Course

Syllabus

Review of Traditional Computer Architecture Basic five stage RISC Pipeline, Cache Memory, Register File, SIMD instructions, GPU architectures – Streaming Multi Processors, Cache Hierarchy, The Graphics Pipeline, Introduction to CUDA programming. Multi-dimensional mapping of dataspace, Synchronization, Warp Scheduling, Divergence, Memory Access Coalescing, Optimization examples: optimizing Reduction Kernels, Optimization examples: Kernel Fusion, Thread and Block. OpenCL basics, OpenCL for Heterogeneous Computing, Application Design: Efficient Neural Network Training/Inferencing.

Text Books / References

1. Jason Sanders and Edward Kandrot, “CUDA by Example: An Introduction to General-Purpose GPU Programming”, Addison-Wesley, 1^st edition, 2010. 
2. Benedict Gaster,Lee Howes, David R. Kaeli, “Heterogeneous Computing with OpenCL”,Morgan Kaufmann; 1^st edition, 2011. 
3. David Kirk and Wen-meiHwu, “Programming Massively Parallel Processors”, Morgan Kaufmann, 3rd edition, 2010. 
4. John L.Hennessy and David A. Patterson, “Computer Architecture — A Quantitative Approach”, Morgan Kaufmann, 5th edition, 2011. 

Pre-requisite: Nil

Course Objectives:

• To introduce modern computer architectures, including RISC pipelines, GPU structures, and CUDA programming basics.
• To develop skills in GPU optimization techniques, including memory coalescing, synchronization, and efficient kernel design.
• To provide an understanding of OpenCL for heterogeneous computing and its application in neural network training and inference.

Course Outcomes:

CO-PO Mapping