Cycles X: main development happening is still regarding tiled and resumable rendering. Sergey has many of the pieces, patches are being incrementally finished and review. No major other changes besides small feature completeness and fixes.
Anisotropic SSS is practically ready, expected to be committed within the next few days. Only needs a few smaller changes to work more robustly with high albedo and anisotropy. Christophe will start on MNEE next.
AMD sent initial code and early drivers for HIP implementation for Cycles X. This is not yet available publicly, and does not yet support hardware ray-tracing. Performance looks promising comparing to CUDA. Cycles developers will look at the code and give early feedback, by the end of August. William will be the main Cycles developer responsible for feedback and testing of this backend.
Kévin created patches to reduce memory usage and copying overhead of positions and normals for texturing and normal mapping. Patches are under review.
Facebook has a patch to avoid copying of geometry arrays for cases like the Alembic procedural, by referencing memory instead of copying it. Patch will be submitted for review.
William still works on improving the PMJ sampling implementation so we can make it the only sampling pattern. Still running benchmarks, performance and noise is getting close to Sobol. It’s a bit slower still, but required for adaptive sampling which can bring much bigger speedups.
Eevee & Viewport
Kévin is still working on OpenSubdiv support in the dependency graph. Also still looking into an issue with compressed normals.
Other
There was a long discussion regarding USD, Hydra, MDL, MaterialX, etc. Much of this is topics to be discussed of the bf-usd mailing list, but it also touches rendering of course. In general the Blender project is interested in USD and MaterialX integration. A few relevant links:
Tangent Animation unfortunately shut down recently, which also means it’s unclear what will happen with their Cycles implementation. Fortunately much of this is open-source, and we can port over useful features. As already agreed previously, we are interested in having a USD hydra delegate in the Cycles repository, it’s just a matter of finding development time for it or a contributor picking this up. A presentation on the Cycles work at Tangent was given recently by Stefan Werner.
Practical Info
This is a weekly video chat meeting for planning and discussion of Blender rendering development. Any contributor (developer, UI/UX designer, writer, …) working on rendering in Blender is welcome to join and add proposed items to the agenda.
For users and other interested parties, we ask to read the meeting notes instead so that the meeting can remain focused.
@sybren We had a long discussion with the Facebook labs groups (they have some great industry experience). That might be worth having a followup video meeting to discuss their workflow ideas in USD. And we’d like to present our ideas with the Hydra addon. Thoughts?
As mentioned in the notes, this is not yet available publicly. We don’t have a date yet when it will be possible for users to test this, probably somewhere in October earliest.
amd want a piece of the pie in the civilian professional market in blender on the project must be done well, if even they can not do well can also expect that third-party manufacturers to help optimize it?
Does anyone know what is happening with eevee development at the moment? I know there is an eevee rewrite branch, but there have been no commits by Clement in a while. Was hoping that this was slated for 3.0 along with CyclesX.
Think it’s important to separate vulkan for the UI and vulkan as a cycles backend these are two rather seperate things, afaik there are no plans to the latter.
hip is just a translation tool to translate cuda into rocm, as far as I know currently does not support windows and rdna architecture graphics card, but the article mentions windows then rocm must be on windows including rdna architecture, or amd did something new, can support windows I think mac is just a matter of time, in addition vulkan theoretically can be used as a universal computing api, but there is really no reference case, the implementation will be very difficult.
hip is just a translation tool to translate cuda into rocm, as far as I know currently does not support windows and rdna architecture graphics card, but the article mentions windows then rocm must be on windows including rdna architecture, or amd did something new, can support windows I think mac is just a matter of time, in addition vulkan theoretically can be used as a universal computing api, but there is really no reference case, the implementation will be very difficult.
).