Andreas Klöckner
Andreas Klöckner
> I tried removing the int arg and am waiting for the CI result but I'm doubtful that is the problem. I'm puzzled. Does it compile for you locally?
Pushed to Gitlab for CI, find results via https://gitlab.tiker.net/inducer/pycuda/-/commits/issue_319
(Pushed once again.)
I think this is legitimate functionality to want. `extract_subst` for now (IIRC) only finds exact subtree matches. I agree that matchpy is likely a better approach than trying to roll...
* With what OpenCL implemenation (ICD)? * Could you run with `python -m pytest -v --capture=no`?
Also, is this a regression, i.e. did 2018.1.x work?
This looks like pocl may have an issue, more than pyopencl. Does pocl work with a hello-world-level OpenCL program (in C)?
Could you compile PyOpenCL with tracing and figure out which call fails?
> Can we write some more docs and examples? Sure, I'd be happy to consider a pull request to improve the docs.
Matching the original strides is not a plausible strategy, since the argument array could be non-contiguous or have negative strides, neither of which would go well in a newly-allocated buffer...