Allow switchable data types for Numpy arrays
Allow to choose between double precision floats (implemented) and single precision. Single precision could speed up CUDA runs significantly without loosing accuracy for the results.
All Numpy types (and platform independent aliases): https://docs.scipy.org/doc/numpy/user/basics.types.html?highlight=data%20type
Requires to check all return values of Numpy functions for the correct type.
List of all explicit types in code:
core |
inerpolate |
mv |
data_io |
---|---|---|---|
Simulation.__init__ |
ChargePlane.__init__ |
initialize_mv |
read |
- np.int64
|
- np.float64
|
- np.float64
|
- np.int64
|
- np.float64
|
- np.complex128
|
- np.float64
|
|
- np.real => returns? |
read_iter |
||
- np.int64
|
|||
- np.float64
|
Watch changes of #1 for ChargePlanes
.
data_io
also needs attention for (de)serialization and (loading)storing Numpy data arrays in files.
test_consistency
needs attention for (de)serialization and (loading)storing Numpy data arrays in files.
Edited by Kayran Schmidt