nopeslide
nopeslide
currently we are using `malloc` & `printf` all over the place, ignoring platform specific implementation. I would like to replace these calls so users can provide their own allocators and...
Got the following proposal regarding threading: - Sets refer to the typeless operator instead of resolver (see issue #40 ) - The typeless operator sets everything up for the actual...
Just thinking out loud regarding our build system. onnx is huge and I think our build system can not scale good enough, so we need to address a few things:...
When integrating connxr in C++ we have to disable the C++ function mangling, so our c functions can be linked correctly. We would need to put all our prototypes or...
I did not remove the original `trace.{c,h}` files and left the inference and testing files as they are. the "old" trace.{c,h} must be integrated into tracing.h or at least all...
One thought I had was to maybe replace the type specific traces (`TRACE_TENSOR` and so on) by inline functions so the normal c typechecker can act upon these, but this...
I found a few memory errors with valgrind in the base code. To prevent such things in the future I propose running all tests with valgrind
- each commit should be tested and pass - a PR should only be merged if all commits pass - rewrite CI scripts to accommodate this behaviour - if possible...
It would be nice to autogenerate the [operator overview](https://github.com/alrevuelta/cONNXr/blob/master/doc/04_OperatorStatus.md)