Hi everyone,
I'm having a performance issue with an FME workspace that I previously used for tree canopy analysis.
The workspace processes classified LiDAR point clouds along with DEM and DSM rasters. It uses transformers including PointCloudFilter, PointCloudToPointCoercer, HullAccumulator (Alpha 0.5), and RasterMosaicker.
The workflow previously took around 8–12 minutes per tile. However, since upgrading to FME 2026.3.1.1, it takes significantly longer and doesn't complete.
What I've observed:
-
FME memory usage increases to around 50 GB, with overall system memory reaching 98%.
-
CPU usage remains relatively low.
-
The log shows that LiDAR filtering completes, but processing continues for a long time without further progress messages.
-
Eventually, I get this warning:
Failed to free sufficient memory to reach the process usage limit.
I've also tested running the child workspace independently, rather than through WorkspaceRunner, but the problem persists.
Has anyone experienced similar performance or memory issues after upgrading to FME 2026.3? Could this be related to changes in point-cloud processing or the HullAccumulator?
Any suggestions on how to identify the cause would be appreciated.
Thanks!


