YOUR ACCOUNT

Login or Register to post new topics or replies
Sphinx.
Filter Optimizer

Posts: 1750
Filters: 39
Please add some sort of performance evaluation functionality to filter forge.

This request is motivated by the two following challenges:
1. being able to identify performance of library filters (I'm not interested in dead slow filters)
2. being able to optimize/construct more optimal filters

Some ideas for a solution:
- Profile the sample time of components inside filters and add it all up
- Do some basic internal rating of components. Count the number of sample draws (it would be very useful if I could see the bomber/whatever component being queried 18 times output sample or so)
- Time a known complex build in filter (preferably something that uses a lot of different components), compare any other filter performance to that reference and use it to put together some absolute performance rating that anyone can relate to.
  Details E-Mail
Betis
The Blacksmith

Posts: 1207
Filters: 76
+1!
Roses are #FF0000
Violets are #0000FF
All my base are belong to you.
  Details E-Mail
raabix
Nodist
Posts: 121
Filters: 45
+2
I would totally support the filter timing - it would also help to figure out, which components are slowing down your filter (for those who don't know yet that blurs, highpass, etc. are expensive)

One idea I am having here: You can enable 'analyze-mode' which will highlight the nodes in your filter that use most cycles. (see the mockup that explains it)

Or do it like the <ahem> substance guys </ahem> who print the timing under each node. I guess that's not something that they have patented. I am often surprised how many noises I can stack, but then one single blur takes away all my performance. Knowing about timings is crucial for ambitious filter designers.

  Details E-Mail
CorvusCroax
CorvusCroax

Posts: 1227
Filters: 18
I'd really like something like this too: any kind of semi-objective way to evaluate performance would really be helpful. It would be especially useful to people who are learning FF.

BTW: here is what the Substance UX looks like: check out the green numbers under the nodes.


I don't think it has to be a *perfect* tool: just something to help the filter-maker decide between different ways of doing things. An 80% solution would be fine; since FF can already calculate the render time, having a little sub-mode where it breaks it into proportional steps seems do-able.
  Details E-Mail
Grimbly
Grimbly
Posts: 68
I'm in complete agreement about this as well. Any node based program like this will always benefit from node level timing like Substance has. It would benefit users for more than just debugging purposes too. In general having this readily displayed for each node would eventually help in the way people construct nodes from the very beginning. They would unconsciously be thinking of these timers and how to build the node graph more efficiently after enough time and exposure to them.
  Details E-Mail
mitaywalle
Great to meet you :)

Posts: 177
Filters: 63
+1 bumbed. Sometime I'm forced to use Substance Designer cause of this feature
Randomize everithing!
  Details E-Mail

Join Our Community!

Filter Forge has a thriving, vibrant, knowledgeable user community. Feel free to join us and have fun!

33,711 Registered Users
+18 new in 30 days!

153,533 Posts
+31 new in 30 days!

15,348 Topics
+73 new in year!

Create an Account

Online Users Last minute:

35 unregistered users.