Dear all,
after starting our GPGPU work with CUDA, we have been increasingly reliant on OpenCL to achieve cross-platform (between brands A and N) and cross-technology (e.g., to MIC) portability. However, looking at SDK 5.0, it appears that OpenCL support is waning. I get this impression from little things (e.g., OpenCL samples no longer included [though downloadable from web site]) to bigger issues (no profiling support for OpenCL programs in Visual Profiler).
Am I misinterpreting these signs and portents, and this unbalance is just temporary (and will get fixed in 5.1), or should we fall-back to plain CUDA on NVIDIA hardware?