Conversation
teqdruid
left a comment
teqdruid
left a comment
There was a problem hiding this comment.
I've kicked off a pycde test run. It more completely tests the ESI bindings. I should really have more complete tests in our integration tests.
|
The pycde tests failed. It seems that something has changed regarding python type conversions... |
|
Was able to repro. This is funny as the changes to support.py (which is used by pycde too) is exactly towards this. The checking should be on the underlying TypeID which I'd not expect to change in PyCDE. I'll instrument the isa methods to check. |
|
I think I see it: seq and esi both don't have the TypeID getters, that's missing. |
|
Sorry about the delay. I pulled and pushed and kicked a build off: https://github.com/llvm/circt/actions/runs/22640967493 |
|
Thanks, looks green in that one (unrelated, I could get cde working locally with the exclusion of the simulation, some interplay with what gRPC needed cmake side with what I have, so thanks for kicking it off as I couldn't check). |
|
Updated PyCDE run https://github.com/llvm/circt/actions/runs/23000617785 |
2 participants
The mlir_*_subclass APIs are soft-deprecated and will be removed in the future.
Exposed more C API methods for the isa methods in HW.
Updated hw test to
[IntegerType(i2), IntegerType(i32), InOutType(!hw.inout<i32>)](Type -> InOutType - I am not sure why this change, but figured it is accurate still, but can dig in if not).Attempted to do as mechanically as possible. But still is going to be a bit of a pain to review.