[Ardour-Users] Denormals

[Q]

On 30/12/11 13:00, Carl Hetherington wrote:
> On Thu, 29 Dec 2011, Q wrote:
>> Why are the "Flush To Zero" and "Denormals are zero" etc options
>> greyed out when these were available in prior versions of Ardour?
>
> If these are greyed out, it means that Ardour thinks your processor
> doesn't support those options. Has your system changed since you were
> using a version which didn't grey these out?

I have recently changed motherboard, but these options were greyed out 
with the old mobo. The processor hasn't been changed, I've been using an 
AMD Phenom Quad-core for the last three years and I'm reasonably sure 
I've been able to use the processor-handling options on this machine at 
some point in the past.

Until November, I had been using Ubuntu Studio 9.10 for a couple of 
years, so I think the only change was upgrading the version of Ardour 
that I was using at some point changed the availability of those options.

Is this an option I would set when building Ardour, or is it something 
to do with the kernel so that Ardour knows the processor can deal with 
denormals?

>
>> Going through and deactivating plugins hasn't shown any obvious culprits.
>
> Do you see the problems with no plugins at all?

I'd no problems when I started the mix (i.e. when there were no, and 
then very few, plugins).

Deactivating them all reduces the CPU load from about 35-40% to around 
5%. This rises a bit when playing, but returns to around 5% when the 
transport is stopped.

Earlier, I tried deactivating plugins track by track to see what had an 
effect on the CPU load. Just when I thought I'd narrowed it down to the 
plugins on particular track or buss (the CPU load would shoot up/down 
when activating/deactivating the plugins on that channel), it would 
shoot up again when they were deactivated, suggesting it was actually 
another plugin elsewhere that was probably the root of the problem.

I'm using EQ10Q (many instances), Invada Compressor (several instances), 
IR (two), TAP Tube Warmth, Invada Stereo Phaser, MS Spatialiser (a 
couple), Invada HPF, C* AMP VTS, C* Cabinet II, Richard Furse Simple 
Delay Line (a couple), Invada Delay Munge (a few), Tape Delay Simulation.

Going through and re-introducing plugins, I thought I'd found one 
instance of EQ10Q which causes the load to go from around 25% to over 
40%. However, playing long enough with the plugin deactivated and the 
load has again shot up. There's no apparent logic to it. I'll keep 
plugging away (ha!) at it...

The only other things I have running at the moment are Firefox and 
Thunderbird. I'm beginning to wonder whether there's some periodic 
system service that might be causing the spikes -- is that possible?

JACK's running in RT mode and at the highest latency it will allow, 92.9 
msec (2048 f/p, 2 p/b).

>> In the time it's taken to write this things have got worse -- with the
>> transport not running the CPU load is 100% all the time and racking up
>> thousands of xruns (does "Stop plugins with transport" have no effect?).
>
> I believe this option is broken in 2.0 and doesn't do anything.

Ah, that might explain why when I observed the session, the CPU load 
would rise and fall over a course of several few minutes even when the 
session wasn't playing.

>
> Carl

Thanks for your help. Any other ideas or suggestions gratefully accepted.

Q