On Wed, 25 Jul 2012, Paul T. Bauman wrote: > 2. Currently, only JxW are stored in FEMap. However, I'd like to also store, > and provide access to, the determinant of the Jacobian of the > map (i.e. J, without xW). This quantity is needed in the curl calculation for > HCurl conforming elements and in the shape function > evaluations in HDiv and L2 conforming elements (among others). Since > typically few FE objects are instantiated (and therefore FEMap > objects), this didn't strike me as a particularly heavy memory overhead
The memory overhead is negligable relative to RAM; maybe less so relative to cache. Don't think of it as just memory, think of it as memory that'll be getting hit with every FE::reinit(). > but would be a big FLOP saver. I haven't added this yet, but would > like to once the final version of this patch goes in. I wouldn't mind seeing it added, regardless of what I said above; not having handy access to plain Jacobians is nonintuitive, and worrying about cache feels like premature optimization. --- Roy ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Libmesh-devel mailing list Libmesh-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-devel
