Tag: roofline
All the articles with the tag "roofline".
CUDA Programming (7): Parallel Reduction and Kernel Optimization Techniques
A comprehensive case study on optimizing parallel reduction in CUDA, covering Roofline modeling, control divergence, shared memory bank conflicts, loop unrolling, warp shuffle instructions, and thread coarsening.
CUDA Programming (5): Program Timing and Performance Metrics
A comprehensive guide to measuring CUDA kernel execution time using CUDA Events, calculating effective memory throughput and compute throughput, and analyzing workloads with the Roofline Model.