I don't know why I failed to get this PCIe Coral TPU working previously but all is good now. Now I just need people to come to the front door so I don't actually have to put clothes on and test it myself.
@mw1cgg [ 
] naked man [97% probability]
@flangey It'll be one of those baby aubergines I can get from the Asian shop.
I have actually put some clothes on now though.
@mw1cgg I see, a classic Megatokyo fan.
@M0KHR No, just lazy and didn't want the neighbours to see me in my curry stained joggers.
@mw1cgg I got embroiled in some Coral/ PCIe / Linux kernel / HP interaction that required a kernel recompile. Once I'd done that, I'd lost the enthusiasm for actually doing the useful part
@flangey Urgh. I don't think I've actually compiled a kernel in 20 years.
@mw1cgg @flangey ah yeah I had to step my dad through this. On the bright side it's not actually a kernel compile, but it does involve DKMS, and the Coral module does not compile against all kernels, so if you are planning to use them you should plan on ensuring that the kernel version does not change
(Also PCIe passthrough is a option, if you use a platform where that is sane, but given I assume you are using frigate or w/e that is likely a bad idea for other reasons)
@benjojo @mw1cgg yeah this was an issue with HP and mtrr and I had to use this patch https://github.com/kiler129/relax-intel-rmrr to pass it through to VMs. alternative is just to run it in dom0 of course.
It all started with the PCIe adaptor needing a track cutting so the machine would even boot. After all that palaver I ran out of steam :(