One missing piece of info: systems with DDR4 memory often need a blob to train the memory. This is the case with:
RK3399 boards like ROCKPro64 (note: the Samsung “kevin” Chromebook that Alyssa uses is not affected because it uses DDR3)
Tegra X1 (+coreboot) devices like the Google Pixel C and the (hacked) Nintendo Switch
i.MX8 devices like the upcoming Purism phone (that’s why they’re looking into running the blob on the additional core on the SoC that’s not the main processors)
It’s not impossible to reimplement this code with a FOSS license. It’s just that nobody has done that yet. And it’s not a big blob, it’s mostly tables IIRC, if you’re paranoid about blobs you could audit the disassembly. I’m okay with these blobs, but of course the FSF wouldn’t be :)
(oh and the article doesn’t mention X1 and i.MX8 – they both have a good graphics situation with nouveau and entaviv)
The article itself contains little to no useful content. What it links to is really the content of the piece.
Perhaps of more interest is the comparison of different single-board computers on Wikipedia.
FSF’s take on it is important to some folks.
One missing piece of info: systems with DDR4 memory often need a blob to train the memory. This is the case with:
It’s not impossible to reimplement this code with a FOSS license. It’s just that nobody has done that yet. And it’s not a big blob, it’s mostly tables IIRC, if you’re paranoid about blobs you could audit the disassembly. I’m okay with these blobs, but of course the FSF wouldn’t be :)
(oh and the article doesn’t mention X1 and i.MX8 – they both have a good graphics situation with nouveau and entaviv)
Also relevant: https://lobste.rs/s/vmphwq/introducing_panfrost_for_mali_midgard
Albeit I don’t think this is mainline yet, so some time off.