Hi
One of the higher level languages that I really would like to see 'enabled' to use the E16/64 capacities is R (http://www.r-project.org)
I have done some initial research on how to accomplish this best, and this is the preliminary conclusion.
1) using E16/64 capacity directly from R, through openCL code (This would probably work more-or-less out of the box with the existing R package: opencl). While this is the most direct path to making the E's available in R, it also means that the R user would have to be versed in openCL kernel programming to use the E16/64. So probably not very appealing to the general R user community (statisticians, scientists of all kind, quants,...)
2) Having R use a linear algebra library (lapack, BLAST) that has been compiled *with E16/64* support in mind ! That would be awesome since the end user would not have to know anything about openCL and parallellization but could still leverage the Epiphanys capacity through R. Probably very appealing to the wider R community. The difficulty here is that one would need to actually write that library(!), for example using the great viennaCL lib http://viennacl.sourceforge.net
3) Maybe there is even a middle way between 1) and 2) but I am not sure what that would look like. Maybe very similar to what D.Eddelbuettel has done in the 'RcppEigen' package. It makes the linear algebra functions provided by the 'Eigen' library available in R.
http://cran.r-project.org/web/packages/ ... uction.pdf
Would be good to hear if anyone here has the same interest and what their thoughts are on this !
Cheers
http://censix.com