Jeremy L Thompson
Jeremy L Thompson
Right.... You highlighted exactly the [wrong] description that I want replaced. Using push like that mutates one array and allows for the mutation of the other. The current solution to...
Note - this branch needs the updates in main for the latest development version of PETSc: #940
You need to force compatible quadrature spaces by setting the quadrature order on coarse levels to match the fine level. See `RatelSetupDMByOrder_FEM()`
Are you using `PetscCall(DMPlexMarkBoundaryFaces(dm, PETSC_DETERMINE, label));`?
Why 1 instead of PETSC_DETERMINE?
Rough code for what I was thinking of. ```C typedef enum { CEED_PRECISION_HALF = 0, CEED_PRECISION_SINGLE = 1, CEED_PRECISION_DOUBLE = 2, CEED_PRECISION_DOUBLE_DOUBLE = 3, } CeedPrecisionType; ... char *data_ptrs[4]; ......
```C // Set two pointers of user arrays to use CeedVectorSetArray(v, CEED_MEM_HOST, CEED_PRECISION_HALF, CEED_USE_POINTER, &h_ptr); CeedVectorSetArray(v, CEED_MEM_HOST, CEED_PRECISION_DOUBLE, CEED_USE_POINTER, &d_ptr); // Copy to half precision // Would contain data from...
Oh, I was just scratching down some ideas. I wouldn't replicate two enums for basically the same thing, I just didn't remember the enum names offhand. I don't think we...
We should probably expect the user to explicitly set the precision of the context data?