TUFLOW Message 3201: Difference between revisions

Content deleted Content added
No edit summary
No edit summary
Line 5:
 
|type=[[ERROR]]
|message_desc=<font color="blue"><tt>Quadtree BC Parallel Inertia Approach </tt></font><font color="red"><tt>== </tt></font><font color="black"><tt>Method B</tt></font> extracts velocities from faces parallel to the model boundary (“QT”, “HT” and “HX” types) to calculate the transverse momentum, or parallel inertia at the boundary. For example, to calculate the parallel inertia at the face highlighted by orange circle in the image below, the transverse velocity from the non-BC-face below is neededrequired. However, if the boundary cells are distributed across multiple CPU or GPU devices (illustrated by the red dot line), itthe cansolver requiremay need to extract velocity information from a face within another device (see the face highlighted by the purple circles). This could causecauses memory access violation during the simulation, thus the ERROR message is generated to stop the simulation.
 
[[File:Proxy face error.PNG|600px|caption]]
 
|suggestions=Revise quadtree refinement polygon to ensure the cell size does not change immediately inside the model boundary. The image below shows a good example, where 2 rows of cells have the same sizesizes and thiswhich ensures the face the boundary faces selectedreferred by the <font color="blue"><tt>Quadtree BC Parallel Inertia Approach </tt></font><font color="red"><tt>== </tt></font><font color="black"><tt>Method B</tt></font> are always within the same CPU/GPU device.
 
[[File:Proxy face error fix.PNG|600px|caption]]