Steven Johnson
Steven Johnson
Agreed, but availability could be an issue -- pybind11 is widely available on many systems already, and (more importantly to me personally), nanobind isn't available inside Google yet, so we'd...
FYI, I took a stab at doing a basic adaptation, but it looks like there are a number of API changes that aren't well-documented yet (eg buffer->tensor) and other things...
(One part of nanobind that is better than pybind11 is that it appears to have support for dlpack built in, which is something we really want/need, not sure if it...
So how confident are we in this change? Do we have a way to verify that the null-check code is actually being exercised?
Monday review ping to @aankit-ca
If @aankit-ca isn't available for this review, we should either find someone else or just land it as-is.
> I guess if anything in Halide.h uses LLVM types, that might not work well. That would be an error if it is currently the case. In theory, only Halide...
I'm claiming this, and (gradually) working on implementing.
As I look at this, there are some preconditions to making this work: - We take a fast-and-loose approach about implementation public API in our header files, sometimes as explicitly...
I think we have (mostly) resolved this via #5659, though it's still an open issue for Windows. Would anyone like to take this on for MSVC so that we can...