Port-arm archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: HOWTO for RPi4 and -current?



> But wonder, with due respect for the immensely popular pi's
> designers, why are some things (like GPU interface) so complicated -
> rather so different that OSS developers aren't able to leverage the
> GPU?

My understanding - which may be wrong - is that GPU interfaces are
insanely complex and insanely buggy, to the point that I had someone
who routinely works with such things under NDA tell me that the errata
document for an unnamed GPU they were working with was something like
three times the size of the (multi-hundred-page) documentation it's the
errata for.

The only plausible explanation I've come up with is that GPU vendors
rush their products to market so fast they don't bother fixing bugs
first and then are somehow, simultaneously, catastrophically ashamed of
how buggy their interfaces are.

> Or, are there commercial factors to this?

Yes.  GPUs usually have no public documentation - what documentation
there is is available only under NDA - and the vendor-provided and
-supported way to access them is "run this (huge) closed-source driver
we provide".

This is, of course, unsuitable for an open-source project.  I don't
understand why supposedly open-source-compatible projects like the Pi
put up with it, and the use of undocumented hardware is a major
fraction of why I've never bought a Pi.  I'd rather have no GPU than an
undocumented GPU.

/~\ The ASCII				  Mouse
\ / Ribbon Campaign
 X  Against HTML		mouse%rodents-montreal.org@localhost
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B


Home | Main Index | Thread Index | Old Index