python control file/function (instead of Makefile)

What would be your opinion to have python control file for the cocotb instead of makefile.

Initially one can start with simple a wrapper for Makefile. A module that will generate a local Makefile based on python code and execute it (subprocess). This should be backward compatible. Later can recede makefiles to just compile liberalises and run simulator directly form python.

Could look something like this at the end of test file:

from cocotb.runcontrol import Run

def test_run():
    Run(
        sources = ['../hdl/dff.v'],
        top_level = 'dff',
        incdir = ['../hdl'],
        defines=['TEST=1'],
    )

if __name__ == "__main__":
    test_run()

This could automatic discoverable by pytest. If we get #517 we could run in parallel.

It needs bit of thinking and trying but would like to know your opinion. Somehow related to #437, #508, #667 In case of positive feedback can make a working demo.

[imphil]

I’ve done something going into this direction with cocotb-testrunner (https://github.com/imphil/cocotb-testrunner). It works like this:

• Add a yaml “manifest” file describing your design to a test directory. (Example here: https://github.com/opensocdebug/osd-hw/blob/osd-next/modules/stm/test/test_stm.manifest.yaml)
• Run cocotb-testrunner. It finds all manifest files, writes specialized Makefiles for it, runs the simulation, and in the end collects its results.

We use this testrunner extensively internally, and in Open SoC Debug and in OpTiMSoC. Unfortunately, none of the test runs are publicly available in CI since we mostly use Synopsys VCS.

[themperek]

I think both works, it’s more a matter of taste. In the end I don’t think it highly matters.

Actually one should be able to do something like this:

from cocotb.runcontrol import Run

def test_run(): 
    Run(**yaml.load('manifest.yaml'))

if __name__ == "__main__":
    test_run()