PETSc: investigate the behaviour of some matrix operations and make it easier to use
To make CardiacMechanics work again, I had to replace some NonZeroPattern::same
by NonZeroPattern::different
, which is deeply unsatisfactory.
Likewise, they are cumbersome to use as if the result is already allocated it may lead to a PETSc error.
I need to understand better PETSc behaviour, and find a way to hide the complexity for the end user.
And of course we should strive to limit NonZeroPattern::different
at minimum (0 case would be the best!)