14:32 <samael> am i safe powering on a beagleplay without antennas plugged? i remember some guy's advice long time ago telling me to avoid powering up wifi hardware without antennas to avoid overheating the rf card or something

14:35 <jakllsch> samael: will be safer to have antennas plugged in

14:36 <samael> jakllsch: can you please elaborate on that? i'm curious about the reasons

14:37 <jakllsch> reflected radio power can damage the transmit amplifier

14:37 <jakllsch> and lack of an antenna causes reflection

14:38 <jakllsch> feel free to web search for: radio transmitter no antenna damage

14:44 <samael> jakllsch: thanks a lot!

22:38 <Guest17> Hello

22:38 <Guest17> I am not that familiar with IRC. So I'll just try to ask my question straight away here. I currently doing some performance measurements on a beaglebone black and notice that the prefetcher is not doing its job (at all?).

22:38 <Guest17> Long story short after invalidating the entire cache in kernel space, I traverse a 4k array linearly. While traversing it I get a cache miss at every 64th byte every single time. Not only for the first 128bytes or so. Note that 64 corresponds to the size of a cache line.

22:38 <Guest17> I find this very weird as the traversal pattern is purely linear, meaning that after some accesses the prefetcher should be able to predict all the future accesses and therefor prefetch them into cache so that there is no penalty. Yet it doesn't as I consistently get a cache miss at every 64th access.

22:38 <Guest17> Why?

23:02 <zmatt> Guest17: did you fill that memory before testing, or did you just allocate it but never write anything to it?

23:02 <zmatt> Guest17: also please don't send private messages

23:05 <zmatt> oh in kernel space, hm

23:05 <zmatt> does the cortex-a8 even do prefetching? I don't really remember

23:08 <zmatt> I mean, it does instruction prefetching, but I'm honestly not sure it even does any sort of data prefetching at all (except when prompted using a PLD instruction)

23:12 <zmatt> I don't see anything in the Cortex-A8 TRM to suggest that it does

23:30 <zmatt> samael: you could instead make sure the rf functions are kept disabled through software or DT, or perhaps terminate the antenna ports with 50 ohm resistors (dunno if u.fl terminators exist though, but maybe)

23:33 <zmatt> wait, the cc1352p on the beagleplay is also a 2.4 GHz transmitter, and there's no coexistence signalling between it and the wifi chip? ... that can't be good