Adaptive scheduling with dynamic partition load balancing for fast partition compilation

Abstract

Methods, systems, and computer-readable storage medium embodiments are disclosed herein for adaptive scheduling with dynamic partition load balancing for fast partition compilation. One embodiment includes detecting, by at least one processor, an amount of available hardware resources usable by an electronic design automation (EDA) process for a design description via a plurality of computing elements, and analyzing the design description to generate an estimate of the amount of hardware resources to be used by the EDA process. In some further embodiments, the at least one processor may compare the estimate to the amount of available hardware resources and adjust a memory allocation for the EDA process or a specified number of computing elements of the plurality of computing elements used in parallel by the EDA process. Further, according to some additional embodiments, the at least one processor may calculate a weighted load average for the plurality of computing elements.

Description

[0001] Cross References to Related Applications [0002] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 072,401, entitled "Adaptive Scheduling with Dynamic Load Balancing for Fast Partition Compilation," filed August 31, 2020, which is incorporated herein by reference in its entirety. technical field [0003] The present invention generally relates to improving the performance of electronic design automation (EDA) tools. More specifically, the present invention relates to improving the efficiency and speed of partition compilation according to design specifications. Background technique [0004] Modern EDA tools may allow multiple partitions to be used to compile a hardware design specification or description (eg, from a higher level hardware description language). While using multiple partitions allows tools to manage complex and parallelized workloads, for designs that can be relatively large and particularly complex, the resulting reso...

AI Technical Summary

Problems solved by technology

While using multiple partitions allows tools to manage complex and parallelized workloads, for designs that can be relatively large and particularly complex, the resulting resource usage and overall performance are often not optimal

[0005] Therefore, attempts to parallelize workloads, even with distributed cloud-based systems, can result in inefficient use of computing resources, adversely affecting other processing using the same pool of computing resources

This impact can include a slowdown in overall compilation time, or even a deadlock that can stall the compilation process

However, it is often not obvious to the user of the tool how to allocate computing resources to complex compilation jobs to improve the efficiency and speed of compilation

[0006] User attempts to improve performance by changing compilation parameters may randomly affect performance metrics, or may cause further performance degradation, further deteriorating performance

Similarly, traditional automation technologies lack sufficient maturity, and they produce similarly problematic results

Images