Why bmi_port.c is a C code? about behavioral-model HOT 2 OPEN

This code is written in C for legacy reasons. It was "copied over" years ago from the previous version of the behavioral model: https://github.com/p4lang/p4factory/tree/master/modules/BMI. There was no strong reason to rewrite this code at the time (although I think it may have been marginally improved over time), as it was working and this is a part of the code that is both essential and hard to test in an automatic way.

That being said, I think rewriting this code in C++ and taking a more modern approach (I would say the select-based approach is probably a bit obsolete now) is a great idea. While I will not work on this, I would be happy to review such a change.
I would say that if anyone wants to work on this, please do not start any implementation work until first presenting your proposed approach here and getting some sort of consensus. I would be fine with something that is based on Boost Asio.

Before approving such a change, and because we don't have automated tests for the I/O code, I would expect the following:

• a performance comparison of the current approach and the new proposed approach using this benchmark: https://github.com/p4lang/behavioral-model/blob/main/docs/performance.md#running-the-benchmark. The new approach should yield at least the same throughput.
• a successful test run of the p4c PTF test suite using the modified bmv2 + veths for packet I/O. Assuming there is such a thing... cc @fruffy

from behavioral-model.

a successful test run of the p4c PTF test suite using the modified bmv2 + veths for packet I/O. Assuming there is such a thing... cc @fruffy

It depends... we do not have many static PTF tests available in the suite, but P4Testgen generates PTF tests for all the BMv2 programs with each CI run. That can be used for testing at least.

from behavioral-model.