QParallel. Support for explicit parallelism in quantum programs similar to OpenMP about qsharp-language HOT 4 OPEN

Above proposal is a joint work by @msoeken, @alexva, @thomashaener.
@sam-jaques, @fvirdia, @cryptosidh, this proposal might be helpful to mitigate the issues with quantum parallelism you have encountered while working on quantum cryptography resource estimates in Q#.

from qsharp-language.

Hi @vadym-kl. Thanks for filing this. This suggestion contains a couple of different requests; would you mind splitting this into individual issues that can be processed separately?

from qsharp-language.

@bettinaheim I beleive there is a conceptual discussion about Q parallel needed first and this issue could be a good place to have it. Once there is a common conceptual ground, perhaps one can create issues for 'parallel for', 'parallel sections', 'section' and 'fanout' discussing further details and referring to this issues for common high level picture.

from qsharp-language.

This would be great, generally. It seems like most of the issues occur in loops of various sorts, so circuit designers probably know what to tell the compiler for each loop (whether to be parallel or not). Would each "Section" use the width- or depth-minimizing compiler? It seems like the width-minimizing compiler would be the better choice, leaving the programmer to manually create the parallelism, but maybe it could be given as a parameter?

A more ambitious feature would be to allow access to the current number of allocated qubits, so the compilation/running of a circuit can dynamically decide whether it should be parallel or not. Having the change suggested in this issue would probably make this extra feature easier to implement and use at some future date.

from qsharp-language.