Closed ShangkunLi closed 2 weeks ago
If you checkout to the latest version of VectorCGRA, you need pymtl3 version of pip install -U git+https://github.com/tancheng/pymtl3.1@yo-struct-list-fix
as shown in the README of VectorCGRA. If you didn't checkout to any version and just use the VectorCGRA within the docker, you don't need to re-install anything IIRC.
Right, that's the issue we wanna fix. The tests are only unit tests and aiming at verification of the existing functionality. We don't have a robust way to guarantee the correctness of the control signals generated from the mapper, as either of the two tools has been developed on its own for a while. Based on our and others experience, it would take a few days to tune the generated control signals to make the RTL simulation run as expected. We are working on DFG-simulator and mapping-verification these days. It would super helpful if you can help contribute on this :-)
Sure, I would be happy to contribute on this functionality extension. Please let me know if you have anything that I can help.
Hi Cheng,
I just walked through your CGRA-Flow using the GUI provided in the docker image. And I have some questions:
I tried to deploy VectorCGRA locally. However, there seems to be some problems with the PyMTL3 package. I tried to install pymtl3 using
pip install pymtl3
orpip install git+https://github.com/tancheng/pymtl3.git
. But in both ways, the classes/functions inpymtl3/stdlib/test_utils
have some little difference with classes/functions inpymtl3/stdlib/test
from your docker image.In terms of the Run tests function in your CGRA-Flow GUI, it firstly runs some tests on a determined CGRA coupled with a specific mapping json file, for the user defined CGRA, it only generates the parameterized CGRA. So how to determine whether the mapping result generated by CGRA-Mapper is correct?
Thanks for your time and help!