danny couture 8844de5ffb Add per asset memory usage approximation during async compilation to offer more precise throttling when memory is scarce or we're processing huge assets ...

- FQueuedThreadPoolWrapper will now try to retract any task it can until the max concurrency goal is achieved to respect priority as much as possible
  - Add memory requirement estimate on IQueuedWork to allow the threadpool wrapper to take action when required
  - Add logic in the asset compiler's threadpool to handle per asset memory requirement instead of using a default for all assets
  - Add estimate for static mesh and textures
  - This change should improve CPU usage when processing small assets on memory constrained systems and should prevent OOM when dealing with several excessively large assets

#jira UETOOL-3082
#rb Francis.Hurteau

[CL 16019139 by danny couture in ue5-main branch]

2021-04-15 05:30:38 -04:00

..

Private

Avoid renderdata of HLOD being recomputed during cook package save because of non-deterministic guid

2021-03-20 20:01:33 -04:00

Public

Add per asset memory usage approximation during async compilation to offer more precise throttling when memory is scarce or we're processing huge assets

2021-04-15 05:30:38 -04:00

MeshDescription.Build.cs

…