Question about porting VCE1 to amdgpu

[Alex Deucher]

On Tue, Jun 27, 2017 at 1:08 PM, Alexandre Demers
<alexandre.f.demers at gmail.com> wrote:
> Just a quick update: things a getting into shape slowly since my paternal
> vacancy is over, but I'm still progressing. I'm debugging things as I can. I
> managed to work around the missing header with the current firmware after
> analyzing what it contains and where it is used.

I'd suggest just tacking the header onto the existing firmware.  The
only thing we use it for on the multi-media firmware is the version
info which you can read out of the raw firmware and populate based on
the code in radeon_vce_init().

Alex

>
> I'm more interested than ever to have VCE and UVD ported to amdgpu, since we
> now have cards from the 300 series at work (one based on GCN 1, the other on
> GCN 2).
>
> By the way, the R7 360 (based on GCN 2, Bonaire Pro) fails to load the
> amdgpu driver because of an error in VCE 2 implementation.
>
> I'll keep you up to date.
>
> Alexandre
>
> On Wed, 14 Jun 2017 at 14:22 Deucher, Alexander <Alexander.Deucher at amd.com>
> wrote:
>>
>>
>>
>> From: Alexandre Demers [mailto:alexandre.f.demers at gmail.com]
>> Sent: Wednesday, June 14, 2017 1:30 PM
>> To: Deucher, Alexander; Christian König; Freedesktop - AMD-gfx
>>
>>
>> Subject: Re: Question about porting VCE1 to amdgpu
>>
>> On Wed, 14 Jun 2017 at 13:09 Deucher, Alexander
>> <Alexander.Deucher at amd.com> wrote:
>>
>>
>>
>> From: amd-gfx [mailto:amd-gfx-bounces at lists.freedesktop.org] On Behalf Of
>> Christian König
>> Sent: Wednesday, June 14, 2017 12:37 PM
>> To: Alexandre Demers; Freedesktop - AMD-gfx
>> Subject: Re: Question about porting VCE1 to amdgpu
>>
>>
>>
>> - Would we need a different firmware version with a different "hdr" for
>> the amdgpu driver?
>>
>> Yes, we should probably release the latest one instead of reusing the one
>> used with radeon.
>>
>> Actually, we should probably stick the same one as radeon for now until we
>> can verify the new firmware in general.  Easier to start with a known
>> working case.
>>
>>
>>
>> OK. Then,  is it expected to have a validation failure with the current
>> firmware? Is the header compatible with how the validation is done under
>> VCE2 and others or should I keep how it was done under radeon?
>>
>> There is no header on the multi-media firmware from radeon, so you'd need
>> to add one.  The header is documented in amdgpu_ucode.h,  All you'd need to
>> do is prepend the header to the existing firmware image.
>>
>>
>>
>> Alex
>>
>>
>>
>>
>>
>>
>> BTW: Does VCE work on CIK? Alex, don't we run into the same issue there as
>> well?
>>
>> VCE works on CIK.  We ported VCE and UVD to CIK as part of the initial
>> amdgpu bring up.
>>
>>
>>
>> I've been using VCE2 port as my template for VCE1. My initial intention
>> was to work on UVD, but I ended up plugging in VCE in the first place. UVD
>> is on my todo list right next, I was expecting to working on it after fixing
>> the VCE part.
>>
>>
>>
>>
>>
>> Alex
>>
>>
>>
>> - Wouldn't it be better to continue loading the driver while having VCE
>> disabled IF we fail when loading or validating the FW? Completely failing to
>> load the driver for this reason seems overkill IMO, since nothing has been
>> loaded in memory and no registry have been modified up to that point.
>>
>> UVD and VCE are actually needed for correct power management. When the
>> blocks fail to initialize you usually sooner or later run into problems with
>> power management (e.g. stuck inside a certain power level).
>>
>>
>>
>> OK, but right now it is disabled, so the situation wouldn't be worst isn't
>> it?
>>
>>
>>
>>
>>
>> - Would it be a good idea to send a patch as a RFC so some of you could
>> help me finish the job and maybe pinpoint where the last modifications need
>> to be done?
>>
>> Well you could, but to be honest without AMD releasing new firmware that
>> is most likely a futile effort.
>>
>>
>>
>> I'll send a patch then, and we'll navigate from there. This will allow me
>> to work on UVD in parallel.
>>
>>
>>
>> Alexandre Demers
>>
>>
>>
>>
>> Regards,
>> Christian.
>>
>> Am 14.06.2017 um 18:22 schrieb Alexandre Demers:
>>
>> Hi,
>>
>>
>>
>> I've been working on porting VCE1 from radeon to amdgpu in the last few
>> weeks. I'm pretty much done and I've enabled the functionality to see how it
>> goes. However, I ended up with an error on the firmware validation (size
>> doesn't seem to fit), thus failing completely in loading the driver. I'm
>> testing on a R9 280X (Tahiti).
>>
>>
>>
>> Three questions then:
>>
>> - Would we need a different firmware version with a different "hdr" for
>> the amdgpu driver?
>>
>> - Wouldn't it be better to continue loading the driver while having VCE
>> disabled IF we fail when loading or validating the FW? Completely failing to
>> load the driver for this reason seems overkill IMO, since nothing has been
>> loaded in memory and no registry have been modified up to that point.
>>
>> - Would it be a good idea to send a patch as a RFC so some of you could
>> help me finish the job and maybe pinpoint where the last modifications need
>> to be done?
>>
>>
>>
>> Thank you!
>>
>> Alexandre Demers
>>
>>
>>
>> _______________________________________________
>>
>> amd-gfx mailing list
>>
>> amd-gfx at lists.freedesktop.org
>>
>> https://lists.freedesktop.org/mailman/listinfo/amd-gfx
>>
>>
>
>
> _______________________________________________
> amd-gfx mailing list
> amd-gfx at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/amd-gfx
>