Overflows when using exponentials

[Bart Bogaerts]

see http://seldon.cs.kuleuven.be/idp/?src=55d607f6725aa0680050

I don't know yet if this is solver or grounder related.

[Bart Bogaerts]

Related issue (failing on the same file but without CP support) is #816

[Bart Bogaerts]

This issue occurs because of overflow problems..

Printgrounding reveals

tseitin_504 <=> var_81 = 640942080.
tseitin_504.
tseitin_505 <=> var_82 = 937459712.
tseitin_505.
tseitin_506 <=> var_83 = -2049400832.
tseitin_506.
tseitin_507 <=> var_84 = 1733216256.
tseitin_507.
tseitin_508 <=> var_85 = 1620131072.
tseitin_508.
tseitin_509 <=> var_86 = 2025163840.

[Bart Bogaerts]

Issue #816 was marked as a duplicate of this issue.

[Bart Bogaerts]

This issue can be resolved as follows:

• In src/structure/NumericOperations.cpp take overlows into account. Carefully program this.

• Make sure that other parts of the code all use these numericoperations. For example in src/Structure/StructureComponents, there still are manual calls to pow, manual sums, ....

[Joachim Jansen]

Idea: re-use ArithmeticUtils? These use the "safeint" library that does these overflow checks for us

Or is there a reason not to reuse these?

Edit: is this only related to ints? Also: safeint3 does not support exponentials, but we could loop around this with multiplication.

[Bart Bogaerts]

als dat al bestaat, goed idee