Jeremy Kun

Just a quick peek: looks like they also do tracing to get per-node min/max stats for poly eval: https://github.com/baahl-nyu/orion/blob/f0581052b28d02a00299cce742949930b3260aa8/orion/core/tracer.py#L149. This may help inform https://github.com/google/heir/issues/1700

Currently we are working on implementing the convolution kernel from Orion.

I think this is the work that @ZenithalHourlyRate and the authors were discussing over email.

There is another Google engineer (@dvadym ) who is doing some of this. See for example https://github.com/google/heir/pull/1772 and https://github.com/google/heir/pull/1770 I believe he is going to work on (a simplified version...

Working through some of the nitty gritty details of this overall layout system, recording some notes here for myself: I was looking at the "materialize the layout" step of the...

I think this would be best phrased as "tensor semantics" vs "ciphertext semantics" (maybe semantics is the wrong word though) and draw a line showing where we lower from one...

IR just before optimize-relinearization: ```mlir module attributes {scheme.bfv} { func.func @two_mul(%arg0: !secret.secret {mgmt.mgmt = #mgmt.mgmt}, %arg1: !secret.secret {mgmt.mgmt = #mgmt.mgmt}) -> (!secret.secret {mgmt.mgmt = #mgmt.mgmt}) { %0 = secret.generic(%arg0: !secret.secret...

I can only imagine this is something we'd want in the case that the inputs to the operand _really_ want to be packed row-major for compatibility with other more expensive...

> I believe I finished making all of the requested changes but please let me know if I missed anything. Due to our strange project setup in re Google, you...

In that test, maybe we can avoid this by just removing one of the operations so it's 15 instead of 16?