Ben Corbett
Ben Corbett
We could also use numpy's built in method https://docs.scipy.org/doc/numpy-1.15.0/reference/generated/numpy.allclose.html which evaluates `|A-B| > ( EPS + RTOL * |B| )`
@rrsettgast Yeah I modified the notation to match the silodiff docs.
I'm just a little skeptical of scaling our tolerances to the largest value in the baseline data set. If all we cared about was the linear solver stuff then we...
One complementary option is we could allow a test to provide different tolerances for different fields. ``` python defaultAtol = 1e-9 defaultRtol = 1e-9 specializations = [(".*/d.*_d.*", (1e-4, 1e-7)] ```
What about using the cosine of the angle between the two vectors to compare them? ``` dotProduct( x, y ) / |x||y| < 1 - tolerence ``` The example above...
The way our current algorithm is we need to know an upper bound at mesh construction time. Currently this is hard coded to 200, but it would be nice if...
Yeah we have a lot of methods that are pascal case still like `GetGroup`. I would say that we should just delete these two methods and similar accessors. Users should...
What's the status? I'm pretty interested in using this.
> We are planning to test on some internal applications prior to merging so that we don't accidentally break them. This has been taking a while but we definitely want...
Hmm we really should have more specific caliper output, it's possible all the time is spent in IO and there's not much we can do about it.