[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <6f11921a-4ee8-4f40-9131-529f548f681a@kernel.org>
Date: Mon, 7 Apr 2025 16:39:26 +0100
From: Bryan O'Donoghue <bod@...nel.org>
To: Konrad Dybcio <konrad.dybcio@....qualcomm.com>,
Renjiang Han <quic_renjiang@...cinc.com>,
Stanimir Varbanov <stanimir.k.varbanov@...il.com>,
Vikash Garodia <quic_vgarodia@...cinc.com>,
Mauro Carvalho Chehab <mchehab@...nel.org>, Rob Herring <robh@...nel.org>,
Krzysztof Kozlowski <krzk+dt@...nel.org>, Conor Dooley
<conor+dt@...nel.org>, Bjorn Andersson <andersson@...nel.org>,
Konrad Dybcio <konradybcio@...nel.org>
Cc: linux-arm-msm@...r.kernel.org, linux-media@...r.kernel.org,
devicetree@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v6 2/4] media: venus: pm_helpers: use opp-table for the
frequency
On 09/01/2025 13:05, Konrad Dybcio wrote:
> On 2.01.2025 6:38 AM, Renjiang Han wrote:
>>
>> On 12/23/2024 10:17 PM, Konrad Dybcio wrote:
>>> On 19.12.2024 6:41 AM, Renjiang Han wrote:
>>>> The frequency value in the opp-table in the device tree and the freq_tbl
>>>> in the driver are the same.
>>>>
>>>> Therefore, update pm_helpers.c to use the opp-table for frequency values
>>>> for the v4 core.
>>>> If getting data from the opp table fails, fall back to using the frequency
>>>> table.
>>>>
>>>> Signed-off-by: Renjiang Han<quic_renjiang@...cinc.com>
>>>> ---
>>>> drivers/media/platform/qcom/venus/pm_helpers.c | 53 +++++++++++++++++++-------
>>>> 1 file changed, 39 insertions(+), 14 deletions(-)
>>>>
>>>> diff --git a/drivers/media/platform/qcom/venus/pm_helpers.c b/drivers/media/platform/qcom/venus/pm_helpers.c
>>>> index 33a5a659c0ada0ca97eebb5522c5f349f95c49c7..b61c0ad152878870b7223efa274e137d3636433b 100644
>>>> --- a/drivers/media/platform/qcom/venus/pm_helpers.c
>>>> +++ b/drivers/media/platform/qcom/venus/pm_helpers.c
>>>> @@ -43,14 +43,20 @@ static int core_clks_enable(struct venus_core *core)
>>>> const struct venus_resources *res = core->res;
>>>> const struct freq_tbl *freq_tbl = core->res->freq_tbl;
>>>> unsigned int freq_tbl_size = core->res->freq_tbl_size;
>>>> + struct device *dev = core->dev;
>>>> + struct dev_pm_opp *opp;
>>>> unsigned long freq;
>>>> unsigned int i;
>>>> int ret;
>>>> - if (!freq_tbl)
>>>> - return -EINVAL;
>>>> -
>>>> - freq = freq_tbl[freq_tbl_size - 1].freq;
>>>> + opp = dev_pm_opp_find_freq_ceil(dev, &freq);
>>>> + if (IS_ERR(opp)) {
>>>> + if (!freq_tbl)
>>>> + return -EINVAL;
>>>> + freq = freq_tbl[freq_tbl_size - 1].freq;
>>>> + } else {
>>>> + dev_pm_opp_put(opp);
>>>> + }
>>> I'm not super convinced how this could have ever worked without
>>> scaling voltage levels, by the way. Perhaps this will squash some
>>> random bugs :|
>>>
>>> Konrad
>> Thanks for your comment.
>> The default value of freq is 0, and then dev_pm_opp_find_freq_ceil is
>> used to match freq to the maximum value in opp-table that is close to
>> 0. The frequency values in opp-table and freq_tbl are the same, and
>> dev_pm_opp_find_freq_ceil is used to assign the minimum value in
>> opp-table to freq. So the logic is the same as before. I'm not sure if
>> I've answered your concern.
>
> We talked offline, but for the record: my concern here was about
> clk_set_rate() not scaling RPM/h voltage corners, which this patch
> fixes
>
> Konrad
Konrad is this an RB from you, do you have any other concerns with this
code ?
Dikshita, Vikash ?
I'll give it a test myself ASAP but any other comments or R/B would be
helpful.
---
bod
Powered by blists - more mailing lists