Although PLCopen normally deals with the software environments, it was the best choice to define a benchmarking tool for measuring the performance of the different CPU architectures. Nowadays the CPU suppliers can tune their architectures to the specific needs. However, the needs of the industrial world in general are not clearly specified for them. The strict need for real-time parallel to other operations are not so much known in other industries and needs clarification between the CPU suppliers and the control suppliers.
PLCopen did benchmarking in the past with a focus to different user applications, including motion control. However the network seemed to be the decisive factor and further development was stopped. Today Benchmarking is back on the drawing board, but with a different scope. In April PLCopen held a successful kick-off meeting on this new approach.
This new PLCopen benchmarking workgroup has a more hardware resource oriented (like processor and computing architecture) approach. With this new initiative, the benchmark can indicate the effective performance of the different processors and architectures as used in the industrial controllers. By the time the IEC program can run, it is already "too late" to tune and much time and effort is already invested.
This workgroup is defining a common language between the parties involved, ranging from CPU suppliers to control suppliers, and combined this with common test scripts for real-time and the influence of the other processes (“Noisy Neighbors”). This will include micro-benchmarks for the CPU specific tasks and subsets of instructions and application oriented benchmarks. Overall the suppliers are not looking for highest performance but for predictable, reliable and deterministic approach at good pricing: “the best bang for the buck!”
Currently on the CPU side this initiative is supported by Intel, AMD and ARM, while others are in the process.
The proposed architecture for the first phase looks like this: