Differential D6891
GPU: Add Image property to allow high bitdepth support on a per image basis Authored by Clément Foucault (fclem) on Feb 19 2020, 6:07 PM. Tags None Subscribers None
Details This adds the Half Float Precision option in the image property panel. Adding a flag inside the imbuf (IB_halffloat) on load is done for EXR and PSD formats that can store half floating point (16bits/channels). Related task T73086
Diff Detail
Event TimelineComment Actions I would consider automatic handling of half float cases in scope of this patch, or at least something that should be committed alongside this. The option should at least be enabled by default for existing files, to avoid having renders that worked before run out of memory. Comment Actions I've added support for detecting the half floats for EXR and PSD as it seems they are the only format that can support them. The other formats are using 16bits normalized int encoding or RGBE encoding (hdr). I don't know about dpx/cineon but I would guess they are normalized ints. Since normalized ints can have better precision than half floats I prefer to keep the option visible. The parameter is using inverted boolean so it's disabled by default (using low precision by default).
| ||||||||||||