Vc::InterleavedMemoryWrapper is missing in std::simd

[dimula73]

Usecase

We have pixel data is R1G1B1A1R2G2B2A2... format. To make compositing we need to deinterleave them into R1R2R3..., G1G2G3..., B1B2B3..., A1A2A3... format.

Implementation in Vc and std-simd

In Vc library we could use Vc::InterleavedMemoryWrapper to achieve this goal. std-simd doesn't provide any efficient solution for the task. We could use generator constructor for it, but it doesn't seem to generate what we want:

Old version with Vc::IntterleavedMemoryWrapper:
https://godbolt.org/z/fEQ3s1

Generator syntax with std-simd:
https://godbolt.org/z/pUtMXm

You can see that generator-based version compiler generates more than twice more instructions. That is really suboptimal

PS:
I think this issue may be somewhat related: VcDevel/std-simd#4