Difference between revisions of "Mesh/Mesh Tests"

From Second Life Wiki
Jump to navigation Jump to search
Line 34: Line 34:
* The "pure physics cost" must be < 32.0 before rounding to make the linkset physical.
* The "pure physics cost" must be < 32.0 before rounding to make the linkset physical.
* Test LSL functions that interact with physics and size and textures.
* Test LSL functions that interact with physics and size and textures.
* Localization


[[Category:Test Plans]]
[[Category:Test Plans]]
[[Category:Mesh]]
[[Category:Mesh]]

Revision as of 10:22, 18 October 2010

Mesh tests required:

  • The walk-throughs function as a Mesh smoke test:
  • Mesh_Import_test Everything about the Mesh upload window.
  • Mesh decomposition tool needs more detail in Mesh_Import_test
  • Mesh_Prim_Equivalence_test How many prims a mesh counts as in the parcel limit and simulator limit.
  • Mesh_Accounting_Test is a starting point for several tests. Break the tests out and expanded on them.
  • uploading a rigged object that contains joint offsets
  • bone offsets
  • Mesh viewer performance
  • Mesh sim performance
  • Prim size change (prim size increased to 64m)
    • (encroachment checking)
    • (allow manual return of encroaching objects)
    • (edit tools/camera when working with larger prims)
    • Viewer Performance testing for larger prims
    • Server Performance testing for larger prims
  • Load testing. assets<->sim, sim data<->client, client rendering, etc
  • Viewer caching of mesh objects
  • Physics shape type (with un/linking, de/rez, duplication, etc.)
  • Target overall scene triangle count at 128m, medium detail (currently 125K/tris)
  • Target viewport streaming cost at 128m, medium detail (currently 2MB)
    • "I got this number by flying around some prim heavy builds (the city north of Ahern) with all rendering disabled but prims, view distance set to 128m, and prim detail set to mid, and watching the peak triangle count per frame in the statistic floater. The average was around 75k/triangles per frame with a peak of 150k, so for meshes a good target would be 125k. On MeshHQ 3, after disabling all rendering but prims (meshes fall under the same category as prims in the render pipeline), the peak triangle count per frame was close to 300k/triangles. On that build, there was no LoD enforcement and a ratio of 8KB/prim for mesh costs, so I'm hoping with a ratio of 4KB/prim and strict LoD enforcement, recreating that build within the new criteria will result in under 125k/triangles per frame."
  • Test swapping meshes onto mesh objects.
    • high prim equiv. mesh onto a physical object
    • drag a boneless mesh onto a boned object on ground (and also attachment)
  • A linkset is not allowed to cost less than 1/2 the number of prims. That's to account for basic work the sim has to do in handling the LLTasks
  • PurePhysicsCost will be the sum of the physics resource costs of the prims before rounding and before accounting for that 1/2 num prims limit on linksets. PurePhysicsCost will determine if you can make a linkset dynamic (e.g., a vehicle)
  • setting the root prim of a prim object to Convex hull forces the object to use the new accounting system. Setting a child prim to None does the same.
  • Test the newly exposed physics options Friction, Density, Restitution
  • The "pure physics cost" must be < 32.0 before rounding to make the linkset physical.
  • Test LSL functions that interact with physics and size and textures.
  • Localization