How to transform Mako template into c- or cuda-type?

[Kiny Wan]

Dear all, 

        In the publication of PyFR (PyFR: An open source framework for solving advection–diffusion type problems on streaming architectures using the flux reconstruction approach),  the idea of Mako is very interesting that extends a small piece of code into c- or cuda-type. I would like to ask about which python file in PyFR controls the transformation? and how to find the generated CUDA source/OpenMP annotated C source related to corresponding Mako template?

Regards

Kiny

[Freddie Witherden]

The translation is mostly done in the various subclasses of
BaseKernelGenerator in:

<https://github.com/PyFR/PyFR/blob/develop/pyfr/backends/base/generator.py>

Some of the Mako related functions are inside makoutil:

<https://github.com/PyFR/PyFR/blob/develop/pyfr/backends/base/makoutil.py>

If you want to see what the generated code looks like you can put a
print statement around:

<https://github.com/PyFR/PyFR/blob/develop/pyfr/backends/base/kernels.py#L80>

Regards, Freddie.

[Kiny Wan]

Hi Freddie, thank you.