add gt::transform or gt::vectorize?

wdmapp / gtensor

GTensor is a multi-dimensional array C++14 header-only library for hybrid GPU development.

BSD 3-Clause "New" or "Revised" License

35 stars 9 forks source link

add gt::transform or gt::vectorize? #111

Open bd4 opened 3 years ago

bd4 commented 3 years ago

This can be implemented via thrust::transform for CUDA/HIP and via std::transform for Intel SYCL

germasch commented 3 years ago

I guess I'm not necessarily a great fan of using transform as the API for this. It's usually based on iterators, a bit awkward to use, and it's not lazy. My suggestion would be to implement gt::vectorize (see numpy/xtensor) instead.