[PATCH] drm/amd/pp: Revert gfx/compute profile switch sysfs
Felix Kuehling
felix.kuehling at amd.com
Wed Feb 28 17:13:58 UTC 2018
Hi Rex,
The purpose of the compute power profile for us is to optimize
performance. Compute workloads tend to be "bursty". The default
heuristics of the SMU don't work well for such work loads. They aren't
able to raise clocks quickly enough to accelerate short bursts of
compute work.
We need to quickly raise clocks when needed. That's why we change
hysteresis parameters, and set a minimum clock. This minimizes the
latency of reaching maximum clocks.
The goal is not to set a fixed frequency, but to minimize the time
needed to reach the highest frequency. Setting fixed DPM level 7 will
not allow us to reach level 8. That will hurt our performance.
Regards,
Felix
On 2018-02-28 06:08 AM, Zhu, Rex wrote:
> I went through the power profiling feature in windows - Auto Wattman supported on Polaris and Vega. Initially apply to Polaris.
>
> The Goal is to maximize performance/watt without impacting performance and no visual impact per system/use case
>
> On smu7, 3 parameters control how SMU responds to sclk/mclk activity, up/down hysteresis and activity threshold.
>
> My understanding is:
> When sclk_activity large than threshold, adjust uphyst/downhyst/activity threshold to go higher dpm level more faster or stay in higher dpm level longer for boost performance.
> Otherwise, for power saving.
>
> Those parameters were used to change the behavior of dpm switching for performance in watt. So we wish to add an independent sysfs to configure them.
>
> When this feature enabled, the performance in watt is improved in most case(low activity and Multimedia) from pplib's test. I do not find compute test result.
>
> So there is a question:
>
> When compute ring begin, if we just fix sclk in level7 directly, ,
> Is there any difference in performance or performance in watt, compared to set compute profile mode (include apply min sclk to level6)?
>
>
> Best Regards
> Rex
>
>
>
>
>
> -----Original Message-----
> From: Kuehling, Felix
> Sent: Wednesday, February 28, 2018 2:55 AM
> To: Alex Deucher; Huang, JinHuiEric
> Cc: Zhu, Rex; Russell, Kent; amd-gfx list
> Subject: Re: [PATCH] drm/amd/pp: Revert gfx/compute profile switch sysfs
>
> On 2018-02-27 11:27 AM, Alex Deucher wrote:
>> On Tue, Feb 27, 2018 at 11:22 AM, Eric Huang <jinhuieric.huang at amd.com> wrote:
>>> As I mentioned in code review for new power profile, old gfx/compute power
>>> profile have two scenarios for auto switching. One is
>>> gfx->compute(default)->gfx and other is gfx->compute(custom)->gfx. New power
>>> profile only satisfies first one, but in second one for user debugging, user
>>> setting of power profile will be over-written when switching to compute
>>> profile.
>> For debugging, the idea would be to select manual mode via
>> force_performance_level and then force a profile via sysfs. If manual
>> mode was selected, none of the automatic profile changing would
>> happen. Is that adequate or do you need the automatic switching even
>> with a custom profile?
> It would result in higher power consumption while no compute work is
> running by applying minimum clocks (if that's still part of the
> profile). If minimum clock is not part of the profile, then that's
> another major regression for us.
>
> I would prefer to add the ability to tweak all the power profiles so we
> can have workload-specific customization, not just for compute. Maybe
> that's something we can add later.
>
> Regards,
> Felix
>
>> Alex
>>
>>
>>> Regards,
>>> Eric
>>>
>>> On 2018-02-27 10:52 AM, Felix Kuehling wrote:
>>>> [+Eric]
>>>>
>>>> Compute profile switching code as well as KFD compute support for most
>>>> GPUs is not upstream yet. As such, there is probably no requirement
>>>> (yet) to keep the compute profile API stable, that we added specifically
>>>> for KFD. Once we are upstream that will change.
>>>>
>>>> If you change it now, we'll have to adapt, like we have often done.
>>>>
>>>> I'll let Eric comment on whether the new power profile code and API is
>>>> an adequate replacement from a functionality perspective. Eric, Kent,
>>>> have we done any testing with Rex's new profile code?
>>>>
>>>> Regards,
>>>> Felix
>>>>
>>>>
>>>> On 2018-02-27 10:41 AM, Alex Deucher wrote:
>>>>> + Kent and Felix for comment
>>>>>
>>>>> On Tue, Feb 27, 2018 at 6:21 AM, Rex Zhu <Rex.Zhu at amd.com> wrote:
>>>>>> The gfx/compute profiling mode switch is only for internally
>>>>>> test. Not a complete solution and unexpectly upstream.
>>>>>> so revert it.
>>>>>>
>>>>>> Change-Id: I1af1b64a63b6fc12c24cf73df03b083b3661ca02
>>>>>> Signed-off-by: Rex Zhu <Rex.Zhu at amd.com>
>>>>>> ---
>>>>>> drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h | 8 -
>>>>>> drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c | 180
>>>>>> ---------------
>>>>>> drivers/gpu/drm/amd/amdgpu/ci_dpm.c | 256
>>>>>> ---------------------
>>>>>> drivers/gpu/drm/amd/amdgpu/ci_dpm.h | 7 -
>>>>>> drivers/gpu/drm/amd/include/kgd_pp_interface.h | 7 -
