google/heir

In #1945 we split the process of lowering polynomial.eval into two steps: constructing an ArithemticDag for the lowered eval and materializing the IR in the pass by running a visitor over the DAG. The first step is good because you can unit test the lowered arithmetic DAG outside…

A continuation of #1376 We added a ct-ct matmul kernel, with modifications from the Tricycle paper. The same paper has a more optimized kernel for ct-pt matmul (6.1), which should go alongside (or replace?) our existing ct-pt matmul kernel that just "stacks" the Halevi-Shoup…