Background
Advances in space technology have led to an increasingly wide range of heavy computation and power requirements specific to the unique objective of each mission. The ambition for superior space computing abilities requires greater processor speeds often increasing power consumption, but the longevity of the mission requires a strict power budget. Consequently, these mission requirements significantly influence many decisions for hardware selection and often lead to a diverse array of embedded processors being used for different missions. Extensive knowledge and engineering centered on specifications of the target processor are required to produce a highly optimized spaceflight embedded systems platform. Additionally, the implementation of various processors for different spaceflight embedded systems reduces the efficiency of mission development. Applying a single processor architecture to various spaceflight embedded systems offers greater platform reuse in future development, simplifying these designs and leading to a reduction in production time and project expenditure. The objective of this research project was to evaluate the capability and performance of Advanced RISC Machine (ARM) processor configurations to determine the viability of using the next-generation space qualified LS1046A high-performance processor, to meet the diverse processing and power requirements of current and future spaceflight system missions.
Approach
For this effort, benchmark testing was performed to evaluate the performance capability and power consumption of the LS1046A compared to SwRI current processor offerings. The benchmarking was separated into two categories.
- Synthetic Legacy Comparison: These benchmarks included a standard synthetic suite of Dhrystone, Whetstone, and CoreMark to compare the legacy processors to the LS1046A.
- Real World Performance: This benchmarking involved evaluating the LS1046A’s performance when running algorithms more commonly used in real world applications. This set of benchmarks was used to evaluate the performance of the LS1046A more closely to that of the MPC8548E which is currently offered as a high-performance solution. These applications included JPEG-LS compression, an image processing pipeline, and a machine learning framework.
Accomplishments
Results from the benchmarking tests demonstrated that the LS1046A is a viable processor for spaceflight applications and that future missions requiring increased or variable computing capability with reduced power consumption will benefit most. The LS1046A exhibited a significant advantage in performance when compared to the legacy processors. Compared to the baseline high-performance processor, the LS1046A demonstrated a roughly 530% increase in performance with reduced power consumption when running the Coremark benchmark at 800MHz. However, the LS1046A may not be suitable for platforms that require extremely low core frequency and power as it was unable to reach the targeted power offered by the low powered legacy options due to frequency scaling constraints. The performance and power reduction provided by the LS1046A along with its ability to provide on demand frequency scaling and core power mode switching, allows the processor to be configured for a variety of spaceflight embedded system missions.