being :)

such a problem may be reduced if we go to bitkeeper.
(or, even subvesion would help much in this regard?)


ciao,

Takashi


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

Re: hdsp and midi problems

[Marcus Andersson]

Takashi Iwai wrote:
> however, the most case in which the compilation breaks is a missing
> check-in of some files (just like above).  and it's difficult to fix,
> because forgetfulness is a nature of human being :)

Yes, I know it's easy to forget things. But why not automate the
checking? That's what we use computers for, automate difficult things.
Missing files are marked with a question mark, and changed files can be
checked with status.

Marcus


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

Re: hdsp and midi problems

[Jaroslav Kysela]

On Mon, 23 Jun 2003, Takashi Iwai wrote:

> At Mon, 23 Jun 2003 17:53:07 +0200,
> Marcus Andersson wrote:
> > 
> > Mark Knecht wrote:
> > > 
> > > Marcus,
> > >    The HDSP 9652 MIDI interface has (for me under Linux) always dropped lots
> > > of notes. Mostly I find it interesting that it drops note off information
> > > and doesn't seem to drop note-on.
> > 
> > That is probably just a conincident depending on the read
> > pattern/processor load of your applications. Now it's just a matter of
> > convincing someone to debug it. I might even do it myself. If the ALSA
> > CVS would just compile. I have never experienced a project that cares so
> > little about code checked in being compilable. Frustrating.
> 
> hmm, i have no problem at all for compiling.  what problem do you
> have?
> 
> except for known compiler-check problem in the configure script, i've

The last issue should be fixed as well.

						Jaroslav

-----
Jaroslav Kysela <perex@suse.cz>
Linux Kernel Sound Maintainer
ALSA Project, SuSE Labs



-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

Re: hdsp and midi problems

[Takashi Iwai]

At Mon, 23 Jun 2003 18:46:36 +0200 (CEST),
Jaroslav wrote:
> 
> On Mon, 23 Jun 2003, Takashi Iwai wrote:
> 
> > At Mon, 23 Jun 2003 17:53:07 +0200,
> > Marcus Andersson wrote:
> > > 
> > > Mark Knecht wrote:
> > > > 
> > > > Marcus,
> > > >    The HDSP 9652 MIDI interface has (for me under Linux) always dropped lots
> > > > of notes. Mostly I find it interesting that it drops note off information
> > > > and doesn't seem to drop note-on.
> > > 
> > > That is probably just a conincident depending on the read
> > > pattern/processor load of your applications. Now it's just a matter of
> > > convincing someone to debug it. I might even do it myself. If the ALSA
> > > CVS would just compile. I have never experienced a project that cares so
> > > little about code checked in being compilable. Frustrating.
> > 
> > hmm, i have no problem at all for compiling.  what problem do you
> > have?
> > 
> > except for known compiler-check problem in the configure script, i've
> 
> The last issue should be fixed as well.

i think another buglet in gentoo's list is not solved yet.
either gcc -v or compiler.h says something like

	xxxx-gcc-x.yz (GCC) x.yz (xxxx)

whether we handle only the case gcc-x.yz (GCC) (xxx).

won't it better to add and option to prevent the gcc version check?
of course, the version check should be enabled as default, though.


Takashi


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

Re: hdsp and midi problems

[Takashi Iwai]

[-- Attachment #1: Type: text/plain, Size: 964 bytes --]

At Mon, 23 Jun 2003 06:38:20 -0700,
Mark Knecht wrote:
> 
> >
> >
> > More information,
> >
> > when I run the program with 100% load using my VIA motherboard midi, I
> > loose no bytes as far as I can see.
> >
> > I use alsa 0.9.4
> >
> > Marcus
> 
> Marcus,
>    The HDSP 9652 MIDI interface has (for me under Linux) always dropped lots
> of notes. Mostly I find it interesting that it drops note off information
> and doesn't seem to drop note-on. I cannot be sure how well it handles
> controller information.

do you mean dropping notes in reading MIDI?
well, IIRC, you reported also that the behavior of MIDI out is
strange...  oh sigh.

> 
>    The same machine booted into Windows runs perfectly, so I'm pretty sure
> it's not a hardware issue.

dropping a note is surely a bug of the driver.
unless the FIFO overflows, no dropping should happen.

btw, could you try the attached patch?
(and in-advance-scheduled excuse if it causes a hang-up :)


Takashi

[-- Attachment #2: hdsp-midi-test.dif --]
[-- Type: application/octet-stream, Size: 2422 bytes --]

Index: alsa-kernel/pci/rme9652/hdsp.c
===================================================================
RCS file: /suse/tiwai/cvs/alsa/alsa-kernel/pci/rme9652/hdsp.c,v
retrieving revision 1.34
diff -u -r1.34 hdsp.c
--- alsa-kernel/pci/rme9652/hdsp.c	18 Jun 2003 10:34:11 -0000	1.34
+++ alsa-kernel/pci/rme9652/hdsp.c	23 Jun 2003 14:04:20 -0000
@@ -1103,7 +1134,7 @@
 	int i;
 
 	spin_lock_irqsave (&hmidi->lock, flags);
-	if ((n_pending = snd_hdsp_midi_input_available (hmidi->hdsp, hmidi->id)) > 0) {
+	while ((n_pending = snd_hdsp_midi_input_available (hmidi->hdsp, hmidi->id)) > 0) {
 		if (hmidi->input) {
 			if (n_pending > (int)sizeof (buf)) {
 				n_pending = sizeof (buf);
@@ -1122,12 +1153,6 @@
 		}
 	}
 	hmidi->pending = 0;
-	if (hmidi->id) {
-		hmidi->hdsp->control_register |= HDSP_Midi1InterruptEnable;
-	} else {
-		hmidi->hdsp->control_register |= HDSP_Midi0InterruptEnable;
-	}
-	hdsp_write(hmidi->hdsp, HDSP_controlRegister, hmidi->hdsp->control_register);
 	spin_unlock_irqrestore (&hmidi->lock, flags);
 	return snd_hdsp_midi_output_write (hmidi);
 }
@@ -3155,35 +3157,38 @@
 
 	hdsp_write(hdsp, HDSP_interruptConfirmation, 0);
 
