Re: [PATCH] media: venus: Fix MSM8998 frequency table

[Konrad Dybcio]

On 6/6/25 2:14 PM, Vikash Garodia wrote:
> Hi,
> 
> On 6/6/2025 5:29 PM, Arnaud Vrac wrote:
>> Le dim. 1 juin 2025 à 09:46, Dmitry Baryshkov
>> <dmitry.baryshkov@xxxxxxxxxxxxxxxx> a écrit :
>>>
>>> On Sat, May 31, 2025 at 02:22:00PM +0200, Konrad Dybcio wrote:
>>>> From: Konrad Dybcio <konrad.dybcio@xxxxxxxxxxxxxxxx>
>>>>
>>>> Fill in the correct data for the production SKU.
>>>>
>>>> Fixes: 193b3dac29a4 ("media: venus: add msm8998 support")
>>>> Signed-off-by: Konrad Dybcio <konrad.dybcio@xxxxxxxxxxxxxxxx>
>>>> ---
>>>>  drivers/media/platform/qcom/venus/core.c | 10 +++++-----
>>>>  1 file changed, 5 insertions(+), 5 deletions(-)
>>>>
>>>
>>> Verified against msm-4.4
>>
>> Hello,
>>
>> The current values are based on
>> https://git.codelinaro.org/clo/la/kernel/msm-4.4/-/blob/caf_migration/kernel.lnx.4.4.r38-rel/arch/arm/boot/dts/qcom/msm8998-vidc.dtsi
>> which we've been using in production for many years.
> 
> I see -v2 updates these to 533/444/.. MHz [1]. If the value changes based on
> board variants, these need to be picked from board DT instead of driver then.

I believe they don't (although there exist some SoCs where fuses determine
maximum frequency for a given target). We really want to move off of in-driver
freq tables but that is still in progress.

> 
> [1]
> https://git.codelinaro.org/clo/la/kernel/msm-4.4/-/blob/caf_migration/kernel.lnx.4.4.r38-rel/arch/arm/boot/dts/qcom/msm8998-v2.dtsi#L1140

Arnaud, as Vikash mentioned, dvfs tables and some other magic values may
get overriden in socname-v2/-v2.1/v3 etc. I'm not a fan that downstream
leaves irrelevant information for old revisions in place, instead of
simply replacing them, but what can I do..

Unless you somehow came into posession of v1 SoCs (which I believe were
totally internal), your hw has not been stretching its legs fully for
all this time.

Konrad