Re: Hummingboard DTB (imx6)
On Sunday 19 November 2017 19:08:13 Alan Corey wrote:
> > SPI support is labeled as broken in debian. And likely will remain
> > so until some one who understands it decides to work on it. Some of
> > the problem is in obtaining the headers for the gpio that is on the
> > individual card.
> >
> > Decent SPI support, as in error free writing at 41 megabaud, in 32
> > bit packets, and reading back, error free, the results at 25
> > megabaud, from a Mesa 7i90HD interface card over a 1" long cable, is
> > currently done ONLY from a user-space module that is now part of
> > linuxcnc, and ONLY on a raspberry Pi 3b. This code can probably be
> > made to run on other platforms, but so far I haven't even been
> > successful at removing the "am I running on a pi-3b checks". After
> > that, then figure out how to get the headers for the gpio part of
> > the SoC thats on the boards you have, of which there are apparently
> > at least a dozen mutually incompatible versions just from Broadcom.
> > I'd certainly bow and give homage to the person who can take rpspi.c
> > and its limited header dependencies, rename it to rkspi.c and make
> > it run on the rock64 board at similar speeds. Although its not that
> > much faster than a pi, it comes without the very small funnel that
> > all non-gpio generated data must get (its an internal usb2 hub) thru
> > on a pi, and which throws away keyboard and mouse events in
> > wholesale amounts. The rock64, compared to the pi, feels to be at
> > least 5x faster than the pi will ever be because of this one
> > difference in internal architecture.
> >
> > A universal driver would be nice, but I suspect the best is going to
> > be a driver per platform. There is not a universal gpio, yet. Maybe
> > another 5 years for the winner to be shook out of the bushes far
> > enough to become a standard.
>
> That's a drag, I bought a PocketBeagle a few months ago. SPI seems to
> be the only video. I'm still getting adapters together and such . It
> has 2 SPI
The actual number of spi slaves can be quite large if the cs lines are
decoded. Generally I think most masters can have up the 3 cs lines right
out of the gate. Properly decoded, that n^2-1 or 8 slaves. Some however
may be used internally but generally the internal stuff has its own
master, which may be on a completely separate set of pins, which may, or
may not be accessed from the real world. Usually those pins are internal
to the board if they even get out of the SoC.
I'm with you on the drag bit. I bought a pair of the 4gig rock64's out of
the 2nd run, and I have stretch running on one of them, the other is
still in its little plastic box serveing as a light paperweight in front
of my monitor. The one running stretch can make an rtai kernel from
scratch in less than 2 hours. One of my pi's doing that would be 2 or
more days...
The difference IS that obvious.
Cheers, Gene Heskett
--
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page <http://geneslinuxbox.net:6309/gene>
Reply to: