hpkfft.com

Hmmm, macos C++ compiler doesn't like `static_assert(a.is_ro);` in the tests. Is there a mac-specific idiom that should be used? If not, I can simply `&&` this into the return value...

Thanks, Roman. I'm now accessing the class static data member `is_ro` through the type of `a` ``` [](nb::ndarray a) { static_assert(decltype(a)::is_ro); ``` and, if `a` is a reference ``` [](const...

Certainly. (I've had project managers that insisted on simple, repetitive, fall-through code in tests.) Please advise (at your convenience) on naming preference, or if the boolean should be removed. I...

Since template parameter packs are recursively examined right to left, given `nb::ndarray a`, the view `a.view()` is still 2x2. This patch changes this so that the shape for the view...

As a brainstorming exercise, I experimented some in PR #498

I re-based this upon the current HEAD (so as to manually resolve merge conflicts with recent commits) and force-pushed.

Would you be kind enough to add documentation (the source for which is found in the `docs` subdirectory) to https://nanobind.readthedocs.io/en/latest/api_cmake.html and also to https://nanobind.readthedocs.io/en/latest/typing.html#command-line-interface The latter would be especially helpful...

Excellent point! My proposal is for this to apply just to real inputs. Given industry adoption of IEEE Std 754-2019, correct rounding **is** actually the case for existing libraries/devices. Sometimes...

I believe the existing accuracy requirements for element-wise arithmetic operations are only intended to apply to real inputs. For example, the usual 6 flop implementation of complex multiplication ``` (a...

For those interested in the accuracy of complex operators/functions, my colleagues and I have done a comparison with different compilers/libraries: [complex.pdf](https://inria.hal.science/hal-04714173)