Conversation
The codimension of the unstable locus does not depend on the denominator, there is no need for this argument
improves euler_form performance
|
This weakens the requirement of theta-coprimality to the requirement of theta being general, that is, semistablity and stability being equivalent notions. This is currently checked via
|
This was only useful using `SVectors{T, Int}`
src/Chow.jl
Outdated
| ) | ||
| out = R(0) | ||
| for sigma in W | ||
| Oscar.add!(out, out, sign_product(sigma) * permute(f, sigma)) |
There was a problem hiding this comment.
Is this where addmul! could be useful for more in-place computations?
There was a problem hiding this comment.
Needs to be tested - it is not so immediate because sign_product(sigma) is an Int, maybe the implementation of addmul! does not like that so much. Also, the sign could be integrated in the permute(f, sigma) function, since it already knows sigma, potentially saving one extra allocation.
There was a problem hiding this comment.
It's an Int, but is it really simply a sign? Then maybe doing add! or sub! is an even better idea.
There was a problem hiding this comment.
Good point, yes. Yes, it is just the sign of the permutation sigma of the variables...
Speedup of an order of magnitude
only warns if `verbose` though.
Useful if one is only interested in Chow ring computations and the HN type enumeration is heavy on memory
This was introducing a lot of overhead in the computation e.g. of chern classes. Ex: `point_class(M; unsafe=true)` of the 11-points on P1 moduli problem goes from 15 seconds to 2.
This was apparently introducing some coercion overhead
Despite the naming convention it does not perform the subtraction in place.
Reduces allocations by ~1%
This enables large computations, analogously to the caching of `all_hn_types`
instead of theta-coprimality we just require theta to be in general position, i.e., semistability and stability must coincide
2 participants
No description provided.