>>>>>> drivers/gpu/drm/amd/powerplay/amd_powerplay.c | 114 ---------
>>>>>> .../gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c | 17 --
>>>>>> drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.c | 2 -
>>>>>> drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c | 84 -------
>>>>>> drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c | 86 -------
>>>>>> .../gpu/drm/amd/powerplay/inc/hardwaremanager.h | 1 -
>>>>>> drivers/gpu/drm/amd/powerplay/inc/hwmgr.h | 9 -
>>>>>> drivers/gpu/drm/amd/powerplay/inc/smumgr.h | 3 -
>>>>>> drivers/gpu/drm/amd/powerplay/smumgr/ci_smumgr.c | 27 ---
>>>>>> drivers/gpu/drm/amd/powerplay/smumgr/fiji_smumgr.c | 66 ------
>>>>>> .../drm/amd/powerplay/smumgr/polaris10_smumgr.c | 64 ------
>>>>>> drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c | 10 -
>>>>>> .../gpu/drm/amd/powerplay/smumgr/tonga_smumgr.c | 64 ------
>>>>>> 18 files changed, 1005 deletions(-)
>>>>>>
>>>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h
>>>>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h
>>>>>> index bd745a4..9c373f8f 100644
>>>>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h
>>>>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h
>>>>>> @@ -341,14 +341,6 @@ enum amdgpu_pcie_gen {
>>>>>>
>>>>>> ((adev)->powerplay.pp_funcs->reset_power_profile_state(\
>>>>>> (adev)->powerplay.pp_handle, request))
>>>>>>
>>>>>> -#define amdgpu_dpm_get_power_profile_state(adev, query) \
>>>>>> - ((adev)->powerplay.pp_funcs->get_power_profile_state(\
>>>>>> - (adev)->powerplay.pp_handle, query))
>>>>>> -
>>>>>> -#define amdgpu_dpm_set_power_profile_state(adev, request) \
>>>>>> - ((adev)->powerplay.pp_funcs->set_power_profile_state(\
>>>>>> - (adev)->powerplay.pp_handle, request))
>>>>>> -
>>>>>> #define amdgpu_dpm_switch_power_profile(adev, type) \
>>>>>> ((adev)->powerplay.pp_funcs->switch_power_profile(\
>>>>>> (adev)->powerplay.pp_handle, type))
>>>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c
>>>>>> b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c
>>>>>> index 9e73cbc..632b186 100644
>>>>>> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c
>>>>>> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c
>>>>>> @@ -734,161 +734,6 @@ static ssize_t
>>>>>> amdgpu_set_pp_power_profile_mode(struct device *dev,
>>>>>> return -EINVAL;
>>>>>> }
>>>>>>
>>>>>> -static ssize_t amdgpu_get_pp_power_profile(struct device *dev,
>>>>>> - char *buf, struct amd_pp_profile *query)
>>>>>> -{
>>>>>> - struct drm_device *ddev = dev_get_drvdata(dev);
>>>>>> - struct amdgpu_device *adev = ddev->dev_private;
>>>>>> - int ret = 0xff;
>>>>>> -
>>>>>> - if (adev->powerplay.pp_funcs->get_power_profile_state)
>>>>>> - ret = amdgpu_dpm_get_power_profile_state(
>>>>>> - adev, query);
>>>>>> -
>>>>>> - if (ret)
>>>>>> - return ret;
>>>>>> -
>>>>>> - return snprintf(buf, PAGE_SIZE,
>>>>>> - "%d %d %d %d %d\n",
>>>>>> - query->min_sclk / 100,
>>>>>> - query->min_mclk / 100,
>>>>>> - query->activity_threshold,
>>>>>> - query->up_hyst,
>>>>>> - query->down_hyst);
>>>>>> -}
>>>>>> -
>>>>>> -static ssize_t amdgpu_get_pp_gfx_power_profile(struct device *dev,
>>>>>> - struct device_attribute *attr,
>>>>>> - char *buf)
>>>>>> -{
>>>>>> - struct amd_pp_profile query = {0};
>>>>>> -
>>>>>> - query.type = AMD_PP_GFX_PROFILE;
>>>>>> -
>>>>>> - return amdgpu_get_pp_power_profile(dev, buf, &query);
>>>>>> -}
>>>>>> -
>>>>>> -static ssize_t amdgpu_get_pp_compute_power_profile(struct device *dev,
>>>>>> - struct device_attribute *attr,
>>>>>> - char *buf)
>>>>>> -{
>>>>>> - struct amd_pp_profile query = {0};
>>>>>> -
>>>>>> - query.type = AMD_PP_COMPUTE_PROFILE;
>>>>>> -
>>>>>> - return amdgpu_get_pp_power_profile(dev, buf, &query);
>>>>>> -}
>>>>>> -
>>>>>> -static ssize_t amdgpu_set_pp_power_profile(struct device *dev,
>>>>>> - const char *buf,
>>>>>> - size_t count,
>>>>>> - struct amd_pp_profile *request)
>>>>>> -{
>>>>>> - struct drm_device *ddev = dev_get_drvdata(dev);
>>>>>> - struct amdgpu_device *adev = ddev->dev_private;
>>>>>> - uint32_t loop = 0;
>>>>>> - char *sub_str, buf_cpy[128], *tmp_str;
>>>>>> - const char delimiter[3] = {' ', '\n', '\0'};
>>>>>> - long int value;
>>>>>> - int ret = 0xff;
>>>>>> -
>>>>>> - if (strncmp("reset", buf, strlen("reset")) == 0) {
>>>>>> - if (adev->powerplay.pp_funcs->reset_power_profile_state)
>>>>>> - ret = amdgpu_dpm_reset_power_profile_state(
>>>>>> - adev, request);
>>>>>> - if (ret) {
>>>>>> - count = -EINVAL;
>>>>>> - goto fail;
>>>>>> - }
>>>>>> - return count;
>>>>>> - }
>>>>>> -
>>>>>> - if (strncmp("set", buf, strlen("set")) == 0) {
>>>>>> - if (adev->powerplay.pp_funcs->set_power_profile_state)
>>>>>> - ret = amdgpu_dpm_set_power_profile_state(
>>>>>> - adev, request);
>>>>>> -
>>>>>> - if (ret) {
>>>>>> - count = -EINVAL;
>>>>>> - goto fail;
>>>>>> - }
>>>>>> - return count;
>>>>>> - }
>>>>>> -
>>>>>> - if (count + 1 >= 128) {
>>>>>> - count = -EINVAL;
>>>>>> - goto fail;
>>>>>> - }
>>>>>> -
>>>>>> - memcpy(buf_cpy, buf, count + 1);
>>>>>> - tmp_str = buf_cpy;
>>>>>> -
>>>>>> - while (tmp_str[0]) {
>>>>>> - sub_str = strsep(&tmp_str, delimiter);
