Closed miklos1 closed 7 years ago
This was a sanity check I put in because people kept on creating non-polynomial integrands that had ludicrous estimated degrees and so the quadrature was massive overkill. I agree that these schemes are entirely legit for high-degree elements. Shall we just remove it?
Perhaps we could add a warning in the form compiler if the estimated degree is more than a factor 10 (?) larger than the degree of the test/trial spaces?
I approve
Thanks, this sounds good.
Fixed in upstream master. firedrakeproject/tsfc#128 is adding a better alternative.
Because of this sanity check. However, high order quadrature degrees are completely legitimate when solving for high-order discretisations. @wence-: what should we do?