On 19 Aug 2009, at 20:03, joel wrote:
DE wrote:
either of cases need some templates specializations but in a slightly different manner (i didn't think so for which is harder to implement All this is doable with a simpel proto grammar + transform
For extra geek points why not implement an entire algebraic simplification engine to do appropriate vector calculus transforms first [*], http://en.wikipedia.org/wiki/Vector_calculus_identities Like a red rag to a bull! Seriously though, I get the feeling that if one's too clever it will become impossible for the user to know what they are actually doing - therefore a simple expression that would normally be carried out by a single BLAS / LAPACK style library call could end up being de- optimally represented by a set of other calls. -ed [*] My tongue is very firmly in my cheek - I'm not being serious!