OpenCL Visual Profiler is a tool that can be used by developers to view information about performance bottlenecks to facilitate problem resolution faster and in a more efficient manner. OpenCL Visual Profiler is available to developers that are NVIDIA GPU Computing Registered. The next CUDA Toolkit contains this multi-functional utility called OpenCL Visual Profiler.
OpenCL (Open Computing Language) Visual Profiler has a number of features, which are handy to developers. Some of these features are:
• Customizable graphs that facilitate attention to important problem areas.
• Importation and exportation to CSV for analysis.
• Auto-analysis with basic functions to present problems related to warp serialization
• Profiling of kernel efficiency, instruction issue rate, and actual hardware signals
• OpenCL Visual Profiler also provides timing of memory copies between the GPU dedicated memory ad system memory.
OpenCL Visual Profiler comes with a guide that contains detailed discussion on the following topics: Control Flow, Memory Optimizations, NDRange Optimizations, Instruction Optimizations, Performance Optimization Strategies, Performance Metrics, and GPU Computing with OpenCL.
OpenCL Visual Profiler presents a number of opportunities for optimization. This tool is designed to help in the programming for CUDA architecture. This utility facilitates implementation of high performance algorithms. OpenCL Visual Profiler is also helpful in understanding GPU computing best practices.