quantumlib
Alternative uniform superposition costs
Our current implementation has a cost of roughly 10 log L T gates, the algorithm in the thc paper (bottom of page 31), which is often used as a reference point has a cost like 3 log L Toffolis (12 T) and uses a phase gradient register.
Opening this to track as I will deprecate the chemistry black boxes which captured this cost, but we maybe want to add this (it's a straightforward circuit to build) and we do need to think about how to propagate down the bit precision when using a phase gradient register.
Can you add github commit links to "our current implementation" and " chemistry black boxes which captured this cost" so we can look at the precise bloqs at a later point in time when we want to fix this?
