the tricks with offsets is needed only for Apple as other platforms do have un-padded float3, maybe that simplifies things? I've rebased the changes on top of master, not started to implement a…
I mean that we could derive offsets/sizes numerically instead of relying on symbols (e.g. memcpy(current_offset, ...); ... ; current_offset += 1; instead of `memcpy(&integrator_state_gpu_.parent_…
I had a quick pass at making this a runtime check, but the changeset started getting quite big. A couple of approaches I started on:
- Define two versions of
IntegratorStateGPU, one for the…
1ee5010cfd
__INTEGRATOR_GPU_PACKED_STATE__ is only enabled on arm64 builds (more explicit #if check)
I'm fine with this problem being solved in either PR, just needs to be done before oneAPI support can land.
I'd like to solve this as a follow on. I couldn't find a concise way to make it…
Out of curiosity I was trying to gain performance improvements, but our performance benchmark did not show anything beyond of regular fluctuations. Attached the result just in case.
Hmm. On…
Wish I could tell you why I added that... 🤔
- Changing a scene setting can cause all materials to be recompiled, and compilation times are already an issue.
Compilation times shouldn't be adversely affected. Adaptive compilation is…
Thanks for taking a look at the CUDA performance, @brecht. It's a shame that it didn't have a positive effect, and I don't exactly know how to explain that.
I've pushed a new revision:
- Only…