>>>>>> - ret = kstrtol(sub_str, 0, &value);
>>>>>> - if (ret) {
>>>>>> - count = -EINVAL;
>>>>>> - goto fail;
>>>>>> - }
>>>>>> -
>>>>>> - switch (loop) {
>>>>>> - case 0:
>>>>>> - /* input unit MHz convert to dpm table unit
>>>>>> 10KHz*/
>>>>>> - request->min_sclk = (uint32_t)value * 100;
>>>>>> - break;
>>>>>> - case 1:
>>>>>> - /* input unit MHz convert to dpm table unit
>>>>>> 10KHz*/
>>>>>> - request->min_mclk = (uint32_t)value * 100;
>>>>>> - break;
>>>>>> - case 2:
>>>>>> - request->activity_threshold = (uint16_t)value;
>>>>>> - break;
>>>>>> - case 3:
>>>>>> - request->up_hyst = (uint8_t)value;
>>>>>> - break;
>>>>>> - case 4:
>>>>>> - request->down_hyst = (uint8_t)value;
>>>>>> - break;
>>>>>> - default:
>>>>>> - break;
>>>>>> - }
>>>>>> -
>>>>>> - loop++;
>>>>>> - }
>>>>>> - if (adev->powerplay.pp_funcs->set_power_profile_state)
>>>>>> - ret = amdgpu_dpm_set_power_profile_state(adev, request);
>>>>>> -
>>>>>> - if (ret)
>>>>>> - count = -EINVAL;
>>>>>> -
>>>>>> -fail:
>>>>>> - return count;
>>>>>> -}
>>>>>> -
>>>>>> -static ssize_t amdgpu_set_pp_gfx_power_profile(struct device *dev,
>>>>>> - struct device_attribute *attr,
>>>>>> - const char *buf,
>>>>>> - size_t count)
>>>>>> -{
>>>>>> - struct amd_pp_profile request = {0};
>>>>>> -
>>>>>> - request.type = AMD_PP_GFX_PROFILE;
>>>>>> -
>>>>>> - return amdgpu_set_pp_power_profile(dev, buf, count, &request);
>>>>>> -}
>>>>>> -
>>>>>> -static ssize_t amdgpu_set_pp_compute_power_profile(struct device *dev,
>>>>>> - struct device_attribute *attr,
>>>>>> - const char *buf,
>>>>>> - size_t count)
>>>>>> -{
>>>>>> - struct amd_pp_profile request = {0};
>>>>>> -
>>>>>> - request.type = AMD_PP_COMPUTE_PROFILE;
>>>>>> -
>>>>>> - return amdgpu_set_pp_power_profile(dev, buf, count, &request);
>>>>>> -}
>>>>>> -
>>>>>> static DEVICE_ATTR(power_dpm_state, S_IRUGO | S_IWUSR,
>>>>>> amdgpu_get_dpm_state, amdgpu_set_dpm_state);
>>>>>> static DEVICE_ATTR(power_dpm_force_performance_level, S_IRUGO |
>>>>>> S_IWUSR,
>>>>>> amdgpu_get_dpm_forced_performance_level,
>>>>>> @@ -916,12 +761,6 @@ static DEVICE_ATTR(pp_sclk_od, S_IRUGO | S_IWUSR,
>>>>>> static DEVICE_ATTR(pp_mclk_od, S_IRUGO | S_IWUSR,
>>>>>> amdgpu_get_pp_mclk_od,
>>>>>> amdgpu_set_pp_mclk_od);
>>>>>> -static DEVICE_ATTR(pp_gfx_power_profile, S_IRUGO | S_IWUSR,
>>>>>> - amdgpu_get_pp_gfx_power_profile,
>>>>>> - amdgpu_set_pp_gfx_power_profile);
>>>>>> -static DEVICE_ATTR(pp_compute_power_profile, S_IRUGO | S_IWUSR,
>>>>>> - amdgpu_get_pp_compute_power_profile,
>>>>>> - amdgpu_set_pp_compute_power_profile);
>>>>>> static DEVICE_ATTR(pp_power_profile_mode, S_IRUGO | S_IWUSR,
>>>>>> amdgpu_get_pp_power_profile_mode,
>>>>>> amdgpu_set_pp_power_profile_mode);
>>>>>> @@ -1767,21 +1606,6 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device
>>>>>> *adev)
>>>>>> return ret;
>>>>>> }
>>>>>> ret = device_create_file(adev->dev,
>>>>>> - &dev_attr_pp_gfx_power_profile);
>>>>>> - if (ret) {
>>>>>> - DRM_ERROR("failed to create device file "
>>>>>> - "pp_gfx_power_profile\n");
>>>>>> - return ret;
>>>>>> - }
>>>>>> - ret = device_create_file(adev->dev,
>>>>>> - &dev_attr_pp_compute_power_profile);
>>>>>> - if (ret) {
>>>>>> - DRM_ERROR("failed to create device file "
>>>>>> - "pp_compute_power_profile\n");
>>>>>> - return ret;
>>>>>> - }
>>>>>> -
>>>>>> - ret = device_create_file(adev->dev,
>>>>>> &dev_attr_pp_power_profile_mode);
>>>>>> if (ret) {
>>>>>> DRM_ERROR("failed to create device file "
>>>>>> @@ -1827,10 +1651,6 @@ void amdgpu_pm_sysfs_fini(struct amdgpu_device
>>>>>> *adev)
>>>>>> device_remove_file(adev->dev, &dev_attr_pp_sclk_od);
>>>>>> device_remove_file(adev->dev, &dev_attr_pp_mclk_od);
>>>>>> device_remove_file(adev->dev,
>>>>>> - &dev_attr_pp_gfx_power_profile);
>>>>>> - device_remove_file(adev->dev,
>>>>>> - &dev_attr_pp_compute_power_profile);
>>>>>> - device_remove_file(adev->dev,
>>>>>> &dev_attr_pp_power_profile_mode);
>>>>>> device_remove_file(adev->dev,
>>>>>> &dev_attr_pp_od_clk_voltage);
>>>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c
>>>>>> b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c
>>>>>> index f82f40f..ddb814f 100644
>>>>>> --- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c
>>>>>> +++ b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c
>>>>>> @@ -3695,40 +3695,6 @@ static int ci_find_boot_level(struct
>>>>>> ci_single_dpm_table *table,
>>>>>> return ret;
>>>>>> }
>>>>>>
>>>>>> -static void ci_save_default_power_profile(struct amdgpu_device *adev)
>>>>>> -{
>>>>>> - struct ci_power_info *pi = ci_get_pi(adev);
>>>>>> - struct SMU7_Discrete_GraphicsLevel *levels =
>>>>>> - pi->smc_state_table.GraphicsLevel;
>>>>>> - uint32_t min_level = 0;
>>>>>> -
>>>>>> - pi->default_gfx_power_profile.activity_threshold =
>>>>>> - be16_to_cpu(levels[0].ActivityLevel);
>>>>>> - pi->default_gfx_power_profile.up_hyst = levels[0].UpH;
>>>>>> - pi->default_gfx_power_profile.down_hyst = levels[0].DownH;
>>>>>> - pi->default_gfx_power_profile.type = AMD_PP_GFX_PROFILE;
>>>>>> -
>>>>>> - pi->default_compute_power_profile =
>>>>>> pi->default_gfx_power_profile;
