lo10K1 not working on x86_64 ?

lo10K1 not working on x86_64 ?

[Corvus Corax]

Hi

It seems like lo10k1 is not working correctly on x86_64

the emu10k1 driver itself works ok, but loading anything (for example
with emu_init) results either in

- no change at all
- sound in front left speaker distorted
- no sound

just running emu_init results in distorted sound

i tried qlo10k1 and loaded several patches manually or unloaded them from the
default provided emu_init. all that changes are different heavy distortions

(distortion = sounds like extreme clipping doe to over amilification)

is the 64bit address stuff causing this, or could as10k1 eventually produce
incorrect DSP machine code? (compiled DSP assembler code on x86_64 seems
to have differed from code previously assembled on x86, but Im not sure if the
sources were 100% identical)

tested with an

0000:01:07.0 Class 0401: 1102:0002 (rev 0a)
        Subsystem: 1102:8065
        Flags: bus master, medium devsel, latency 32, IRQ 21
        I/O ports at a000 [size=32]
        Capabilities: [dc] Power Management version 1
(
0000:01:07.0 Multimedia audio controller: Creative Labs SB Live! EMU10k1 (rev
0a)        Subsystem: Creative Labs SBLive! 5.1 Digital Model SB0220
        Flags: bus master, medium devsel, latency 32, IRQ 21
        I/O ports at a000 [size=32]
        Capabilities: [dc] Power Management version 1
)
on an AMD64 X2 machine

custom patches previously worked flawlessly with the OSS emu10k1 driver under
x86, OSS dspmgr (both compiled for 32 and 64) segfaults when run with the 64bit
kernel compiled oss emu10k1 driver

I did not test alsa ld10k1 with the old system


regards

Corvus


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php

Re: lo10K1 not working on x86_64 ?

[Mikael Magnusson]

On Wed, Nov 02, 2005 at 09:20:09PM +0100, Corvus Corax wrote:
> 
> Hi
> 
> It seems like lo10k1 is not working correctly on x86_64
> 
> the emu10k1 driver itself works ok, but loading anything (for example
> with emu_init) results either in
> 
> - no change at all
> - sound in front left speaker distorted
> - no sound
> 
> just running emu_init results in distorted sound
> 
> i tried qlo10k1 and loaded several patches manually or unloaded them from the
> default provided emu_init. all that changes are different heavy distortions
> 
> (distortion = sounds like extreme clipping doe to over amilification)
> 
> is the 64bit address stuff causing this, or could as10k1 eventually produce
> incorrect DSP machine code? (compiled DSP assembler code on x86_64 seems
> to have differed from code previously assembled on x86, but Im not sure if the
> sources were 100% identical)

Which version of alsa-driver and alsa-tools are you trying. You will
need at least version 1.0.10rc1, since it contains bug fixes to ld10k1
that are related to 64bit systems.

/Mikael



-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php

Re: lo10K1 not working on x86_64 ?

[Corvus Corax]

Am Thu, 3 Nov 2005 13:01:17 +0100
schrieb Mikael Magnusson <mikaelmagnusson@glocalnet.net>:

> On Wed, Nov 02, 2005 at 09:20:09PM +0100, Corvus Corax wrote:
> > 
> > Hi
> > 
> > It seems like lo10k1 is not working correctly on x86_64
> > 
> > the emu10k1 driver itself works ok, but loading anything (for example
> > with emu_init) results either in
> > 
> > - no change at all
> > - sound in front left speaker distorted
> > - no sound
> > 
> > just running emu_init results in distorted sound
> > 
> > i tried qlo10k1 and loaded several patches manually or unloaded them from
the
> > default provided emu_init. all that changes are different heavy distortions
> > 
> > (distortion = sounds like extreme clipping doe to over amilification)
> > 
> > is the 64bit address stuff causing this, or could as10k1 eventually produce
> > incorrect DSP machine code? (compiled DSP assembler code on x86_64 seems
> > to have differed from code previously assembled on x86, but Im not sure if
the
> > sources were 100% identical)
> 
> Which version of alsa-driver and alsa-tools are you trying. You will
> need at least version 1.0.10rc1, since it contains bug fixes to ld10k1
> that are related to 64bit systems.
> 
> /Mikael
> 

I think I had it running with 1.0.10rc1 on both driver and tools, but Im not
completely sure about alsa-driver anymore.

I have now tested again with both at 1.0.10rc2, everything compiled 64bit.

init_live still creates a setup with heavily distorted output, but when cleaning
the DSP setup with qlo10k1 and setting up a single master volume control slider
and routing from PCM to different speakers, I have been able to change their
volume successfully afterwards - so most of the backend seems to be working.

Maybe specific patches have problems, or the init_live scrip doesnt fit my card,
I'll give it some further tests, trying to find a working routing with adequate
mixer controls. Thanks for answer :-)

regards

Corvus


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php

Re: lo10K1 not working on x86_64 ?

