Isuru Fernando
Isuru Fernando
It is a one-dimensional list, not a list of lists.
I approve
> FWIW, multibuild is backtracking and will also create thin wheels. Not true. multibuild has always created thin wheels and did wheel fusion.
Kernel copying takes the longest time. I'm not sure why copying takes that amount of time.
I don't even understand why `LoopKernel.all_inames()` doesn't return `frozenset(self.inames.keys())`.
@inducer, what are your thoughts on memoizing `get_var_dict` and `get_var_names` in islpy?
> This seems indicating the most of the time is spent on knl.set_args and cl.enqueue_nd_range_kernel. Any idea how to speed up that? Not really. If you do the following, you'll...
Reported the issue in Oclgrind at https://github.com/jrprice/Oclgrind/issues/180#issuecomment-578326894 Can you do `conda install clang=9.0.0 -n nengoenv` until this bug is fixed?
This seems to be an issue with the opencl implementation in 10.14. https://github.com/magnumripper/JohnTheRipper/issues/3434. I can reproduce on a fresh VM on travis-ci OSX
Here's the pyinstrument profile, ``` 136.566 loopy_reproduce.py:1 ├─ 73.198 generate_code_v2 loopy/codegen/__init__.py:404 │ ├─ 32.329 preprocess_kernel loopy/preprocess.py:2030 │ │ ├─ 26.890 wrapper loopy/transform/iname.py:1218 │ │ │ └─ 25.818 realize_reduction loopy/preprocess.py:881 │...