I’m putting this up for comment now. I still need to dot a few Is and cross a few Ts before it’s ready to go. Background This is following on from a previous discourse topic a prior stan-dev:math pull request stan-dev:math issue #123 std::complex<var> and std::complex<fvar<T>> I believe I h…

[image] ChrisChiasson: rpath does not work on Windows I dont want to derail this topic but not sure what you mean by “not working”? In order to do dynamic linking on Windows TBB needs to be in the users or systems PATH variable. [image] ChrisChiasson: TBB likes to spawn 3 threads just fo…

[image] rok_cesnovar: I dont want to derail this topic but not sure what you mean by “not working”? In order to do dynamic linking on Windows TBB needs to be in the users or systems PATH variable. rpath, if I understood it correctly, is different from the PATH environment variable. I believe …

A finite set of random tests is still measure zero. How about rather than having unit tests, we run a million trials of something random under development? That way, we don’t have random inexplicable failures of unit tests unrelated to the branch being developed.

[image] ChrisChiasson: I have been wondering, why are you calling it CRTP? I called it CRTP-like in the doc. It would be an instance of CRTP if the template type were std::complex rather than T; but it’s used just like the CRTP in terms of casting to the subclass. [image] ChrisChiasson: …

Also, as of now, everything’s broken while I try to sort out the ambiguities in the header tests.

[image] Bob_Carpenter: I may not have been testing that case. Indeed, pow(var, int) is also ambiguous for me with clang++: auto fx = pow(x, 2); ^~~ ./stan/math/rev/core/std_complex.hpp:2599:26: note: candidate function inline std::complex<var> pow(const var& lhs, const std::com…

I’ve pushed fixes so that the header tests now pass and pow(int, var) and pow(var, int) work. The unit tests are still failing to compile in various places due to ambiguities.

[image] Bob_Carpenter: I’ll check out what’s going on. I may not have been testing that case. I can try to fire up your PR and see if it passes all of my tests. I thought this was going to be a lot easier than it is. It would be interesting to see what would happen testing my PR. In many way…

[image] Bob_Carpenter: The main difference is that I never had to take extreme measures to set things to zero. I just defined the constructors in std::complex<var> and std::complex<fvar<T>> to do that. For me, because of the choice to delegate back to the standard library implementation, and …

[image] ChrisChiasson: Also, I may be slow to respond next week due to an industry meeting I have to attend. Apologies in advance. No worries. I’m also traveling and haven’t been doing a lot around this. I’ll be back in the office next week myself.

Fully functional std::complex specializations and overloads

Developers

mitzimorris December 14, 2019, 4:27pm 49

problems described here - best guess - old crap from macports - any of y’all please advise -

Topic		Replies	Views
Help testing release candidate for std::complex Developers complex-math	85	4431	June 5, 2020
Complex number support in Math library autodiff Developers complex-math	53	4126	October 9, 2018
Complex numbers and matrix inversion in Stan Modeling stan-math , complex-math	21	2283	May 1, 2020
Using std::pow inside stan::math namespace Developers	38	2367	June 10, 2020
Am I Leaking? [Eigen autodiff] Developers	12	2025	April 29, 2019

Fully functional std::complex specializations and overloads

Related topics