Drew Hubley

See benchmark code below: ``` #include #include #include #include double mean_milliseconds_from_total(std::chrono::nanoseconds total, size_t num_repeats) { std::chrono::duration total_ms = total; return total_ms.count() / (double)num_repeats; } int main() { size_t num_repeats =...

> Thanks a lot for tackling this! I think the linear / not linear iteration was supposed to be handled in the flat_storage_adaptor but we probably missed some cases. >...

> Thanks a lot for tackling this! I think the linear / not linear iteration was supposed to be handled in the flat_storage_adaptor but we probably missed some cases. >...

I've been wondering how to fix this. The default lambda parser is definitely not compliant with constexpr captures. How did you know that adding would fix this?

Thanks! This will help my code base. @JohanMabille please review! 🙂

This is the source of your issue:  In the case of contiguous containers this implementation works fine. We don't care about strides because adding a scalar to the whole...

Your issue is that there is an implicit cast from `xt::equal(x, x)` to `xt::xarray` add a cast to do the conversion. ``` #define XTENSOR_USE_XSIMD #include #include #include "xtensor/xarray.hpp" #include "xtensor/xtensor_forward.hpp"...

This would be amazing. We currently use a reverse proxy to use a single IP address and HTTPS connection to serve multiple services. I would love to be able to...

@vakokako Have you tried implementing this with `xt::noalias`? Each time you run this you could be reallocating memory for the output buffer.

Hi @emmenlau. I'm aware of the patch. I should have been more clear and asked what version of xtensor was used in the test results. If it's a new version...