Hi Phil, > Thanks Karl ! > This will allow serve us when we deal with multiple GPU ;)
I'm pretty happy with the model now, basically extending the concept of a 'context' in OpenCL beyond OpenCL boundaries: Create vectors as follows: viennacl::vector<T> x(42); //vector in default context viennacl::context host_context(viennacl::MAIN_MEMORY); viennacl::vector<T> y(42, host_context); //vector in main RAM // vector in i-th OpenCL context, size will be specified later viennacl::context opencl_context(viennacl::ocl::get_context(i)); viennacl::vector<T> z(opencl_context); viennacl::context cuda_context(viennacl::CUDA_MEMORY); viennacl::vector<T> w(42, cuda_context); //vector on CUDA device and if you cant to migrate a vector to a different context, simply: viennacl::switch_memory_context(z, host_context); viennacl::switch_memory_context(w, opencl_context); and all the internals are taken care of automatically (including conversions if e.g. int and cl_int don't have the same bit representation). Similarly for the various matrix types. What does not yet work is something tricky like y = z + w; where vectors reside in different contexts. I think it's reasonable to use the lvalue as the target context, but it may also create dangerous performance pitfalls if such implicit transfers happen silently... Best regards, Karli ------------------------------------------------------------------------------ Get your SQL database under version control now! Version control is standard for application code, but databases havent caught up. So what steps can you take to put your SQL databases under version control? Why should you start doing it? Read more to find out. http://pubads.g.doubleclick.net/gampad/clk?id=49501711&iu=/4140/ostg.clktrk _______________________________________________ ViennaCL-devel mailing list ViennaCL-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/viennacl-devel
