Overview

MindStudio Kernel Performance Prediction (msKPP) provides functions such as performance modeling and analysis, calling of msOpGen operator projects, and automatic tuning based on the Ascend C template library.

• Performance Modeling: Before developing an operator, write an operator expression of an operator implementation solution using the mathematical logic of the operator and the APIs provided by msKPP to obtain the operator performance modeling result of the solution. Since performance prediction requires only the execution time of corresponding algorithms based on input and output sizes, and not actual computation, performance modeling results can be provided within seconds.
• Calling the msOpGen Operator Project: The mskpp.tiling_func and mskpp.get_kernel_from_binary APIs provided by msKPP can directly call the msOpGen operator project.
• Auto Tuning: msKPP provides the generation, compilation, and running of the template library kernel delivery code, and code replacement and auto tuning in the kernel.