Has anyone benchmarked OpenCV on the Xeon Phi Coprocessor? Has there been any work planned/underway/completed on implementing Phi SIMD intrinsics in OpenCV? Does anyone know if Intel's compiler is able to translate SSE instructions to equivalent Phi vector instructions to leverage existing manual vectorization in the code? Trying to decide if purchasing one of these cards would be worth it for OpenCV based research projects!
Josh - OpenCV is working on Xeon Phi here at Intel. It is compiled for the offload model so workloads can migrate from MIC to the host and back. But there is little vectorization at this point so performance is not near the speed of the host. But because both host and MIC are working, overall throughput can be better with MIC. The key thing is that the Xeon Phi is able to run all the OpenCV algorithms and now the challenge is to optimize the important ones with compiler directives.
The SSE instructions present in OpenCV are not present on MIC. IPP is not supported on MIC so that option is out. Algorithms with floating point should benefit from the Intel compiler.
This post has NOT been accepted by the mailing list yet.
Hi Bobdavies2000us and Everyone,
I would like to know how to run opencv programs by using the intel xeon phi co processor card? What are my options? May i use the opencl module? Also, how can i work with the Transparent API and Xeon Phi ? Any suggestions or thoughts in this regard will be greatly appreciated.
I would like to know how to build OpenCV library for the Intel xeon phi coprocessor. I have read on this thread that OpenCV is working on Xeon Phi at Intel. How do I compile OpenCV for the offload model ? May I use the OCL module of Xeon Phi to compile programs for Xeon Phi? A few insights into this would be helpful.