In environments such as space or particle accelerators, soft errors (i.e. correctable bitflips) caused by radiation are much more frequent. Adding elements such as Triple Modular Redundancy (TMR) or Error Correcting Codes (ECC) can detect and correct these errors, however they can induce a large overhead or are not always applicable in case data changes. Furthermore, for certain applications (e.g. handling imprecise data), fault tolerance is not always needed, or only partially beneficial.
We offer a variety of projects augmenting and extending the PULP platform for fault tolerance.
- e-mail: email@example.com
- ETZ J71.2
If you have any other project or ideas relating to fault tolerance, feel free to send Michael Rogenmoser an email or drop by his office to discuss possibilities.