ARMv8 processors are becoming more common in the industrial market – also thanks to theire use in recent FPGAs like the Xilinx Zynq UltraScale+ MPSoC. It is useful to understand how well these processors perform with Preempt-RT. In the presentation Preempt-RT Latency Benchmarking of the Cortex-A53 processor Paul Thomas from AMSC compares the CortexA53 utilized in the Xilinx Zynq UltraScale+ MPSoC processors in the Enclustra Mercury XU5 module with a Cortex-A9 based module.

Three different latency tests were performed: standard cyclictest latency, external hardware interrupt latency (as utilized by the I/O-subsystem) and Ethernet UDP latency. Analyzing the performance of a hardware interrupt can be tricky, the method of utilizing a timer with a capture function is detailed in the presentation.