[Corvus Corax]

I have tested it again now, and failed to get a useable configuration doe to the
following problems.

As soon as I load a more complex patch set - by init_live - or just by inserting
multiple master_volume patches (Am I correct that the 6 channel master volume is
the only one with a single slider to control 2 channels i.e. left and right
component? because the 2-channel Vol sets up separate mixer sliders for each
control) I get heavy distortion and the actual output seems to have nothing to
do anymore with the intention of the patchset, I have then to load a different
saved set from disk, or clear the DSP / unload all patches manually, and load
new ones to get normal output again.

That happens already when loading the patchsets, usually before actually
connecting their I/O channels together. Its almost like they would overwrite
eachother on loading, creating garbage in the DSP memory...

However a single patch does work as expected as it seems, at least for volume
controls and routings.

I haven't tested the FX-BUS because I dont know how that one is supposed to
work. On loading that patch over an existing routing I usually get the mentioned
distortion and most mixer controls cease to function.

By the way, is there a way to control custom volume sliders with OSS mixer
controls? Does the OSS emulation driver have to be re-loaded after loading the
patchset? I get mixer program hangs when trying to use OSS mixer controls, and
when overriding Alsa PCM or Master sliders with new ones, their OSS counterparts
still control only the left/right channels like with no patchset.

regards

Corvus



-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php

Re: lo10K1 not working on x86_64 ?

[D. Hugh Redelmeier]

| From: Corvus Corax <corvus-ml@cybertrench.com>
| 
| I have tested it again now, and failed to get a useable configuration doe to the
| following problems.

I know nothing about what ld10k1 does nor have I run it, but I did
submit some fixes to its code.  These fixes correct a bug that would
manifest itself on x86_64.  They are now in CVS (as of today).

See https://bugtrack.alsa-project.org/alsa-bug/view.php?id=1533

Perhaps this fixes the problems you observed.  I have no way of
knowing.  You might consider testing HEAD.


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php

Re: lo10K1 not working on x86_64 ?

[Peter Zubaj]

Hi,

Could you send output of lo10k1 when this happen ?

lo10k1 --debug 1
lo10k1 --debug 2
lo10k1 --debug 3
lo10k1 --debug 4
lo10k1 --debug 5
lo10k1 --debug 6
lo10k1 --debug 7
lo10k1 --debug 8
lo10k1 --debug 9

and for every loaded effect (less loaded effect will be better)
lo10k1 --debug 100
lo10k1 --debug 101

up to last loaded (for example if there are 5 loaded last will be 104)

lo10k1 --debug 104

and content of fx8010_acode from /proc/asound card directory) too

Peter Zubaj

On Wed, 2005-11-09 at 14:13 +0100, Corvus Corax wrote:
> As soon as I load a more complex patch set - by init_live - or just by inserting
> multiple master_volume patches (Am I correct that the 6 channel master volume is
> the only one with a single slider to control 2 channels i.e. left and right
> component? because the 2-channel Vol sets up separate mixer sliders for each
> control) I get heavy distortion and the actual output seems to have nothing to
> do anymore with the intention of the patchset, I have then to load a different
> saved set from disk, or clear the DSP / unload all patches manually, and load
> new ones to get normal output again.






-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php

Re: lo10K1 not working on x86_64 ?

[Corvus Corax]

[-- Attachment #1: Type: text/plain, Size: 931 bytes --]

Am Wed, 09 Nov 2005 20:02:33 +0100
schrieb Peter Zubaj <pzad@pobox.sk>:

> Could you send output of lo10k1 when this happen ?
> 
> lo10k1 --debug 1
> lo10k1 --debug 2
> lo10k1 --debug 3
> lo10k1 --debug 4
> lo10k1 --debug 5
> lo10k1 --debug 6
> lo10k1 --debug 7
> lo10k1 --debug 8
> lo10k1 --debug 9
> 
> and for every loaded effect (less loaded effect will be better)
> lo10k1 --debug 100
> lo10k1 --debug 101
> 
> up to last loaded (for example if there are 5 loaded last will be 104)
> 
> lo10k1 --debug 104
> 
> and content of fx8010_acode from /proc/asound card directory) too

yep, tar.gz has been attached for 5 different setups,
2 of them in "distorted condition".

when glancing at it I noticed that parts of the DSP memory seem to be zeroed out
once the distortion occurs, looks like something goes seriously wrong on
loading the patches.

tell me if you need any additional information.

thanks and best regards

Corvus

[-- Attachment #2: emu10k1_debug_corvus.tgz --]
[-- Type: application/x-compressed-tar, Size: 25511 bytes --]

Re: lo10K1 not working on x86_64 ?

[Corvus Corax]

Am Wed, 9 Nov 2005 13:47:55 -0500 (EST)
schrieb "D. Hugh Redelmeier" <hugh@mimosa.com>:

> I know nothing about what ld10k1 does nor have I run it, but I did
> submit some fixes to its code.  These fixes correct a bug that would
> manifest itself on x86_64.  They are now in CVS (as of today).