>>>>>> - pi->default_compute_power_profile.type = AMD_PP_COMPUTE_PROFILE;
>>>>>> -
>>>>>> - /* Optimize compute power profile: Use only highest
>>>>>> - * 2 power levels (if more than 2 are available), Hysteresis:
>>>>>> - * 0ms up, 5ms down
>>>>>> - */
>>>>>> - if (pi->smc_state_table.GraphicsDpmLevelCount > 2)
>>>>>> - min_level = pi->smc_state_table.GraphicsDpmLevelCount -
>>>>>> 2;
>>>>>> - else if (pi->smc_state_table.GraphicsDpmLevelCount == 2)
>>>>>> - min_level = 1;
>>>>>> - pi->default_compute_power_profile.min_sclk =
>>>>>> - be32_to_cpu(levels[min_level].SclkFrequency);
>>>>>> -
>>>>>> - pi->default_compute_power_profile.up_hyst = 0;
>>>>>> - pi->default_compute_power_profile.down_hyst = 5;
>>>>>> -
>>>>>> - pi->gfx_power_profile = pi->default_gfx_power_profile;
>>>>>> - pi->compute_power_profile = pi->default_compute_power_profile;
>>>>>> -}
>>>>>> -
>>>>>> static int ci_init_smc_table(struct amdgpu_device *adev)
>>>>>> {
>>>>>> struct ci_power_info *pi = ci_get_pi(adev);
>>>>>> @@ -3874,8 +3840,6 @@ static int ci_init_smc_table(struct amdgpu_device
>>>>>> *adev)
>>>>>> if (ret)
>>>>>> return ret;
>>>>>>
>>>>>> - ci_save_default_power_profile(adev);
>>>>>> -
>>>>>> return 0;
>>>>>> }
>>>>>>
>>>>>> @@ -6753,222 +6717,6 @@ static int ci_dpm_set_mclk_od(void *handle,
>>>>>> uint32_t value)
>>>>>> return 0;
>>>>>> }
>>>>>>
>>>>>> -static int ci_dpm_get_power_profile_state(void *handle,
>>>>>> - struct amd_pp_profile *query)
>>>>>> -{
>>>>>> - struct amdgpu_device *adev = (struct amdgpu_device *)handle;
>>>>>> - struct ci_power_info *pi = ci_get_pi(adev);
>>>>>> -
>>>>>> - if (!pi || !query)
>>>>>> - return -EINVAL;
>>>>>> -
>>>>>> - if (query->type == AMD_PP_GFX_PROFILE)
>>>>>> - memcpy(query, &pi->gfx_power_profile,
>>>>>> - sizeof(struct amd_pp_profile));
>>>>>> - else if (query->type == AMD_PP_COMPUTE_PROFILE)
>>>>>> - memcpy(query, &pi->compute_power_profile,
>>>>>> - sizeof(struct amd_pp_profile));
>>>>>> - else
>>>>>> - return -EINVAL;
>>>>>> -
>>>>>> - return 0;
>>>>>> -}
>>>>>> -
>>>>>> -static int ci_populate_requested_graphic_levels(struct amdgpu_device
>>>>>> *adev,
>>>>>> - struct amd_pp_profile *request)
>>>>>> -{
>>>>>> - struct ci_power_info *pi = ci_get_pi(adev);
>>>>>> - struct ci_dpm_table *dpm_table = &(pi->dpm_table);
>>>>>> - struct SMU7_Discrete_GraphicsLevel *levels =
>>>>>> - pi->smc_state_table.GraphicsLevel;
>>>>>> - uint32_t array = pi->dpm_table_start +
>>>>>> - offsetof(SMU7_Discrete_DpmTable, GraphicsLevel);
>>>>>> - uint32_t array_size = sizeof(struct SMU7_Discrete_GraphicsLevel)
>>>>>> *
>>>>>> - SMU7_MAX_LEVELS_GRAPHICS;
>>>>>> - uint32_t i;
>>>>>> -
>>>>>> - for (i = 0; i < dpm_table->sclk_table.count; i++) {
>>>>>> - levels[i].ActivityLevel =
>>>>>> -
>>>>>> cpu_to_be16(request->activity_threshold);
>>>>>> - levels[i].EnabledForActivity = 1;
>>>>>> - levels[i].UpH = request->up_hyst;
>>>>>> - levels[i].DownH = request->down_hyst;
>>>>>> - }
>>>>>> -
>>>>>> - return amdgpu_ci_copy_bytes_to_smc(adev, array, (uint8_t
>>>>>> *)levels,
>>>>>> - array_size, pi->sram_end);
>>>>>> -}
>>>>>> -
>>>>>> -static void ci_find_min_clock_masks(struct amdgpu_device *adev,
>>>>>> - uint32_t *sclk_mask, uint32_t *mclk_mask,
>>>>>> - uint32_t min_sclk, uint32_t min_mclk)
>>>>>> -{
>>>>>> - struct ci_power_info *pi = ci_get_pi(adev);
>>>>>> - struct ci_dpm_table *dpm_table = &(pi->dpm_table);
>>>>>> - uint32_t i;
>>>>>> -
>>>>>> - for (i = 0; i < dpm_table->sclk_table.count; i++) {
>>>>>> - if (dpm_table->sclk_table.dpm_levels[i].enabled &&
>>>>>> - dpm_table->sclk_table.dpm_levels[i].value >=
>>>>>> min_sclk)
>>>>>> - *sclk_mask |= 1 << i;
>>>>>> - }
>>>>>> -
>>>>>> - for (i = 0; i < dpm_table->mclk_table.count; i++) {
>>>>>> - if (dpm_table->mclk_table.dpm_levels[i].enabled &&
>>>>>> - dpm_table->mclk_table.dpm_levels[i].value >=
>>>>>> min_mclk)
>>>>>> - *mclk_mask |= 1 << i;
>>>>>> - }
>>>>>> -}
>>>>>> -
>>>>>> -static int ci_set_power_profile_state(struct amdgpu_device *adev,
>>>>>> - struct amd_pp_profile *request)
>>>>>> -{
>>>>>> - struct ci_power_info *pi = ci_get_pi(adev);
>>>>>> - int tmp_result, result = 0;
>>>>>> - uint32_t sclk_mask = 0, mclk_mask = 0;
>>>>>> -
>>>>>> - tmp_result = ci_freeze_sclk_mclk_dpm(adev);
>>>>>> - if (tmp_result) {
>>>>>> - DRM_ERROR("Failed to freeze SCLK MCLK DPM!");
>>>>>> - result = tmp_result;
>>>>>> - }
>>>>>> -
>>>>>> - tmp_result = ci_populate_requested_graphic_levels(adev,
>>>>>> - request);
>>>>>> - if (tmp_result) {
>>>>>> - DRM_ERROR("Failed to populate requested graphic
>>>>>> levels!");
>>>>>> - result = tmp_result;
>>>>>> - }
>>>>>> -
>>>>>> - tmp_result = ci_unfreeze_sclk_mclk_dpm(adev);
>>>>>> - if (tmp_result) {
>>>>>> - DRM_ERROR("Failed to unfreeze SCLK MCLK DPM!");
>>>>>> - result = tmp_result;
>>>>>> - }
>>>>>> -
>>>>>> - ci_find_min_clock_masks(adev, &sclk_mask, &mclk_mask,
>>>>>> - request->min_sclk, request->min_mclk);
>>>>>> -
>>>>>> - if (sclk_mask) {
>>>>>> - if (!pi->sclk_dpm_key_disabled)
>>>>>> - amdgpu_ci_send_msg_to_smc_with_parameter(
>>>>>> - adev,
>>>>>> - PPSMC_MSG_SCLKDPM_SetEnabledMask,
>>>>>> - pi->dpm_level_enable_mask.
