On 05/01/2020 22:37, Tony Lindgren wrote:
Hi,
* Tony Lindgren <tony@xxxxxxxxxxx> [200104 05:51]:
Just changing the alingment fixes the issue. Looks like the minimum
alignment we currently allow is 128, I think 512 was the minimum
that worked for me, so maybe the right fix would be to just change
the minimum to 512 with no specific need to use 4096 for now.
So Matthijs and I chatted about this a bit on irc, and here's what
we concluded so far:
1. We have at least three different alignment needs for tiler
- Linux use of tiler aligned to 128 bytes
- SGX use of tiler aligned to 4096 bytes (or 512 bytes?)
- Fast userspace mapping aligned to 4096 bytes
2. The alignment need may need to be configured by the tiler consumer
in #1 above
3. The alignment need for SGX seems to be based on SGX MMU page size
4. The issue I'm seeing with stellarium on droid4 may be a stride
issue as about one out of 3 or 4 frames is OK and aligning to
512 also fixes the issue maybe because it happens to make
multiple frames align to 4096
So let's wait on this patch until we have more info and know how
the different alignments should be handled.
I don't know if these are related to the issue you have, and it's been a while since I looked at
TILER, but two thoughts after looking at the thread:
- If the usergart is not used, the width of each line has to be expanded to be page size aligned,
right? Do the patches from Matthijs do this?
- Afaik, there's no safe way to use TILER 2D buffers with other IPs. dmabuf exposes the buffers as
contiguous, i.e. including also the area outside the buffer itself (the TILER "stride"). The IPs
have no idea that they should not be touching the are outside the buffer itself. My experience with
this is with the capture drivers, and I don't know if the SGX driver deals with the buffers the same
way, though. If there's special TILER 2D support code in the SGX driver, then it might be safe.
Tomi
--
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
_______________________________________________
dri-devel mailing list
dri-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/dri-devel