Sorry, I can't seem to get CVS source access

pserver access seems to be prohibited for some reason, and my sourceforge user
has no cvs lock rights on the also project cvs:

$  cvs -z3 -d:pserver:corvuscorax@cvs.sourceforge.net:/cvsroot/alsa login
Logging in to :pserver:corvuscorax@cvs.sourceforge.net:2401/cvsroot/alsa
CVS password: 
cvs [login aborted]: connect to cvs.sourceforge.net(66.35.250.207):2401 failed:
Connection refused


$ cvs -z3 -d:ext:corvuscorax@cvs.sourceforge.net:/cvsroot/alsa co alsa-tools
cvs checkout: warning: cannot write to history file
/cvsroot/alsa/CVSROOT/history: Permission denied cvs checkout: Updating
alsa-tools cvs checkout: failed to create lock directory for
`/cvsroot/alsa/alsa-tools' (/cvsroot/alsa/alsa-tools/#cvs.lock): Permission
denied cvs checkout: failed to obtain dir lock in repository
`/cvsroot/alsa/alsa-tools' cvs [checkout aborted]: read lock failed - giving up


Since when doesn't SF allow anonymous pserver access? Is this permanent?!?


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php

Re: lo10K1 not working on x86_64 ?

[Corvus Corax]

Am Wed, 9 Nov 2005 23:18:21 +0100
schrieb Corvus Corax <corvus-ml@cybertrench.com>:

> 
> Sorry, I can't seem to get CVS source access
> 

sorry, never mind for now, got pserver access to cvs1 out from shell.sf.net,
still IMHO this is a bit complicated just to get HEAD sources -
what if I hadn't been already member of at least some sf projects with shell
server access?



-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php

Re: lo10K1 not working on x86_64 ?

[Corvus Corax]

Am Wed, 9 Nov 2005 13:47:55 -0500 (EST)
schrieb "D. Hugh Redelmeier" <hugh@mimosa.com>:

> 
> I know nothing about what ld10k1 does nor have I run it, but I did
> submit some fixes to its code.  These fixes correct a bug that would
> manifest itself on x86_64.  They are now in CVS (as of today).
> 
> See https://bugtrack.alsa-project.org/alsa-bug/view.php?id=1533
> 
> Perhaps this fixes the problems you observed.  I have no way of
> knowing.  You might consider testing HEAD.
> 

Confirmed, the problem seems NOT to occur with ld10k1 compiled from the current
CVS HEAD

init_live now produces a working routing.

Thanks all for your help :-))


best regards


Corvus Corax


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php

Re: lo10K1 not working on x86_64 ?

[D. Hugh Redelmeier]

| From: Corvus Corax <corvus-ml@cybertrench.com>

| Am Wed, 9 Nov 2005 13:47:55 -0500 (EST)
| schrieb "D. Hugh Redelmeier" <hugh@mimosa.com>:

| > See https://bugtrack.alsa-project.org/alsa-bug/view.php?id=1533

| >  You might consider testing HEAD.
| 
| Confirmed, the problem seems NOT to occur with ld10k1 compiled from the current
| CVS HEAD

Great!  Glad to hear that your problem is solved.

My bug fix came from reading a small code fragment.  I never figured
out the big picture of what the symptoms would be.  I'd like to know
if the symptoms you encountered were due to the bug I fixed or due to
some other change between the earlier version you tested and HEAD.

It would be great if you would test your working version but with the
file bitops.h reverted to what it was before the fix (version 1.2).

I realise that this is asking a lot -- it won't benefit you since you
already have a working system.  But it should help me understand the
consequences of the fix.

Thanks.


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php

Re: lo10K1 not working on x86_64 ?

[Corvus Corax]

Am Thu, 10 Nov 2005 12:09:32 -0500 (EST)
schrieb "D. Hugh Redelmeier" <hugh@mimosa.com>:

> 
> Great!  Glad to hear that your problem is solved.
> 
> My bug fix came from reading a small code fragment.  I never figured
> out the big picture of what the symptoms would be.  I'd like to know
> if the symptoms you encountered were due to the bug I fixed or due to
> some other change between the earlier version you tested and HEAD.
> 
> It would be great if you would test your working version but with the
> file bitops.h reverted to what it was before the fix (version 1.2).
> 
> I realise that this is asking a lot -- it won't benefit you since you
> already have a working system.  But it should help me understand the
> consequences of the fix.
> 
> Thanks.
> 

Confirmed!

I tried the packpatch to version 1.2 of bitops.h,
with that the error occurs like described, DSP meory gets garbed when a certain
amount or combination of patches is loaded,
resultuing in heavy distorted sound output and non-functioning controls.

(See tar.gz attached to other mail in this thread for DSP patch memory dumps)

With the actual version, no corruption occurs, patches get loaded successfully
and work as expected.

regards

Corvus Corax

PS: thanks for fixing that bug :)


-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php