Guys, Has anyone read about RootBeer - https://github.com/pcpratts/rootbeer1
My understanding is that it can compile Java bytecode -> CUDA GPU code. A Java class that you want to run on your CUDA GPU must subclass the RootBeer Kernel interface. So, I wondering what would happen if I 1) implement Kernel with some simple Clojure code that just e.g. adds a couple of numbers, get the class out onto disc, compile it with the Rootbeer compiler and try running the result on my GPU. 2) Pull the RootBeer compiler into a running Clojure image and figure out how to do the same thing on-the-fly. 3) Write e.g. (cuda-map f coll) which embeds f in a Kernel, copies the contents of coll and the new kernel up to your CUDA GPU and executes f on the contents of coll in parallel producing a new coll which is copied back into Clojure world and returned... Of course, there are lots of things to be resolved here, I have only had time today to Compile up the CUDA download, clone the RootBeer git repo, compile it and try to run the example program - ArraySumApp. I got an error which I am thinking indicates that RootBeer doesn't like my "1.7.0_03-icedtea-mockbuild_2012_06_13_11_56-b00" - the current version of Java on my Fedora16 box, but... think how much fun I could have with this little project :-). Anyway, I thought I would just announce my intentions to play with this on the list, in case anyone else was interested or already doing so and fancied collaborating.... or, maybe someone will be able to tell me why it will never work and save me the effort :-) all the best, Jules -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
