[gpz]

I made a very simple modification in r1702 for trying to resolve the threading issue of the discrete triggers being called from Updater process. The fix works in a way as James instructed: the trigger is being flagged by Updater process, while the sound command is run by the Sound process. Searching for the trigger is still done by Updater process.

Could you please try if it really works?

[James Ross]

 gpz, on 05 August 2013 - 11:36 AM, said:

Could you please try if it really works?

Cool, seems to have done the trick. Profiler is showing no unnecessary file accesses from render/updater processes now.

Would be nice to hear from Rob too.

[Matej Pacha]

I have this Dell X15 notebook with a very poor performance when running on its primary display. This is a pretty cool tool for the low-end PC debugging. I've noticed that there is some kind of pulse that causes the adhesion model to get crazy for a while. I've tried to edit the model, but I need to set some parameters too far from reality to get some stable behavior. I have one more ace in my sleeve - I can make a floating average filter with variable buffer to eliminate the oscilations. It should help for such a low FPS.
There are some pictures. Notice the Motive force oscilations - it should be constant for all the time.
BTW I love this in-game chart with names and units. Thanks a lot ;)



Matej

[James Ross]

 Matej Pacha, on 06 August 2013 - 04:38 AM, said:

I have one more ace in my sleeve - I can make a floating average filter with variable buffer to eliminate the oscilations.

Feel free to have a look at the SmoothedData class, which performs a (time-based) weighted-moving-average on data. It defaults to a 3s "weight" (I'm not entirely sure how you'd describe it), which we use for FPS and some other data, but you can construct it with a lower value to have it react quicker. If it does what you'd like, use it, but if not please create something new. :)

 Matej Pacha, on 06 August 2013 - 04:38 AM, said:

BTW I love this in-game chart with names and units. Thanks a lot ;)

No problem; it's really fun for me to have visual insight in to my own area (graphics) and others like physics.

[Matej Pacha]

Added some stuff in v1704. Let's see if the problem remains...

[roeter]

 James Ross, on 05 August 2013 - 12:47 PM, said:

Would be nice to hear from Rob too.

And here I am.

Below is how playing the horn now looks - same situation as above, but this time no peak in update and related drop in render process. The action has clearly moved to the sound process.



The loading of other files has not changed, but the fluctuation in frame-rates now no longer causes wheelslip.

Well done indeed ;) .

Regards,

Rob Roeterdink

[James Ross]

 roeter, on 06 August 2013 - 02:13 PM, said:

The loading of other files has not changed, but the fluctuation in frame-rates now no longer causes wheelslip.

Glad to hear it. ;)

[Matej Pacha]

Oh yeah!

[gpz]

 roeter, on 06 August 2013 - 02:13 PM, said:

no peak in update and related drop in render process. The action has clearly moved to the sound process.

By using a ConcurrentBag object of .net 4 the rest of this searching could be moved to sound process too.