>>>>>> - sclk_dpm_enable_mask &
>>>>>> - sclk_mask);
>>>>>> - }
>>>>>> -
>>>>>> - if (mclk_mask) {
>>>>>> - if (!pi->mclk_dpm_key_disabled)
>>>>>> - amdgpu_ci_send_msg_to_smc_with_parameter(
>>>>>> - adev,
>>>>>> - PPSMC_MSG_MCLKDPM_SetEnabledMask,
>>>>>> - pi->dpm_level_enable_mask.
>>>>>> - mclk_dpm_enable_mask &
>>>>>> - mclk_mask);
>>>>>> - }
>>>>>> -
>>>>>> -
>>>>>> - return result;
>>>>>> -}
>>>>>> -
>>>>>> -static int ci_dpm_set_power_profile_state(void *handle,
>>>>>> - struct amd_pp_profile *request)
>>>>>> -{
>>>>>> - struct amdgpu_device *adev = (struct amdgpu_device *)handle;
>>>>>> - struct ci_power_info *pi = ci_get_pi(adev);
>>>>>> - int ret = -1;
>>>>>> -
>>>>>> - if (!pi || !request)
>>>>>> - return -EINVAL;
>>>>>> -
>>>>>> - if (adev->pm.dpm.forced_level !=
>>>>>> - AMD_DPM_FORCED_LEVEL_AUTO)
>>>>>> - return -EINVAL;
>>>>>> -
>>>>>> - if (request->min_sclk ||
>>>>>> - request->min_mclk ||
>>>>>> - request->activity_threshold ||
>>>>>> - request->up_hyst ||
>>>>>> - request->down_hyst) {
>>>>>> - if (request->type == AMD_PP_GFX_PROFILE)
>>>>>> - memcpy(&pi->gfx_power_profile, request,
>>>>>> - sizeof(struct amd_pp_profile));
>>>>>> - else if (request->type == AMD_PP_COMPUTE_PROFILE)
>>>>>> - memcpy(&pi->compute_power_profile, request,
>>>>>> - sizeof(struct amd_pp_profile));
>>>>>> - else
>>>>>> - return -EINVAL;
>>>>>> -
>>>>>> - if (request->type == pi->current_power_profile)
>>>>>> - ret = ci_set_power_profile_state(
>>>>>> - adev,
>>>>>> - request);
>>>>>> - } else {
>>>>>> - /* set power profile if it exists */
>>>>>> - switch (request->type) {
>>>>>> - case AMD_PP_GFX_PROFILE:
>>>>>> - ret = ci_set_power_profile_state(
>>>>>> - adev,
>>>>>> - &pi->gfx_power_profile);
>>>>>> - break;
>>>>>> - case AMD_PP_COMPUTE_PROFILE:
>>>>>> - ret = ci_set_power_profile_state(
>>>>>> - adev,
>>>>>> - &pi->compute_power_profile);
>>>>>> - break;
>>>>>> - default:
>>>>>> - return -EINVAL;
>>>>>> - }
>>>>>> - }
>>>>>> -
>>>>>> - if (!ret)
>>>>>> - pi->current_power_profile = request->type;
>>>>>> -
>>>>>> - return 0;
>>>>>> -}
>>>>>> -
>>>>>> -static int ci_dpm_reset_power_profile_state(void *handle,
>>>>>> - struct amd_pp_profile *request)
>>>>>> -{
>>>>>> - struct amdgpu_device *adev = (struct amdgpu_device *)handle;
>>>>>> - struct ci_power_info *pi = ci_get_pi(adev);
>>>>>> -
>>>>>> - if (!pi || !request)
>>>>>> - return -EINVAL;
>>>>>> -
>>>>>> - if (request->type == AMD_PP_GFX_PROFILE) {
>>>>>> - pi->gfx_power_profile = pi->default_gfx_power_profile;
>>>>>> - return ci_dpm_set_power_profile_state(adev,
>>>>>> - &pi->gfx_power_profile);
>>>>>> - } else if (request->type == AMD_PP_COMPUTE_PROFILE) {
>>>>>> - pi->compute_power_profile =
>>>>>> - pi->default_compute_power_profile;
>>>>>> - return ci_dpm_set_power_profile_state(adev,
>>>>>> - &pi->compute_power_profile);
>>>>>> - } else
>>>>>> - return -EINVAL;
>>>>>> -}
>>>>>> -
>>>>>> -static int ci_dpm_switch_power_profile(void *handle,
>>>>>> - enum amd_pp_profile_type type)
>>>>>> -{
>>>>>> - struct amdgpu_device *adev = (struct amdgpu_device *)handle;
>>>>>> - struct ci_power_info *pi = ci_get_pi(adev);
>>>>>> - struct amd_pp_profile request = {0};
>>>>>> -
>>>>>> - if (!pi)
>>>>>> - return -EINVAL;
>>>>>> -
>>>>>> - if (pi->current_power_profile != type) {
>>>>>> - request.type = type;
>>>>>> - return ci_dpm_set_power_profile_state(adev, &request);
>>>>>> - }
>>>>>> -
>>>>>> - return 0;
>>>>>> -}
>>>>>> -
>>>>>> static int ci_dpm_read_sensor(void *handle, int idx,
>>>>>> void *value, int *size)
>>>>>> {
>>>>>> @@ -7053,10 +6801,6 @@ static int ci_dpm_read_sensor(void *handle, int
>>>>>> idx,
>>>>>> .set_mclk_od = ci_dpm_set_mclk_od,
>>>>>> .check_state_equal = ci_check_state_equal,
>>>>>> .get_vce_clock_state = amdgpu_get_vce_clock_state,
>>>>>> - .get_power_profile_state = ci_dpm_get_power_profile_state,
>>>>>> - .set_power_profile_state = ci_dpm_set_power_profile_state,
>>>>>> - .reset_power_profile_state = ci_dpm_reset_power_profile_state,
>>>>>> - .switch_power_profile = ci_dpm_switch_power_profile,
>>>>>> .read_sensor = ci_dpm_read_sensor,
>>>>>> };
>>>>>>
>>>>>> diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.h
>>>>>> b/drivers/gpu/drm/amd/amdgpu/ci_dpm.h
>>>>>> index 84cbc9c..91be299 100644
>>>>>> --- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.h
>>>>>> +++ b/drivers/gpu/drm/amd/amdgpu/ci_dpm.h
>>>>>> @@ -295,13 +295,6 @@ struct ci_power_info {
>>>>>> bool fan_is_controlled_by_smc;
>>>>>> u32 t_min;
>>>>>> u32 fan_ctrl_default_mode;
>>>>>> -
>>>>>> - /* power profile */
>>>>>> - struct amd_pp_profile gfx_power_profile;
>>>>>> - struct amd_pp_profile compute_power_profile;
