Different render outcome for same input

Software Luminance HDR
1

I am still on 2.4 because 2.5.-branch gives me extend noise and artifacts in Fattal operator.

For 2.4, I am already noticing for more than a year that let’s say doing a Fattal render (but I have also saw this behaviour for other operators) for 10 times. 6/10 would be a same good stable result. 2 would be different e.g. more saturated and the other 2 would be really messed up with RGB noise in light areas.

Anyone else who experienced this?

(Different render outcome for same input · Issue #121 · LuminanceHDR/LuminanceHDR · GitHub)

2
  1. There is no 5.5 branch.
  2. When reporting bugs, always test using the latest development version.
  3. Report bugs in the bug tracker.
  4. Specifics are important, see this for all the info which is missing (its for RT but the same applies to any program): How to write useful bug reports - RawPedia
3

Typo fixed.
Asking here to see if more users are having this issue, then see what the next step is.
Using the latest version is useless for me since Fattal is useless in that version.

4

@Morgan_Hardwood

In 2.5 nightly this bug seems to be gone. As you know I am getting bad results with 2.5-branch thus will stay on 2.4 until there is a fix or workaround.

What I did is:
Times 10: Start 2.4, import the sample set.
Record the screen while browsing the results.

Hope it’s clear now that the result of import differs A LOT.

5
  1. So is the bug gone or not?
  2. There is no 2.5-branch.
  3. If you found a bug in the latest code in the master branch (which at the time of writing is v.2.5.1-245-g7dc448a6) then please report it in the bug tracker: Issues · LuminanceHDR/LuminanceHDR · GitHub
  4. To see which version you’re using, type git describe. The branch + hash are the most important bits of info.
6

Using v.2.5.1-245-g7dc448a6, the HDR images generated multiple times from the same images using the same settings are consistent.

screenshot_20180901_132806
screenshot_20180901_1328061960×1088 1.24 MB

7

@Morgan_Hardwood
Yes it is safe to assume this bug is not in 2.5.2 5f582a. (Installer ‘master v.2.5.1-241-g5f582a3e.’)
Since 2.5* doesn’t deliver the quality/result I got with 2.4, I am sticking with 2.4 for now. (2.5.1 doing very bad with my images compared to 2.4)