[Gta04-owner] Graphics SDK 5_01_01_01

Dr. H. Nikolaus Schaller hns at goldelico.com
Thu May 1 15:45:03 CEST 2014

Am 01.05.2014 um 15:28 schrieb Josua Mayer:

> Hi,
> Am 25.04.2014 19:33, schrieb Dr. H. Nikolaus Schaller:
>> Hi, sorry for the late reply.
>> Am 19.04.2014 um 12:42 schrieb Josua Mayer:
>>> # Now with attachment
>>> Am 18.04.2014 22:54, schrieb Josua Mayer:
>>>> Hello everyone,
>>>> a while ago there was a new graphics sdk released by TI and I
>>>> now took the time to get it to build for our 3.12 kernel, 
>>>> 3.12-wip-pvrsgx. I also checked if it would build against the 
>>>> 3.12.7 branch now, but it doesn't, the reset-framework patches
>>>> in the wip branch are still needed.
>>> ....
>>>> I haven't yet done any testing on the device, but will do that
>>>> as soon as I have set myself up a fresh system on sdcard. I
>>>> will then report if there are any problems running the GLES
>>>> demos, or if everything is just working fine.
>>> Hello again,
>>> So I have performed 2 test runs now using a fresh debian-wheezy 
>>> system, and my self-compiled kernels.
>>> First test: goldelico 3.12-wip-pvrsgx merged with 3.12.7, pvr
>>> modules builtin, debug kernel panic at startup, see attached
>>> wip3127.log
>> Yes, that one was abandoned because I was not able to get it up and
>> running.
>>> Second test: goldelico 3.12-wip-pvrsgx merged with 3.12.7, pvr
>>> modules not built pvr modules compiled directly from graphics
>>> sdk boots fine modprobe pvrsrvkm fails with the following
>>> message: root at gta04:/usr/local/lib# modprobe pvrsrvkm [
>>> 359.127838] PVR: PVRCore_Init [  359.139831] PVR:
>>> PVRSRVDriverProbe(pDevice=bf26cb90) [  359.154907] pvrsrvkm
>>> pvrsrvkm: PVRSRVDriverProbe: error: reset_control_get [
>>> 359.170318] pvrsrvkm: probe of pvrsrvkm failed with error -2 [
>>> 359.177551] PVR: PVRCore_Init: major device 245
>> Yes, I remember that there is something wrong with the "reset
>> framework".
>> TI isn't doing a really good job in making things compatible to
>> mainline Linux (which is to some extent understandable - but makes
>> it very complex if you don't want to use a TI kernel).
> So I have been reading a lot of google and opened a few datasheets,
> some grep, etc...
> and stumbled on these:
> https://lkml.org/lkml/2013/10/8/566
> https://git.ti.com/wilink8-wlan/wilink8-wlan-ti-linux-kernel/commit/4422e028116e147ac190de0d17e4f05ddb95f307?format=patch
> The ti_reset.c seems to contain processor-specific information about
> the reset controller, but there is no dm or omap part mentioned,
> though apparently DM and AM parts should be identical in this aspect.
> And, the reset controller driver has to be loaded, I think thats what
> the dts change is meant to achieve. I havent seen any hint on how to
> do that in a boardfile (except arch/arm/mach-imx/src.c,
> reset_controller_register).

Thanks for these links! As soon as we have fixed the major issues of
the 3.15 kernel I can take a look into it. Maybe 3.15 is already a little
better developed than 3.12 in that respect.

> This kernel stuff is still far beyond me sadly, so maybe this
> information helps someone else to have a look.
>>> I intend to do another test of 3.12-wip-pvrsgx without merged
>>> 3.12.7, tomorrow, though I guess there is little point.
>>> Maybe anyone wants to invest soem time into PVR (or just swear
>>> loudly, feel free!) and find out whats going on here.
>> Generally my hope is that the 5.0 SDK can be made working on the
>> 3.12 (or 3.15) kernel.
>>> So far have a nice weekend everyone
>> Thanks and same for this new weekend!
>> BR, Nikolaus


More information about the Gta04-owner mailing list