See #913 (comment) for some relevant discussion on the potential shortcomings of the current method and a proposed alternative.
We will encounter this once we add more cost keys, especially for costs where the computation of the cost would depend upon the structure of the composite bloq obtained via bloq.decompose_bloq() and not just on the counts of unique callees that can be obtained via bloq.call_graph().
An example of such a cost is circuit depth. We can close this issue once we add such a cost to Qualtran and show that it scales for large instances of bloqs people often care about (eg: 5000 bit Adder which currently takes ~2 min to decompose into a composite bloq)
See #913 (comment) for some relevant discussion on the potential shortcomings of the current method and a proposed alternative.
We will encounter this once we add more cost keys, especially for costs where the computation of the cost would depend upon the structure of the composite bloq obtained via
bloq.decompose_bloq()and not just on the counts of unique callees that can be obtained viabloq.call_graph().An example of such a cost is circuit
depth. We can close this issue once we add such a cost to Qualtran and show that it scales for large instances of bloqs people often care about (eg: 5000 bit Adder which currently takes ~2 min to decompose into a composite bloq)