[Freedreno] Freedreno on i.MX53

Rob Clark robdclark at gmail.com
Fri Jul 31 06:08:30 PDT 2015


On Fri, Jul 31, 2015 at 7:02 AM, Martin Fuzzey <mfuzzey at parkeon.com> wrote:
> On 30/07/15 20:14, Ilia Mirkin wrote:
>>
>> On Thu, Jul 30, 2015 at 1:32 PM, Martin Fuzzey <mfuzzey at parkeon.com>
>> wrote:
>>>
>>> Comparing with the command stream from the blob driver is complicated by
>>> the
>>> blob driver using the hardware binning commands.
>>
>> At least on a3xx/a4xx you can disable binning (and texture tiling)
>> with a few statements in an adreno_config.txt file in the cwd.
>>
>> binning=sw
>> textureTiling=0
>>
>> That should match things up better if the a2xx driver uses the same
>> things.
>
>
> Thanks for that.
>
> My blob uses "yamato_panel.txt" as the config file and understands
> "binning=sw" but not "textureTiling" (and strings doesn't show anything
> similar).
>
> Interestingly, while  using that does make the binning more like the
> freedreno way (there's no longer an initial "binning shader" and there is
> now a DRAW_INDX per tile) the VGT_CURRENT_BIN_ID_MIN/MAX registers are still
> being set and CP_DRAW_INDX_BIN commands are still being issued.

just fwiw, even on a3xx and later blob version, when you disable
binning the blob still writes some binning related registers
unconditionally.. but those seem to be harmless.  The CP_DRAW_INDX_BIN
part is a bit more strange.

Somtimes I run a test against blob w/ render target small enough to
fit in a single bin, which in your case might be enough to get it to
skip the extra CP_DRAW_INDX_BIN packet.  In either case, getting
things working with a single bin is a useful first step.

BR,
-R


> I'll have to do some more digging to see if this is important..
>
>
> Martin
>


More information about the Freedreno mailing list