It has been our aim to simulate two problems and compare the time taken for simulating them by varying the following parameters:
- The data type for computing the FFT, which is necessary to calculate the Demagnetization energy, was changed form 'long double' to 'double' to 'float'
- These simulations were then carried out on machines with different configurations viz.
- Intel(R) Pentium(R) D, 2.80 GHz, 1 GB RAM
- Intel(R) Core(TM) 2 Duo, 2.80 GHz, 2 GB RAM
- Intel (R) Core(TM) i3, 2.93 GHz, 4 GB RAM
- Intel (R) Core(TM) i5, 3.2 GHz, 2 GB RAM
The problems simulated are given by the following links:
Please refer to the MIF Format page wiki http://www.ee.iitm.ac.in/redmine/wiki/dynamag/MIFformat for details of the problem specification.
The initial magnetization used in the latter problem was obtained by simulating the first problem with the FFT routine using 'long double' data type and then allowing the last step(at 0 torque) to relax at a torque of 1E-4. THe OMF file generated for this step was then converted to an OVF file(initmag2.ovf) which essentially contained the magnetization values at all points of the geometry.
The change, in terms of data type modification, to the FFT routine may be found in the following file:
The data obtained, about the time taken for simulating all these, is given in the following spreadsheet:
The time comparison can also be shown with the following bar charts:
Please note that in both the cases the Function FFT::BaseDecFreqForward(MyComplex*) has been considered for comparing the time taken