Oleksandr "Alex" Zinenko

Thanks for the report! > Clay "parser.h" no such file or directory It sounds like a problem in Clay itself, does it build on its own? > C++14 needs to...

Because Scop already owns the description of in/out tensors (through Halide in/out parameters), it makes sense for it to also own the stride information. It is more consistent than passing...

This also needs support in polyhedral dependence analysis, putting a block for now

I pushed the polyhedral support for scattering to this branch directly. Requesting @nicolasvasilache to review my code.

> Given that the tests already passed without the polyhedral support, should there be new tests that exercise it? Good call! I added polyhedral tests and by doing so found...

Reviving this. I made Sema look inside the subscript expressions on the LHS so it no longer complains at that level. In particular, it accepts things like `O(i + 2)...

> My inclination would be to make O(A(i)) = B(i) work if A(i) has known bounds (e.g. it's a uint8). And infer 0:256 as bounds for `O`? What about `O(i...

TC was never intended as an IR so there's no IRBuilder. At some point in the pipeline, there is a Halide-based representation that should be possible to construct programmatically. Look...

I think that allowing inference to read _values_ of the inputs and all the related changes to the execution model is an overcompilcated solution to a rather simple problem, and...

Anything will confuse users :) The mere fact of us having this discussion is evidence that the choices are not obvious for everybody... ``` def example(float(N) input, int32 K, int32(2)...