Page MenuHome

EEVEE Cryptomatte Processing Performance
Confirmed, NormalPublicTO DO

Description

Current implementation uses a single core. During implementation it was always kept in mind that we should do the processing in a threaded way. It should also be possible to render the next sample during the integration of the previous sample in the accumulation buffer.

Another solution would be to do the sorting on the GPU using compute shaders. D10913: GPU: Compute Pipeline.

Event Timeline

Jeroen Bakker (jbakker) changed the task status from Needs Triage to Confirmed.Nov 10 2020, 11:05 AM
Jeroen Bakker (jbakker) created this task.
Jeroen Bakker (jbakker) changed the subtype of this task from "Design" to "To Do".

hopefully this will improve performance of eevee cryptomattes as current implementation has massive performance penalty up to 500% depending on setup