cudaMemsetAsync?

[staticimport]

Does this not work? I need to zero out my output outside my main kernel since the kernel is multiple thread blocks (unless I use cooperative kernel sync, which is also giving me issues). If I create a trivial kernel to zero it out and launch it before my main kernel on same stream, all is fine. But if I cudaMemsetAsync using that same stream I get from ffi call, it runs but my output is wrong.

[staticimport]

False alarm, works just fine when I get the bytes right...