Service Description

This is the live, 1-on-1 version of the Accelerated Computing using CUDA C/C++ and GPUs Course Outline: - nVidia GPU Architecture to Support CUDA C - Intro to CUDA C and Host program format - CUDA threads, blocks, and indexing > nvcc compiler > kernel launch > memory management > kernel and host code synchronization - Programming assignments to illustrate index assignments for thread level, concurrent processing - CUDA Memory Hierarchy - Shared memory and thread synchronization - DRAM Circuit Operation Considerations: Access Types, Latency Caused by Non-Batch Fetches - Performance Considerations - Tensor Cores: Architecture and NN Application - Brief Coverage of PyTorch with CUDA, TensorRT & ONNX, cuDNN, and cuVSLAM - Learn and use Jupyter Notebook/Colab for Python programming and free use of GPU via internet browser. - Iterative Optimizations with the NVIDIA (Nsight Systems) Command Line Profiler nsys - Learn to use the Nvidia Nsight Systems Visual based performance data analysis for both CPU and GPU code - Obtaining and using GPU's number of Streaming Multiprocessors (SMs) and warp size to further optimize run time (if possible) - Memory Fault Analysis and Asynchronous Streaming of Data - A project demonstrating everything learned is required to obtain a certificate of completion! .