We want to evaluate the expected performance of the server using speccpu2017. In our working scenario, an application instance will be independently deployed on each numa node, and there will be no data communication between instances. However, progress belong to same instance may have shared variables and shared memory within a numa node.
Therefore, when using speccpu for evaluation, we hope to also simulate such scenarios, so that speccpu can avoid the impact of data transmission across numa nodes when running benchmark.
How could I configure it? I did not find the corresponding information in the speccpu document.