Post #379328
2026-02-23 04:39 UTC
Replies (5)
-
@ai6yr@m.ai6yr.org 2026-02-23 04:40
@EricFielding@mastodon.social Ooh, interesting!
-
@skewray@mathstodon.xyz 2026-02-23 05:05
@EricFielding@mastodon.social JPL did a study with Qualcomm after Mars Helicopter where they radiation tested a particular Snapdragon processor and mapped out all the questionable sections. Then they designed a board with two Snapdragons and an FPGA that compares the results of the two CPUs to ensure nothing is going awry. I wouldn't be surprised if the design is open source. I doubt that JPL itself will ever use it. I was PI on a NASA proposal at JPL, prototyping an instrument using this technology for use at Venus, to detect quakes from orbit. Got laid off a few weeks before submission and the project died. 😢
-
@lain_7@tldr.nettime.org 2026-02-23 12:14
@EricFielding@mastodon.social Be sure to read to the end: “Tapping into the HBS computer has had its challenges. To address reliability, the team developed a “sanity check”: The algorithm runs on the HBS multiple times before one of the rover’s main computers checks to ensure the results match. During testing, the team repeatedly found the rover’s position was off by 1 millimeter. They discovered damage to about 25 bits — a minuscule fraction of the processor’s 1 gigabyte of memory — and developed a solution to isolate those bits while the algorithm runs.” One millimeter discrepancy. 25 flakey bits. And they found them.
-
@penguin42@mastodon.org.uk 2026-02-23 14:24
@EricFielding@mastodon.social Now I want to know how a bunch of bit failures ended up with such a small but well characterised error. I'm more used to things like that ending up with either (a crash) or (touching memory the program doesn't use)
-
@Fedihacker@masto.es 2026-02-23 15:06
@EricFielding@mastodon.social Of course, the generative AI buzzword couldn't be passed.