Compared to implementation of the van der Pol problem, use of accelerInt with pyJac is quite simple.

File Geneneration

After installing pyJac (via github or pypi), the required C/CUDA implementation files can be generated e.g. for the included H_2/CO model via:

  python -m pyjac -i examples/pyJac/h2.cti -l c -b examples/pyJac/out/ -ic 1000,1,H2=1,O2=0.5,N2=1.76
  python -m pyjac -i examples/pyJac/h2.cti -l cuda -b examples/pyJac/out/ -ic 1000,1,H2=1,O2=0.5,N2=1.76

Note: the -ic parameter sets the initial conditions, similar to set_same_initial_conditions() in the "vdp van der Pol" example.

Building Solvers

Next the solvers can be built via:

  scons SAME_IC=True LOG_OUTPUT=True LOG_END_ONLY=False t_end=1e-3 t_step=1e-6 mechanism_dir=examples/pyJac/out/ -j 2

The j parameter allows scons to use 2 threads during compilation (vary as necessary)
The mechanism_dir option tells scons to look in the pyJac\out example directory for the required dydt() and eval_jacob() functions.
We log the output to a file (generated by default in accelerInt-root/log/solvername.bin), for plotting.
The time step t_step is set to 10^-6 second, while the end time end_time is set to 10^-4 seconds.
The default error tolerances ATOL/RTOL = 10^-10 and 10^-6 respectively are used.
Additionaly the PRINT=True option tuns on logging to the screen (if so desired)
Alternatively, randomized initial conditions generated by generate_ics.py may be used by copying the resulting ign_data.bin file to the root accelerInt folder, and compiling without the SAME_IC option.

Running the Solvers and Plotting Results

The CPU solvers can be called using:

    ./solver_name [num_threads] [num_odes]

While the GPU solvers are called via:

    ./solver_name [num_odes]

For example, we call:

    ./exprb43-int 1 1
    ./cvodes-int 1 1
    ./radau2a-int-gpu 1

Next, we load the data and plot (plotter.py), resulting in:

Solver output for Constant Pressure H2 Ignition at 1000 K, 1 atm, phi = 1