>>>>>> - struct amd_pp_profile default_gfx_power_profile;
>>>>>> - struct amd_pp_profile default_compute_power_profile;
>>>>>> - enum amd_pp_profile_type current_power_profile;
>>>>>> };
>>>>>>
>>>>>> #define CISLANDS_VOLTAGE_CONTROL_NONE 0x0
>>>>>> diff --git a/drivers/gpu/drm/amd/include/kgd_pp_interface.h
>>>>>> b/drivers/gpu/drm/amd/include/kgd_pp_interface.h
>>>>>> index 22c2fa3..7dfba2d 100644
>>>>>> --- a/drivers/gpu/drm/amd/include/kgd_pp_interface.h
>>>>>> +++ b/drivers/gpu/drm/amd/include/kgd_pp_interface.h
>>>>>> @@ -260,13 +260,6 @@ struct amd_pm_funcs {
>>>>>> int (*get_pp_table)(void *handle, char **table);
>>>>>> int (*set_pp_table)(void *handle, const char *buf, size_t
>>>>>> size);
>>>>>> void (*debugfs_print_current_performance_level)(void *handle,
>>>>>> struct seq_file *m);
>>>>>> -
>>>>>> - int (*reset_power_profile_state)(void *handle,
>>>>>> - struct amd_pp_profile *request);
>>>>>> - int (*get_power_profile_state)(void *handle,
>>>>>> - struct amd_pp_profile *query);
>>>>>> - int (*set_power_profile_state)(void *handle,
>>>>>> - struct amd_pp_profile *request);
>>>>>> int (*switch_power_profile)(void *handle,
>>>>>> enum amd_pp_profile_type type);
>>>>>> /* export to amdgpu */
>>>>>> diff --git a/drivers/gpu/drm/amd/powerplay/amd_powerplay.c
>>>>>> b/drivers/gpu/drm/amd/powerplay/amd_powerplay.c
>>>>>> index 5feb91b..5f5fbb6 100644
>>>>>> --- a/drivers/gpu/drm/amd/powerplay/amd_powerplay.c
>>>>>> +++ b/drivers/gpu/drm/amd/powerplay/amd_powerplay.c
>>>>>> @@ -1014,58 +1014,6 @@ static int pp_dpm_read_sensor(void *handle, int
>>>>>> idx,
>>>>>> return NULL;
>>>>>> }
>>>>>>
>>>>>> -static int pp_dpm_reset_power_profile_state(void *handle,
>>>>>> - struct amd_pp_profile *request)
>>>>>> -{
>>>>>> - struct pp_hwmgr *hwmgr;
>>>>>> - struct pp_instance *pp_handle = (struct pp_instance *)handle;
>>>>>> -
>>>>>> - if (!request || pp_check(pp_handle))
>>>>>> - return -EINVAL;
>>>>>> -
>>>>>> - hwmgr = pp_handle->hwmgr;
>>>>>> -
>>>>>> - if (hwmgr->hwmgr_func->set_power_profile_state == NULL) {
>>>>>> - pr_info("%s was not implemented.\n", __func__);
>>>>>> - return 0;
>>>>>> - }
>>>>>> -
>>>>>> - if (request->type == AMD_PP_GFX_PROFILE) {
>>>>>> - hwmgr->gfx_power_profile =
>>>>>> hwmgr->default_gfx_power_profile;
>>>>>> - return hwmgr->hwmgr_func->set_power_profile_state(hwmgr,
>>>>>> - &hwmgr->gfx_power_profile);
>>>>>> - } else if (request->type == AMD_PP_COMPUTE_PROFILE) {
>>>>>> - hwmgr->compute_power_profile =
>>>>>> - hwmgr->default_compute_power_profile;
>>>>>> - return hwmgr->hwmgr_func->set_power_profile_state(hwmgr,
>>>>>> - &hwmgr->compute_power_profile);
>>>>>> - } else
>>>>>> - return -EINVAL;
>>>>>> -}
>>>>>> -
>>>>>> -static int pp_dpm_get_power_profile_state(void *handle,
>>>>>> - struct amd_pp_profile *query)
>>>>>> -{
>>>>>> - struct pp_hwmgr *hwmgr;
>>>>>> - struct pp_instance *pp_handle = (struct pp_instance *)handle;
>>>>>> -
>>>>>> - if (!query || pp_check(pp_handle))
>>>>>> - return -EINVAL;
>>>>>> -
>>>>>> - hwmgr = pp_handle->hwmgr;
>>>>>> -
>>>>>> - if (query->type == AMD_PP_GFX_PROFILE)
>>>>>> - memcpy(query, &hwmgr->gfx_power_profile,
>>>>>> - sizeof(struct amd_pp_profile));
>>>>>> - else if (query->type == AMD_PP_COMPUTE_PROFILE)
>>>>>> - memcpy(query, &hwmgr->compute_power_profile,
>>>>>> - sizeof(struct amd_pp_profile));
>>>>>> - else
>>>>>> - return -EINVAL;
>>>>>> -
>>>>>> - return 0;
>>>>>> -}
>>>>>> -
>>>>>> static int pp_get_power_profile_mode(void *handle, char *buf)
>>>>>> {
>>>>>> struct pp_hwmgr *hwmgr;
>>>>>> @@ -1124,65 +1072,6 @@ static int pp_odn_edit_dpm_table(void *handle,
>>>>>> uint32_t type, long *input, uint3
>>>>>> return hwmgr->hwmgr_func->odn_edit_dpm_table(hwmgr, type,
>>>>>> input, size);
>>>>>> }
>>>>>>
>>>>>> -static int pp_dpm_set_power_profile_state(void *handle,
>>>>>> - struct amd_pp_profile *request)
>>>>>> -{
>>>>>> - struct pp_hwmgr *hwmgr;
>>>>>> - struct pp_instance *pp_handle = (struct pp_instance *)handle;
>>>>>> - int ret = -1;
>>>>>> -
>>>>>> - if (!request || pp_check(pp_handle))
>>>>>> - return -EINVAL;
>>>>>> -
>>>>>> - hwmgr = pp_handle->hwmgr;
>>>>>> -
>>>>>> - if (hwmgr->hwmgr_func->set_power_profile_state == NULL) {
>>>>>> - pr_info("%s was not implemented.\n", __func__);
>>>>>> - return 0;
>>>>>> - }
>>>>>> -
>>>>>> - if (request->min_sclk ||
>>>>>> - request->min_mclk ||
>>>>>> - request->activity_threshold ||
>>>>>> - request->up_hyst ||
>>>>>> - request->down_hyst) {
>>>>>> - if (request->type == AMD_PP_GFX_PROFILE)
>>>>>> - memcpy(&hwmgr->gfx_power_profile, request,
>>>>>> - sizeof(struct amd_pp_profile));
>>>>>> - else if (request->type == AMD_PP_COMPUTE_PROFILE)
>>>>>> - memcpy(&hwmgr->compute_power_profile, request,
