Reworking renderer statistics

[cwichura]

Is it me, or when you reload a film file, does it no longer tell you how many samples the film file has? Given how broken network render statistics are, I would use this loading of the film file to get the total samples the film file ultimately ended up with. But now when I load a film file, the statistics pane at the bottom is blank.

[Omniflux]

when you reload a film file, does it no longer tell you how many samples the film file has?

This is correct. I am working on a solution.

Given how broken network render statistics are

The fix for this has been commited. It will be available in the next weekly build.

[SATtva]

One more issue i'm observing is stats fluctuate highly during rendering. In the past stats data was averaged, making the readings progressively stabilize, but it seems now only the last-second values are displayed without averaging.

Btw, averaging all the data since the beginning of rendering isn't the best approach either. Better is a rolling average, i.e. accumulate data only for the last, say, 10-minute window and average them. This way induced speed slowdowns (due to other concurrent processes, or changes in the number of rendering threads) will be dropped off after some time, and not affect the stats values forever.

[jeanphi]

Hi,

Instead of accumulation, the easiest way to do a rolling average is to apply a simple first order filtering.

Jeanphi

[Meelis]

Hi

When i hit pause for master and slave renders on then the total S/s stats goes wrong.
My slave prints out logs in slave window 0.00 - 1.20 - 1.60 and even 3.60 S/s but my total shows 50 S/s until slave sends new flm to master after 1000 sec interval while master is not paused.
And when i set my master display intervall for a really long time then i see master speed after first update (not when i resume master).

Also for the ERTP is there more accurate measure of speed than this small numbers.
With the ERTP it has been allways like that but the render looks way better than with other rendering methods that display much hihger speed numbers.

[Omniflux]

Fixes for the following issues have been commited:

• Stored samples are no longer displayed when loading a flm (as opposed to resuming)
• Statistics display incorrectly when pausing with slave nodes

One more issue i'm observing is stats fluctuate highly during rendering. In the past stats data was averaged, making the readings progressively stabilize, but it seems now only the last-second values are displayed without averaging.

I thought I had implemented the window the same way, but I guess not. I will go back and review the original code to see how to restore this.

[Dade]

One more issue i'm observing is stats fluctuate highly during rendering. In the past stats data was averaged, making the readings progressively stabilize, but it seems now only the last-second values are displayed without averaging.

I have the same problem, I was used to compare performance looking at the stats but there is too much variance at the moment.

[SATtva]

I thought I had implemented the window the same way, but I guess not. I will go back and review the original code to see how to restore this.

Thank you Omniflux.

[Omniflux]

I have added a first order filter over a sixty second window. Please let me know if this is sufficient or if it is still too jittery.

[SATtva]

Thanks a lot Omniflux, will test it shortly.