Closed SForeKeeper closed 1 year ago
@SForeKeeper Can you please provide steps to run this PR?
It's still unknown that how should I use Buddy-mlir, it should be a subproject in my opinon, but currently no method is given using it.
I plan to use buddy-mlir domain libraries just like opencv style, which will be ready when I finish redesigning the buddy-mlir structure.
How the python code should be run.
This is an open topic. Should we just use the python script or wrap them as a tool? or other better ideas?
Another thing we need to determine is what format tool we need to use to format the python code. Should we use a tool like black, or can we configure clang-format to make it able to format python code?
@SForeKeeper Thanks for this PR!
When I run the validation Python file, I encountered the following error:
// Navigate to buddy-benchmark/utils/validation/Python
$ python main.py
... ...
OSError: cannot load library '../../../build/utils/validation/AudioProcessing/libCWrapper':
../../../build/utils/validation/AudioProcessing/libCWrapper: cannot open shared object file: No such file or directory.
Additionally, ctypes.util.find_library() did not manage to locate a library called
'../../../build/utils/validation/AudioProcessing/libCWrapper'
And when I run the DAP benchmark in the build/bin
dir, I encountered the following error:
terminate called after throwing an instance of 'std::bad_alloc'
what(): std::bad_alloc
Aborted (core dumped)
Our upstream version can work well now, did I miss something important to run this PR?
Validation using scipy and numpy for buddy-fir and future works. Easily, the framework could be extended by using current code as examples.
There are some problems to be discussed:
c-wrapper
should be built to generate a dynamic library. The python code would also need the path of the dynamic library, since build path will not always bebuild
, it needs cmake information to confirm location.