linux-kernel - Re: [PATCH v5 4/7] drm/msm: adreno: find bandwidth index of OPP and set it along freq index

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <80bed70e-7802-4555-a15e-e06fe46214c6@quicinc.com>
Date: Fri, 13 Dec 2024 18:42:47 +0530
From: Akhil P Oommen <quic_akhilpo@...cinc.com>
To: <neil.armstrong@...aro.org>,
        Konrad Dybcio
	<konrad.dybcio@....qualcomm.com>,
        Rob Clark <robdclark@...il.com>, Sean Paul
	<sean@...rly.run>,
        Konrad Dybcio <konradybcio@...nel.org>,
        Abhinav Kumar
	<quic_abhinavk@...cinc.com>,
        Dmitry Baryshkov <dmitry.baryshkov@...aro.org>,
        Marijn Suijten <marijn.suijten@...ainline.org>,
        David Airlie
	<airlied@...il.com>, Simona Vetter <simona@...ll.ch>,
        Bjorn Andersson
	<andersson@...nel.org>, Rob Herring <robh@...nel.org>,
        Krzysztof Kozlowski
	<krzk+dt@...nel.org>,
        Conor Dooley <conor+dt@...nel.org>
CC: <linux-arm-msm@...r.kernel.org>, <dri-devel@...ts.freedesktop.org>,
        <freedreno@...ts.freedesktop.org>, <linux-kernel@...r.kernel.org>,
        <devicetree@...r.kernel.org>
Subject: Re: [PATCH v5 4/7] drm/msm: adreno: find bandwidth index of OPP and
 set it along freq index

On 12/13/2024 3:07 AM, Neil Armstrong wrote:
> On 12/12/2024 21:21, Konrad Dybcio wrote:
>> On 11.12.2024 9:29 AM, Neil Armstrong wrote:
>>> The Adreno GPU Management Unit (GMU) can also scale the DDR Bandwidth
>>> along the Frequency and Power Domain level, until now we left the OPP
>>> core scale the OPP bandwidth via the interconnect path.
>>>
>>> In order to enable bandwidth voting via the GPU Management
>>> Unit (GMU), when an opp is set by devfreq we also look for
>>> the corresponding bandwidth index in the previously generated
>>> bw_table and pass this value along the frequency index to the GMU.
>>>
>>> The GMU also takes another vote called AB which is a 16bit quantized
>>> value of the floor bandwidth against the maximum supported bandwidth.
>>>
>>> The AB is calculated with a default 25% of the bandwidth like the
>>> downstream implementation too inform the GMU firmware the minimal
>>> quantity of bandwidth we require for this OPP.
>>>
>>> Since we now vote for all resources via the GMU, setting the OPP
>>> is no more needed, so we can completely skip calling
>>> dev_pm_opp_set_opp() in this situation.
>>>
>>> Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
>>> Reviewed-by: Akhil P Oommen <quic_akhilpo@...cinc.com>
>>> Signed-off-by: Neil Armstrong <neil.armstrong@...aro.org>
>>> ---
>>>   drivers/gpu/drm/msm/adreno/a6xx_gmu.c | 39 ++++++++++++++++++++++++
>>> +++++++++--
>>>   drivers/gpu/drm/msm/adreno/a6xx_gmu.h |  2 +-
>>>   drivers/gpu/drm/msm/adreno/a6xx_hfi.c |  6 +++---
>>>   drivers/gpu/drm/msm/adreno/a6xx_hfi.h |  5 +++++
>>>   4 files changed, 46 insertions(+), 6 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gmu.c b/drivers/gpu/drm/
>>> msm/adreno/a6xx_gmu.c
>>> index
>>> 36696d372a42a27b26a018b19e73bc6d8a4a5235..46ae0ec7a16a41d55755ce04fb32404cdba087be 100644
>>> --- a/drivers/gpu/drm/msm/adreno/a6xx_gmu.c
>>> +++ b/drivers/gpu/drm/msm/adreno/a6xx_gmu.c
>>> @@ -110,9 +110,11 @@ void a6xx_gmu_set_freq(struct msm_gpu *gpu,
>>> struct dev_pm_opp *opp,
>>>                  bool suspended)
>>>   {
>>>       struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
>>> +    const struct a6xx_info *info = adreno_gpu->info->a6xx;
>>>       struct a6xx_gpu *a6xx_gpu = to_a6xx_gpu(adreno_gpu);
>>>       struct a6xx_gmu *gmu = &a6xx_gpu->gmu;
>>>       u32 perf_index;
>>> +    u32 bw_index = 0;
>>>       unsigned long gpu_freq;
>>>       int ret = 0;
>>>   @@ -125,6 +127,37 @@ void a6xx_gmu_set_freq(struct msm_gpu *gpu,
>>> struct dev_pm_opp *opp,
>>>           if (gpu_freq == gmu->gpu_freqs[perf_index])
>>>               break;
>>>   +    /* If enabled, find the corresponding DDR bandwidth index */
>>> +    if (info->bcms && gmu->nr_gpu_bws > 1) {
>>
>> if (gmu->nr_gpu_bws)
> 
> gmu->nr_gpu_bws == 1 means there's not BW in the OPPs (index 0 is the
> "off" state)
> 
>>
>>> +        unsigned int bw = dev_pm_opp_get_bw(opp, true, 0);
>>> +
>>> +        for (bw_index = 0; bw_index < gmu->nr_gpu_bws - 1; bw_index+
>>> +) {
>>> +            if (bw == gmu->gpu_bw_table[bw_index])
>>> +                break;
>>> +        }
>>> +
>>> +        /* Vote AB as a fraction of the max bandwidth */
>>> +        if (bw) {
>>
>> This seems to only be introduced with certain a7xx too.. you should
>> ping the GMU with HFI_VALUE_GMU_AB_VOTE to check if it's supported
> 
> Good point

No no. Doing this will trigger some assert in pre-A750 gmu firmwares. We
learned it the hard way. No improvisation please. :)

-Akhil.

> 
>>
>>> +            u64 tmp;
>>> +
>>> +            /* For now, vote for 25% of the bandwidth */
>>> +            tmp = bw * 25;
>>> +            do_div(tmp, 100);
>>> +
>>> +            /*
>>> +             * The AB vote consists of a 16 bit wide quantized level
>>> +             * against the maximum supported bandwidth.
>>> +             * Quantization can be calculated as below:
>>> +             * vote = (bandwidth * 2^16) / max bandwidth
>>> +             */
>>> +            tmp *= MAX_AB_VOTE;
>>> +            do_div(tmp, gmu->gpu_bw_table[gmu->nr_gpu_bws - 1]);
>>> +
>>> +            bw_index |= AB_VOTE(clamp(tmp, 1, MAX_AB_VOTE));
>>> +            bw_index |= AB_VOTE_ENABLE;
>>> +        }
>>> +    }
>>
>> BTW, did you dump the values we send to the GMU here and in the RPMh
>> builder part & validate against downstream?
> 
> It matches how downstream builds the Ab vote yes
> 
>>
>> Konrad
>