... | ... | @@ -56,7 +56,7 @@ For hoppings given by a function `func` we store: |
|
|
|
|
|
(fam_to, fam_from, func, [i1, i2, ..., iN], [j1, j2, ..., jN])
|
|
|
|
|
|
As for the on-site value functions, `func` can be evaluated just once if vectorised, or
|
|
|
As for the on-site value functions, `func` can be evaluated just once if vectorized, or
|
|
|
called for each pair `i<n>, j<n>` if not.
|
|
|
|
|
|
---------------------
|
... | ... | @@ -64,7 +64,7 @@ called for each pair `i<n>, j<n>` if not. |
|
|
[1]: Should the indices be the *global* site indices, or just the offset from the start of the sites
|
|
|
from the current site family?
|
|
|
|
|
|
[2]: The functions may expect a bona-fine `SiteArray` in the case of a finalized `Builder`,
|
|
|
[2]: The functions may expect a bona-fide `SiteArray` in the case of a finalized `Builder`,
|
|
|
can we efficiently construct such a thing? Seems a bit of a waste to construct a new
|
|
|
array and then immediately discard it. If the sites were ordered by their value functions
|
|
|
then we could just use a slice.
|
... | ... | |