Rate This Document
Findability
Accuracy
Completeness
Readability

Calculation Result Differences

The Weather Research and Forecasting (WRF) Model is designed for fine-scale weather simulation and forecasting. It is one of major high-performance computing (HPC) application scenarios. Weather forecasts are closely related to enterprise production, people's life, and disaster prevention and mitigation. To protect people's lives and property to the greatest extent, the forecasts should be as accurate and stable as possible.

In bidding activities of meteorological bureaus, they require that the forecast results of a new cluster be consistent with those of the existing cluster to ensure that the new cluster environment is accurate, stable, and efficient to be put into use. Therefore, a large amount of manpower was invested in cluster server replacement in the past. When a meteorological cluster is ported to the Kunpeng platform, the running result may be inconsistent with that of the existing cluster.

In addition to the meteorological field, in an HPC program, the results may be inconsistent when the cluster, CPU architecture, or compiler is replaced. A small difference is amplified after a large number of iterative calculations in HPC. As a result, the calculation results differ greatly.