>>>>>> - sizeof(struct amd_pp_profile));
>>>>>> - else
>>>>>> - return -EINVAL;
>>>>>> -
>>>>>> - if (request->type == hwmgr->current_power_profile)
>>>>>> - ret =
>>>>>> hwmgr->hwmgr_func->set_power_profile_state(
>>>>>> - hwmgr,
>>>>>> - request);
>>>>>> - } else {
>>>>>> - /* set power profile if it exists */
>>>>>> - switch (request->type) {
>>>>>> - case AMD_PP_GFX_PROFILE:
>>>>>> - ret =
>>>>>> hwmgr->hwmgr_func->set_power_profile_state(
>>>>>> - hwmgr,
>>>>>> - &hwmgr->gfx_power_profile);
>>>>>> - break;
>>>>>> - case AMD_PP_COMPUTE_PROFILE:
>>>>>> - ret =
>>>>>> hwmgr->hwmgr_func->set_power_profile_state(
>>>>>> - hwmgr,
>>>>>> - &hwmgr->compute_power_profile);
>>>>>> - break;
>>>>>> - default:
>>>>>> - return -EINVAL;
>>>>>> - }
>>>>>> - }
>>>>>> -
>>>>>> - if (!ret)
>>>>>> - hwmgr->current_power_profile = request->type;
>>>>>> -
>>>>>> - return 0;
>>>>>> -}
>>>>>> -
>>>>>> static int pp_dpm_switch_power_profile(void *handle,
>>>>>> enum amd_pp_profile_type type)
>>>>>> {
>>>>>> @@ -1590,9 +1479,6 @@ static int pp_set_mmhub_powergating_by_smu(void
>>>>>> *handle)
>>>>>> .set_mclk_od = pp_dpm_set_mclk_od,
>>>>>> .read_sensor = pp_dpm_read_sensor,
>>>>>> .get_vce_clock_state = pp_dpm_get_vce_clock_state,
>>>>>> - .reset_power_profile_state = pp_dpm_reset_power_profile_state,
>>>>>> - .get_power_profile_state = pp_dpm_get_power_profile_state,
>>>>>> - .set_power_profile_state = pp_dpm_set_power_profile_state,
>>>>>> .switch_power_profile = pp_dpm_switch_power_profile,
>>>>>> .set_clockgating_by_smu = pp_set_clockgating_by_smu,
>>>>>> .notify_smu_memory_info = pp_dpm_notify_smu_memory_info,
>>>>>> diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c
>>>>>> b/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c
>>>>>> index 33480de..f06f8f4 100644
>>>>>> --- a/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c
>>>>>> +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c
>>>>>> @@ -128,23 +128,6 @@ int phm_force_dpm_levels(struct pp_hwmgr *hwmgr,
>>>>>> enum amd_dpm_forced_level level
>>>>>> return ret;
>>>>>> }
>>>>>>
>>>>>> -int phm_reset_power_profile_state(struct pp_hwmgr *hwmgr)
>>>>>> -{
>>>>>> - int ret = 0;
>>>>>> -
>>>>>> - if (hwmgr->hwmgr_func->set_power_profile_state) {
>>>>>> - if (hwmgr->current_power_profile == AMD_PP_GFX_PROFILE)
>>>>>> - ret =
>>>>>> hwmgr->hwmgr_func->set_power_profile_state(
>>>>>> - hwmgr,
>>>>>> - &hwmgr->gfx_power_profile);
>>>>>> - else if (hwmgr->current_power_profile ==
>>>>>> AMD_PP_COMPUTE_PROFILE)
>>>>>> - ret =
>>>>>> hwmgr->hwmgr_func->set_power_profile_state(
>>>>>> - hwmgr,
>>>>>> - &hwmgr->compute_power_profile);
>>>>>> - }
>>>>>> - return ret;
>>>>>> -}
>>>>>> -
>>>>>> int phm_apply_state_adjust_rules(struct pp_hwmgr *hwmgr,
>>>>>> struct pp_power_state *adjusted_ps,
>>>>>> const struct pp_power_state *current_ps)
>>>>>> diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.c
>>>>>> b/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.c
>>>>>> index 95ab772..ed3bd15 100644
>>>>>> --- a/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.c
>>>>>> +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.c
>>>>>> @@ -247,8 +247,6 @@ int psm_adjust_power_state_dynamic(struct pp_hwmgr
>>>>>> *hwmgr, bool skip,
>>>>>> if (!phm_force_dpm_levels(hwmgr, hwmgr->request_dpm_level))
>>>>>> hwmgr->dpm_level = hwmgr->request_dpm_level;
>>>>>>
>>>>>> - phm_reset_power_profile_state(hwmgr);
>>>>>> -
>>>>>> return 0;
>>>>>> }
>>>>>>
>>>>>> diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c
>>>>>> b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c
>>>>>> index f09c76a..cb75c4f 100644
>>>>>> --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c
>>>>>> +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c
>>>>>> @@ -4622,89 +4622,6 @@ static int smu7_get_clock_by_type(struct pp_hwmgr
>>>>>> *hwmgr, enum amd_pp_clock_type
>>>>>> return 0;
>>>>>> }
>>>>>>
>>>>>> -static void smu7_find_min_clock_masks(struct pp_hwmgr *hwmgr,
>>>>>> - uint32_t *sclk_mask, uint32_t *mclk_mask,
>>>>>> - uint32_t min_sclk, uint32_t min_mclk)
>>>>>> -{
>>>>>> - struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend);
>>>>>> - struct smu7_dpm_table *dpm_table = &(data->dpm_table);
>>>>>> - uint32_t i;
>>>>>> -
>>>>>> - for (i = 0; i < dpm_table->sclk_table.count; i++) {
>>>>>> - if (dpm_table->sclk_table.dpm_levels[i].enabled &&
>>>>>> - dpm_table->sclk_table.dpm_levels[i].value >=
>>>>>> min_sclk)
>>>>>> - *sclk_mask |= 1 << i;
>>>>>> - }
>>>>>> -
>>>>>> - for (i = 0; i < dpm_table->mclk_table.count; i++) {
>>>>>> - if (dpm_table->mclk_table.dpm_levels[i].enabled &&
>>>>>> - dpm_table->mclk_table.dpm_levels[i].value >=
