Mapping and Partitioning software task graphs on hardware platforms



The mapping and partitioning of task graphs without violating system constraints is a non-negotiable procedure and, if ignored, it affects time to market and reduces revenue. In this webinar, we will explore a new quantitative approach to mapping and partitioning these task graphs to get accurate estimates of resource usage, response times, and power consumption compared to the traditional approach. Using a quantitative approach, the task graph can be effectively partitioned to preserve the trade-off between cost and efficiency. This approach to product design will address system sizing, number of resources, assignment of tasks to CPU/GPU/TPU/AI devices, memory requirements, bus transfer rate,  bus width, and a cost function that measures the quality of a particular solution. The webinar will discuss how to achieve the best solution for partitioning while preserving each of these limitations and will be useful for systems engineering, architects, designers and marketing.

YouTube: https://youtu.be/npLgTdf9NLw