Cooking optimization issue

Bugs
1

2023.12370 Win 10, RTX 3070

Hi,
Here is an example to reproduce:
cookSwitchBug.toe (403.1 KB)

  1. Open probe with GPU time mode
  2. change switch1 index from 0 to 4 to 0 sequentially so all four base COMPs cook at least once.
  3. Whatch probe: base COMPs are stopping cooking differently, and base3 never stop.

All four base COMPs are almost identical, but base3 COMP will remain seriously cooking regardless of switch1 index value - which is mostly likely caused by a combination of Feedback loop + soft switch + me.inputs[0].width in glslTOP uniform par expression.

Screenshot (251)
Screenshot (251)3513×1875 432 KB

Best,
Alex

2

Hi @letsius,

thanks for sending this and I can replicate. We’ll have a look.
Seems like some state is still stored in the file. I went in and temporarily connected the null1 TOP to the out2 TOP in base3, watched the cooking behavior and reconnected the glsl1 TOP to the out2 TOP. Now saving it out and reopening, the overcooking behavior is gone.

cookSwitchBug.7.toe (414.5 KB)
cheers
Markus

1 Like
3

Yes, it works. Thank you!