Delight Renderer : Register cloud rendering variant #5833
Conversation
|
Thanks @johnhaddon! Accepted the invite and will look into the PR and do some tests with the CI builds later today. |
Great, thanks! I've added you as a reviewer now too... |
|
Tested both under Linux and Windows and everything LGTM. One thing to note is that currently 3Delight Cloud on Windows has a critical bug - when 3Delight accesses certain files for upload to the cloud it runs into the following error which prevents the normal functioning of the cloud render:
The error doesn't happen under Linux and is not related to anything that Gaffer is doing. Houdini on Windows runs into the same issue as well. Reported the error to Aghiles and the 3Delight team are looking into it. |
The addresses we pass into the parameter list need to be valid until the `NSIBegin()` call returns. I think it's time we started to switch over to the NSI C++ API - the C API is too error-prone.
In an ideal world perhaps, we'd have implemented this as an option specified by DelightOptions, but that would put us in the chicken-and-egg situation where we can't make an NSI context without the cloud option, and we can't handle the other options until we have an NSI context. Although there _are_ strategies for dealing with that (see CyclesRenderer) they're complex and it would be preferable to avoid them. So instead this is implemented by registering a second "3Delight Cloud" renderer type, with the justification being that cloud rendering is a different enough operation that it's reasonable to present it to a user this way. In practicals terms there is no difference between this and a potential option, because the renderer can be specified using the standard `render:renderer` option.
johnhaddon
force-pushed
the
delightCloudRenderer
branch
from
024b924
to
cf12842
Compare
Thanks very much for testing! Merging now...
Ah, good to know that it's a general problem - fingers crossed for a quick fix. Thanks for investigating... |
@gkocov, I can't add you as a reviewer without you being a member of the GafferHQ team on GitHub, but I'd welcome your input anyway. I've also sent you an invite to be a "collaborator" on GitHub which I think would add you to the list of potential reviewers in future - feel free to accept or decline as you see fit.