I think this is not implemented at the right level (but maybe that is planned as a further step already?). In this PR a device seems to fully complete rendering a slice before moving on to the…
I noticed many of the Cycles tests were crashing. Here is a call stack from the transparent_shadow_hair.blend test
@ 0x43b9af4 ccl::GPUDevice::generic_free()
@ …I did a performance profile of this using all the benchmark scenes. The spreadsheet is available [here](https://docs.google.com/spreadsheets/d/1hycot3yqauzFa4L7A5WEk-OL3w3wfY4LrCAQf_Dx70U/edit?usp=…
Yes the reason they are similar is that I decided to keep the scene changes as is so that the performance differences could be compared directly. There are some performance wins from the…
Ok now MacOS is always disabled and otherwise the supportsRaytracing flag is used.
This change might enable MetalRT on Intel GPUs where it was always disabled before. For the purpose of this bugfix, it would be safest to keep it always disabled for Intel, because it may have…
Ok, I have rebased the changes to the blender-v3.6-release branch and clang-formatted the code changes. So it should be ready to go.
According to the documentation is should report if the device supports ray tracing correctly. Otherwise, it…
I've started looking into this in detail, trying to split off changes in separate commits to understand what is going on. I started doing that in #108454, but hit against deeper design…
