Léo Ducas
Léo Ducas
Darn. Am I right to understand that you are running a 32 bit system ? I've only tested 64 bits systems... Maybe there is another int with unspecified size somewhere...
Did you solved this in the end ?
Note: This likely impact the result for the SCA attack on Frodo.
Tentatively fixed in PR #10.
It appears that they diverge on the computation of `log(volB)`. The correct value is `1400 * ln(2^24) = 23289.7` matching `dbdd_predict`. I suspect a numerical stability issue for determinant computation...
`dbdd_predict_diag` should be more than enough for that purpose indeed. The following option should also allow to match the new estimator "CN11" simulator (usvp attack): `estimate_attack(probabilistic=True)` Though you might have...
@joerowell What is the status here ?
No, SVol is the (logarithm of) determinant of the ellipsoid that defines the error distribution.
Unfortunately 128 bits of precision might still be too low for lattices of dimension 600.
"JS mumble, template mumble mumble." I guess I should leave this to the pros... But I can surely help crawling through all files to detect and mark all texts needing...