Re: [PATCH v2 3/6] media: qcom: camss: Add Kaanapali compatible camss driver

[Vijay Kumar Tumati <vijay.tumati@oss.qualcomm.com>]

On 10/19/2025 11:45 PM, Hangxiang Ma wrote:
> On 10/16/2025 4:55 PM, Bryan O'Donoghue wrote:
>> On 15/10/2025 03:56, Hangxiang Ma wrote:
>>> +static const struct resources_icc icc_res_kaanapali[] = {
>>> +    /* Based on 4096 x 3072 30 FPS 2496 Mbps mode */
>>> +    {
>>> +        .name = "ahb",
>>> +        .icc_bw_tbl.avg = 925857,
>>> +        .icc_bw_tbl.peak = 925857,
>>> +    },
>>
>> Looking at other implementations I realise we've been adding avg and 
>> peak without too much review however, wouldn't 925857 / 2 => 462928 
>> be a better value for the average ?
>>
>> ---
>> bod
>
> Ack. Thanks.
Just adding my thoughts on this, as you know the peak/avg bandwidth 
should primarily depends on the sensor data rate and additionally, the 
average BW vote should depend on the buffer sizes in the NIUs/NOCs and 
(although irrelevant here) whether it's an RT or NRT module (file system 
reads/writes from the NRT modules can be averaged and controlled 
better). Fundamentally, the votes from all modules go into the 
calculation of the DDR clock. The current driver does not take into 
account anything. So either way, it is not ideal I think. We can discuss 
and come up with a cleaner approach in a different patch series covering 
all chip sets. Thanks.
>
> ---
> Hangxiang
>