+	spin_lock (&hdsp->lock);
+
 	midi0status = hdsp_read (hdsp, HDSP_midiStatusIn0) & 0xff;
 	midi1status = hdsp_read (hdsp, HDSP_midiStatusIn1) & 0xff;
 	
 	if (audio) {
 		if (hdsp->capture_substream) {
+			spin_unlock (&hdsp->lock);
 			snd_pcm_period_elapsed(hdsp->pcm->streams[SNDRV_PCM_STREAM_CAPTURE].substream);
+			spin_lock (&hdsp->lock);
 		}
 		
 		if (hdsp->playback_substream) {
+			spin_unlock (&hdsp->lock);
 			snd_pcm_period_elapsed(hdsp->pcm->streams[SNDRV_PCM_STREAM_PLAYBACK].substream);
+			spin_lock (&hdsp->lock);
 		}
 	}
 	
 	if (midi0 && midi0status) {
-		/* we disable interrupts for this input until processing is done */
-		hdsp->control_register &= ~HDSP_Midi0InterruptEnable;
-		hdsp_write(hdsp, HDSP_controlRegister, hdsp->control_register);
 		hdsp->midi[0].pending = 1;
 		schedule = 1;
 	}
 	if (midi1 && midi1status) {
-		/* we disable interrupts for this input until processing is done */
-		hdsp->control_register &= ~HDSP_Midi1InterruptEnable;
-		hdsp_write(hdsp, HDSP_controlRegister, hdsp->control_register);
 		hdsp->midi[1].pending = 1;
 		schedule = 1;
 	}
 	if (schedule)
-	    tasklet_hi_schedule(&hdsp->midi_tasklet);
+		tasklet_hi_schedule(&hdsp->midi_tasklet);
+
+	spin_unlock (&hdsp->lock);
+
 	return IRQ_HANDLED;
 }
 

RE: hdsp and midi problems

[Mark Knecht]

 > Marcus,
> >    The HDSP 9652 MIDI interface has (for me under Linux) always
> dropped lots
> > of notes. Mostly I find it interesting that it drops note off
> information
> > and doesn't seem to drop note-on. I cannot be sure how well it handles
> > controller information.
>
> do you mean dropping notes in reading MIDI?
> well, IIRC, you reported also that the behavior of MIDI out is
> strange...  oh sigh.

Yes, but I tested Alsa (a long time ago) with both MIDI-IN driving soft
synths, and then MIDI-IN/MIDI-OUT driving hardware synths and got the same
'note stuck on' problem. For that reason I say it's a 'MIDI-IN dropped note'
problem, but certainly I do not know this for sure.

>
> dropping a note is surely a bug of the driver.
> unless the FIFO overflows, no dropping should happen.
>
> btw, could you try the attached patch?
> (and in-advance-scheduled excuse if it causes a hang-up :)
>

I would be happy to try a bit of patching. However, I am on
Redhat/PlanetCCRMA and Gentoo, both of which provide prebuilt Alsa
environments. I can try patching the Gentoo one which is pretty recent, but
it may take a few days before I can get back to you.




-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

Re: hdsp and midi problems

[Marcus Andersson]

Takashi Iwai wrote:
> btw, could you try the attached patch?
> (and in-advance-scheduled excuse if it causes a hang-up :)

I will try it. Do you think I can safely apply it to 0.9.4, or should I
use cvs?

Marcus


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php

Re: hdsp and midi problems

[Marcus Andersson]

Marcus Andersson wrote:
> 
> Takashi Iwai wrote:
> > btw, could you try the attached patch?
> > (and in-advance-scheduled excuse if it causes a hang-up :)
> 
> I will try it. Do you think I can safely apply it to 0.9.4, or should I
> use cvs?

Tested with current cvs. It locks up my computer. I used alsa-lib 0.9.4,
as I can't build the latest:

mkdir .libs
rm -fr .libs/libcontrol.la .libs/libcontrol.* .libs/libcontrol.*
(cd . && ln -s cards.lo cards.lo)
(cd . && ln -s hcontrol.lo hcontrol.lo)
(cd . && ln -s control.lo control.lo)
(cd . && ln -s control_hw.lo control_hw.lo)
(cd . && ln -s control_shm.lo control_shm.lo)
(cd . && ln -s setup.lo setup.lo)
(cd . && ln -s control_symbols.lo control_symbols.lo)
ar cru .libs/libcontrol.al cards.lo hcontrol.lo control.lo control_hw.lo
control_shm.lo setup.lo control_symbols.lo
ar: cards.lo: Too many levels of symbolic links
make[2]: *** [libcontrol.la] Error 1
make[2]: Leaving directory `/usr/local/src/alsa/alsa-lib/src/control'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/usr/local/src/alsa/alsa-lib/src'
make: *** [all-recursive] Error 1
debian:/usr/local/src/alsa/alsa-lib#

Probably my system configuration that is screwed up.

But I don't think it matters, as I open() the device directly.

Marcus


-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting http://www.inetu.net/partner/index.php