How to C inline the fipy programme?

[Sandip433]

Hi, I am using fipyTWO TEMPERATURE MODEL.txt (3.4.3) to simulate TWO TEMPERATURE MODEL . But the program is slow. I want to use C inline and --no cache to speed up. I could not make it work. Here is the code I am using.

[Sandip433]

Also I was wondering other ways to speed up python code(like using numba package, cython or codon ) could work with fipy.

[guyer]

The weave module was removed from scipy many years ago, although it still exists. FiPy --inline has known issues (#629) and we do not test it; we should update the documentation to reflect that. Despite what the docs say, I only ever found the speed gains to be modest, at best (factor of 2 or so?); @wd15 may have seen better, but not consistently.

• What does "I could not make it work" mean?
• "slow" compared to what?
• How did you pick your time steps?

[Sandip433]

i) Sir I imported inline command line from fipy.tools but it did not affect the overall time required to perform calculations . ii) When I try to run the program for 100 pico seconds , it takes several hours to complete the task . If I use Labview software( this software I am using to control the machine , also to simulate the experiment theoretically) , the calculations are done in minutes. iii) I tried different time steps to get similar results which I got from the experiment. I tried different cell sizes and grid spacing . Initially I was instructed by my supervisor to use 0.5 fs as time step but I got consistent results from 4 fs and below. This result I was trying to get .

[guyer]

I know nothing about the two temperature model (it is very hard to help when you neither state nor reference what equations you're trying to solve). That said:

• Why are you modeling a depth of 1.0e-4#m when your laser depth will be of the order of $5 w0$ to $10 w0$, or about $10^{-7}~\mathrm{m}$? If I had to guess, LabView is not modeling this entire domain at the same resolution. FiPy does not have mesh adaptivity.
• Your equations are extremely non-linear:
  • You need to sweep to convergence.
  • You should try linear or constant coefficients for $C_e$ and $C_p$ before attempting the full expressions.
  • Likewise, a simpler expression for $S$ is advisable until you get results that make sense.
• Why are you worried about how fast it goes when it doesn't even do what you want?

[Sandip433]

Fipy works well to solve Two Temperature Model TTM1.pdf i) You are write about laser depth . I corrected that part and the program runs well. ii)This figure I got solving Two Temperature Model with co-efficients using Fipy program specified above. Fipy has no issue of inaccuracy or instability for non - linearity. I just want see if there is any other way to run the program faster than current speed.