Questions on multicore Linux, DO-178B and RTOS performance

In many contexts – blog posts, seminars, webinars, via email – I am presented with questions about embedded software matters. I do my best to provide answers to the questioners. However, sometimes I think that the Q&A might have wider interest, so I might publish a few here …

If Linux is installed on two cores in an SoC, will processes be allocated to each core as needed/available?

The simple answer is: maybe. It depends on what software configuration you choose. If you use an SMP [Symmetric MultiProcessing] version of Linux, a single instance runs across both cores and processes will be allocated to a core on which to execute. The way this allocation is performed is a configuration issue. On the other hand, if Linux is installed individually on each core [it is now an AMP – Asymmetric MultiProcessing – system], code execution is managed separately and each process “belongs” to a specific OS instance and, hence, on a specific core.

Is the OS qualified to DO-178B?

It is not possible for an OS to be DO-178B qualified. The standard [which is used for mil-aero applications] requires that the complete software package be certified – both operating system and application code. It is not possible to certify just parts of the code [like the OS] individually. If you need DO-178B, the best questions to ask are:

• Do you have other users of your OS who have done this certification and to what level?
• Do you provide source code? [This is a requirement for certification. Some vendors provide it at no extra cost; some make a charge; some do not make it available.]
• Can you provide any assistance/documentation to help with certification of our application incorporating your OS?

Can we compare performances of various RTOSs?

In theory, the performance of one RTOS may be compared with another, but this can only be done in a really meaningful way by performing the measurements yourself. There are two reasons for this. First, the specific aspects of the performance that relate to your application are what count; each RTOS may perform better in different areas. Second, there is no common language or performance measurement standards. Information quoted on data sheets is hard to compare, as each vendor has their own definition for a number of common terms, and this is only the start of a trail of confusion. This white paper offers guidance on this tricky topic.

This entry was posted on Monday, March 16th, 2020 at 7:27 am. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.