Add norms module #9
Conversation
joewallwork
moved this from In Progress
to Ready for review
in Mesh Adaptation development board
| if isinstance(v, fd.Cofunction): | ||
| v = cofunction2function(v) | ||
| condition = condition or fd.Constant(1.0) | ||
| norm_codes = {"l1": 0, "l2": 2, "linf": 3} |
There was a problem hiding this comment.
| norm_codes = {"l1": 0, "l2": 2, "linf": 3} | |
| norm_codes = { | |
| "l1": PETSc.NormType.NORM_1, | |
| "l2": PETSc.NormType.NORM_2, | |
| "linf": PETSc.NormType.NORM_INFINITY | |
| } |
| v.interpolate(condition * v) | ||
| with v.dat.vec_ro as vv: | ||
| if norm_type == "Linf": | ||
| return vv.max()[1] |
There was a problem hiding this comment.
Does this need an abs()? I presume "Linf" is in fact the same as "linf" ?
| integrand = { | ||
| "h1": lambda w: ufl.inner(w, w) + ufl.inner(ufl.grad(w), ufl.grad(w)), | ||
| "hdiv": lambda w: ufl.inner(w, w) + ufl.div(w) * ufl.div(w), | ||
| "hcurl": lambda w: ufl.inner(w, w) | ||
| + ufl.inner(ufl.curl(w), ufl.curl(w)), | ||
| }[norm_type.lower()](v) |
There was a problem hiding this comment.
I'm probably missing something, but what's the point of the lambda's? If you're worried about the cost of ufl symbolic assembly - I would just change it to an if block (or match case)
| if boundary: | ||
| not_impl_err = "lp errors on the boundary not yet implemented." | ||
| raise NotImplementedError(not_impl_err) | ||
| v.interpolate(condition * v) |
There was a problem hiding this comment.
Hm, this clobbers the input Function v - I'm not sure I like that...
| # Case 1: point-wise norms | ||
| if norm_type[0] == "l": | ||
| v = u | ||
| v -= uh |
There was a problem hiding this comment.
If I'm not mistaken -= for Functions is overloaded and basically does v.assign(v -uh) which again means that the values of the input u are overwritten as v is the same as u at this point. So then if you call this errornorm twice with the same u and uh, you get the wrong thing on the 2nd call.
| :returns: the error norm value | ||
| :rtype: :class:`float` | ||
|
|
||
| Any other keyword arguments are passed to :func:`fd.norms.errornorm`. |
There was a problem hiding this comment.
That's not true it seems
Closes #3.
This PR moves over the
normanderrornormfunctions from Animate, which overload the Firedrake equivalents to account for vectorl1,l2, andlinfnorms, as well as integral norms.