GPU acceleration is awesome, but writing fast CUDA can be challenging. This comprehensive two day course will enable you to become proficient in writing and optimizing applications for GPU acceleration.
Partnering closely with NVIDIA, AccelerEyes CUDA training courses are the fastest way for you to become proficient at CUDA programming for NVIDIA GPUs. AccelerEyes is uniquely equipped to provide training for NVIDIA CUDA GPUs due to our extensive experience programming ArrayFire. We have helped thousands of organizations speedup their code and our primary objective is to help you increase productivity while maximizing the return on your hardware.
Course Goodie BagAll training courses include the following:
Instruction by an excellent and interesting expert. Many hands-on exercisesUse of a laptop with CUDA capable GPUChoice of Linux or Windows operating systemPrinted manual of lecture materialElectronic copy of programming exercisesCertificate of Completion
CUDA Training Course SyllabusDay 1: Introduction to CUDA
Lectures:GPU Computing OverviewThe CUDA Programming ModelBasic Dataset Mapping TechniquesCUDA Libraries, ArrayFireAsynchronous OperationProfiling ToolsPractice:A Simple CUDA KernelEquivalent ArrayFire ExampleUsing CUDA LibrariesMonte Carlo Pi EstimationTiming CUDA and ArrayFireDebugging CUDA Code
Day 2: CUDA Optimization
Lectures:CUDA Architecture: Grids,Blocks,and ThreadsCUDA Memory Model: Global,Shared and Constant MemoryAdvanced Mapping TechniquesCUDA Streams: Asynchronos Launches and Concurrent ExecutionArrayFire: Lazy Evaluation and Code VectorizationPractice:Matrix TransposeOptimization Using Shared MemoryMedian FilterOptimization Using Constant MemoryCUDA Stream ExampleArrayFire Example: Nearest Neighbor Algorithm