asraa
asraa
Superseeded by https://github.com/google/heir/pull/1269
I think the ones introducing arith ops (https://github.com/llvm/llvm-project/blob/30f5a3ca150e98d482abc6a4d0e3fe6c12f77695/mlir/lib/Dialect/Polynomial/IR/PolynomialCanonicalization.td#L59-L64, and the corresponding sub one) are also incorrect in the case where there's a modulus or the coefficients might wrap around the...
When I remove the first function of that test, the test is no longer flakey!
I'm guessing there's a secretness lattice issue then - somehow it's probably recalculating the secretness analysis when the other function is present to make the secretness of the static index...
> @asraa Do you have a fix in the works for this, or should I start looking at this? My fix was to move this https://github.com/google/heir/blob/d58ccf8dde341803f69d2eef8fb796a19cba829e/lib/Transforms/ConvertSecretInsertToStaticInsert/ConvertSecretInsertToStaticInsert.cpp#L80 to the end of...
> @asraa I think it's fair to not devote too much time to this right now, I can write up a better doc and walk you through it after I'm...
Getting around to merging this now, thank you for the wait! I'll take care of the rebase internally since it needs a BUILD fix for the dependency.
> Do you think this could also be part of a compliance tests (as an Action we provide or something else) the TRW authors run as pre-submit? The predicate is...
Yes, more relevant to multiple predicate types.
That's hilarious. I didn't realize that there was an EvalLinearTransform API in OpenFHE. It doesn't have a great docstring but from the implementation that looks like a rotate_and_reduce operation. I'm...