Dear Firedrakers, As you will have observed from your diary, we (the core Firedrake people at Imperial) are scheduled to spend Wednesday to Friday on a documentation sprint. I have a couple of proposals about this 1. The starting assumption is that individuals will be working on documenting bits of Firedrake they are working on. This hopefully maximises the benefit to individuals and means documentation is written by the people with the relevant expertise. 2. We should meet mid-morning tomorrow to coordinate exactly what we're doing. Regards, David -- Dr David Ham Departments of Mathematics and Computing Imperial College London http://www.imperial.ac.uk/people/david.ham
I didn't read this but it might be interesting: http://arxiv.org/abs/1401.3301 --cjc
On 15/01/14 09:07, Colin Cotter wrote:
I didn't read this but it might be interesting: http://arxiv.org/abs/1401.3301
It's a little hard to tell, but they seem to have reinvented the local matrix approach (e.g. Markall, et al 2013, http://onlinelibrary.wiley.com/doi/10.1002/fld.3648/full) and then built a slightly smarter version that does some kind of block-wise insertion (to better utilise cache? it's unclear). The numbers they report indicate that their optimised versions go faster than the unoptimised ones. There's no analysis of whether that is actually fast. Also, results only for P1, "Pk is exactly the same", except that it's not (e.g. Cantwell, Sherwin, Kirby, Kelly 2011, From h to p efficiently: Strategy selection for operator evaluation on hexahedral and tetrahedral elements). Lawrence
participants (3)
-
Colin Cotter
-
David Ham
-
Lawrence Mitchell