Michael Shoemate

More likely, translate materialize -> cast -> clamp -> impute -> resize -> (mean, variance) Into a beam pipeline. I'm happy to add custom data loading components if beam doesn't...

See https://docs.opendp.org/en/latest/user/measurement-constructors.html#floating-point

On the other side of things, the geometric mechanism has an optional argument, `bounds`. When set, the algorithm should execute in constant-time. Any of the mechanisms that depend on MPFR...

I converted this back to draft, because it doesn't use the row-by-row proof.

For progress, see [325-exponential-mechanism](https://github.com/opendp/opendp/tree/325-exponential-mechanism).

This relation could be made tighter with a coupling argument. See [Definition 16, 17](https://arxiv.org/pdf/2007.05157.pdf). Expected to have c-stability of 1. Should also consider if random sampling is necessary.

I don't have a lot of knowledge about embedded platforms, but aarch64 is listed as a [rustc supported platform](https://doc.rust-lang.org/nightly/rustc/platform-support.html). You can use [this page](https://rust-lang.github.io/rustup/overrides.html#default-toolchain) for info on how to set...

Sorry about the slow response, I've been traveling. Try adding the `--no-default-features` flag to help you get things bootstrapped. This drops the gmp/mpfr dependency, and along with it the continuous...

[Here is an example of the python bindings (which uses the FFI interface).](https://github.com/opendp/opendp/blob/main/docs/source/examples/pums-data-analysis.ipynb) When it comes to using the FFI interface directly from C, I'll point you to a few...

About approach 2, here's an example crate for wiring your own ffi up directly: https://github.com/opendp/opendp/commit/8561d7e57e960eb72fffa9f24e2dbe54bb6084bb