Closed stylewarning closed 4 years ago
26q hadamard benchmark goes from 7s (multi-threaded, interpreted) to 4s (single-threaded, interpreted). Compiled mode still beats both at 3.6s.
pan de muerto 🤤sam's not super into it, but i like the vaguely fruity flavor a lot
pan de muerto 🤤sam's not super into it, but i like the vaguely fruity flavor a lot
🍞 💀 I like the hint-of-orange as well, but find it too sugary. I usually dust the sugar off and it's just right 👌
unfortunately, i have a sweet tooth, so i don't often object to extra sugar. old age (rather: dental work) has given me the wisdom not to mind if it isn't sweet to start with, though.
The point of this PR is to allow smaller qubit simulations to be faster. The QVM is pretty fast for large numbers of qubits, but definitely doesn't do the best job for small numbers, which is an important case to many people.
This PR adds macros for definiting qubit application operators as "stupid" loops that kick in when there aren't enough qubits for the full-blown parallelized code.
NOTE: This code is still buggy, hence the draft PR.