Design #778
CRTMill::myAddInfo accept modulus 1 or not?
Description
Should the mem fn CRTMill::myAddInfo accept a residue-modulus pair with modulus equal to 1?
Such a pair actually adds no information at all. Currently, normal mode accepts it, but debugging mode produces an "assertion failure".
Which behaviour is better?
Related issues
History
#1
Updated by John Abbott over 8 years ago
Updated by JAA found the bug when trying to run a CRT loop in CoCoA-5. The initial CRT residue-modulus pair was naturally (0,1); then the very first call to CRT (in CoCoA-5) produced an assertion failure because CRTMill::myAddInfo as called with one input modulus being 1 (and by chance JAA had compiled with debugging active).
It would certainly be convenient to accept modulus 1 (and the residue is completely ignored) so that CRT loops can easily be written.
Adding a residue-modulus pair with modulus=1 adds no information (but is also quite harmless); it would probably be a mistake except as the very first pair.
#2
Updated by John Abbott over 8 years ago
- Status changed from New to In Progress
The possible approaches are:
(A) always give error when an input modulus is 1 (but awkward for starting a CRT loop)
(B) always accept an input modulus of 1 (pointless but harmless; current behaviour)
(C) accept modulus of 1 only if internal modulus is 1 (i.e. just as the first pair)
Approach (C) is tempting, but it does mean that in CoCoA-5 a call like CRT(0,1,A,B) works as expected while CRT(A,B,0,1) would produce an error (with the current impl of CRT...)
Comments? Ideas?
#3
Updated by John Abbott about 8 years ago
- Related to Design #620: Redesign CRTMill added
#4
Updated by John Abbott about 8 years ago
- Target version changed from CoCoALib-0.99540 Feb 2016 to CoCoALib-1.0