>>>>>> min_mclk)
>>>>>> - *mclk_mask |= 1 << i;
>>>>>> - }
>>>>>> -}
>>>>>> -
>>>>>> -static int smu7_set_power_profile_state(struct pp_hwmgr *hwmgr,
>>>>>> - struct amd_pp_profile *request)
>>>>>> -{
>>>>>> - struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend);
>>>>>> - int tmp_result, result = 0;
>>>>>> - uint32_t sclk_mask = 0, mclk_mask = 0;
>>>>>> -
>>>>>> - if (hwmgr->chip_id == CHIP_FIJI) {
>>>>>> - if (request->type == AMD_PP_GFX_PROFILE)
>>>>>> - smu7_enable_power_containment(hwmgr);
>>>>>> - else if (request->type == AMD_PP_COMPUTE_PROFILE)
>>>>>> - smu7_disable_power_containment(hwmgr);
>>>>>> - }
>>>>>> -
>>>>>> - if (hwmgr->dpm_level != AMD_DPM_FORCED_LEVEL_AUTO)
>>>>>> - return -EINVAL;
>>>>>> -
>>>>>> - if (smum_is_dpm_running(hwmgr)) {
>>>>>> - if (!data->sclk_dpm_key_disabled)
>>>>>> - smum_send_msg_to_smc(hwmgr,
>>>>>> PPSMC_MSG_SCLKDPM_FreezeLevel);
>>>>>> -
>>>>>> - if (!data->mclk_dpm_key_disabled)
>>>>>> - smum_send_msg_to_smc(hwmgr,
>>>>>> PPSMC_MSG_MCLKDPM_FreezeLevel);
>>>>>> - }
>>>>>> -
>>>>>> - tmp_result = smum_populate_requested_graphic_levels(hwmgr,
>>>>>> request);
>>>>>> - PP_ASSERT_WITH_CODE(!tmp_result,
>>>>>> - "Failed to populate requested graphic levels!",
>>>>>> - result = tmp_result);
>>>>>> -
>>>>>> - if (smum_is_dpm_running(hwmgr)) {
>>>>>> - if (!data->sclk_dpm_key_disabled)
>>>>>> - smum_send_msg_to_smc(hwmgr,
>>>>>> PPSMC_MSG_SCLKDPM_UnfreezeLevel);
>>>>>> -
>>>>>> - if (!data->mclk_dpm_key_disabled)
>>>>>> - smum_send_msg_to_smc(hwmgr,
>>>>>> PPSMC_MSG_MCLKDPM_UnfreezeLevel);
>>>>>> - }
>>>>>> -
>>>>>> - smu7_find_min_clock_masks(hwmgr, &sclk_mask, &mclk_mask,
>>>>>> - request->min_sclk, request->min_mclk);
>>>>>> -
>>>>>> - if (sclk_mask) {
>>>>>> - if (!data->sclk_dpm_key_disabled)
>>>>>> - smum_send_msg_to_smc_with_parameter(hwmgr,
>>>>>> - PPSMC_MSG_SCLKDPM_SetEnabledMask,
>>>>>> - data->dpm_level_enable_mask.
>>>>>> - sclk_dpm_enable_mask &
>>>>>> - sclk_mask);
>>>>>> - }
>>>>>> -
>>>>>> - if (mclk_mask) {
>>>>>> - if (!data->mclk_dpm_key_disabled)
>>>>>> - smum_send_msg_to_smc_with_parameter(hwmgr,
>>>>>> - PPSMC_MSG_MCLKDPM_SetEnabledMask,
>>>>>> - data->dpm_level_enable_mask.
>>>>>> - mclk_dpm_enable_mask &
>>>>>> - mclk_mask);
>>>>>> - }
>>>>>> -
>>>>>> - return result;
>>>>>> -}
>>>>>> -
>>>>>> static int smu7_notify_cac_buffer_info(struct pp_hwmgr *hwmgr,
>>>>>> uint32_t virtual_addr_low,
>>>>>> uint32_t virtual_addr_hi,
>>>>>> @@ -5154,7 +5071,6 @@ static int smu7_set_power_profile_mode(struct
>>>>>> pp_hwmgr *hwmgr, long *input, uint
>>>>>> .get_clock_by_type = smu7_get_clock_by_type,
>>>>>> .read_sensor = smu7_read_sensor,
>>>>>> .dynamic_state_management_disable = smu7_disable_dpm_tasks,
>>>>>> - .set_power_profile_state = smu7_set_power_profile_state,
>>>>>> .avfs_control = smu7_avfs_control,
>>>>>> .disable_smc_firmware_ctf = smu7_thermal_disable_alert,
>>>>>> .start_thermal_controller = smu7_start_thermal_controller,
>>>>>> diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c
>>>>>> b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c
>>>>>> index f5df20a..1416d2a 100644
>>>>>> --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c
>>>>>> +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c
>>>>>> @@ -2405,34 +2405,6 @@ static int
>>>>>> vega10_populate_and_upload_avfs_fuse_override(struct pp_hwmgr *hwmgr)
>>>>>> return result;
>>>>>> }
>>>>>>
>>>>>> -static int vega10_save_default_power_profile(struct pp_hwmgr *hwmgr)
>>>>>> -{
>>>>>> - struct vega10_hwmgr *data = (struct vega10_hwmgr
>>>>>> *)(hwmgr->backend);
>>>>>> - struct vega10_single_dpm_table *dpm_table =
>>>>>> &(data->dpm_table.gfx_table);
>>>>>> - uint32_t min_level;
>>>>>> -
>>>>>> - hwmgr->default_gfx_power_profile.type = AMD_PP_GFX_PROFILE;
>>>>>> - hwmgr->default_compute_power_profile.type =
>>>>>> AMD_PP_COMPUTE_PROFILE;
>>>>>> -
>>>>>> - /* Optimize compute power profile: Use only highest
>>>>>> - * 2 power levels (if more than 2 are available)
>>>>>> - */
>>>>>> - if (dpm_table->count > 2)
>>>>>> - min_level = dpm_table->count - 2;
>>>>>> - else if (dpm_table->count == 2)
>>>>>> - min_level = 1;
>>>>>> - else
>>>>>> - min_level = 0;
>>>>>> -
>>>>>> - hwmgr->default_compute_power_profile.min_sclk =
>>>>>> - dpm_table->dpm_levels[min_level].value;
>>>>>> -
>>>>>> - hwmgr->gfx_power_profile = hwmgr->default_gfx_power_profile;
>>>>>> - hwmgr->compute_power_profile =
>>>>>> hwmgr->default_compute_power_profile;
>>>>>> -
>>>>>> - return 0;
>>>>>> -}
>>>>>> -
>>>>>> /**
>>>>>> * Initializes the SMC table and uploads it
>>>>>> *
>>>>>> @@ -2576,7 +2548,6 @@ static int vega10_init_smc_table(struct pp_hwmgr
>>>>>> *hwmgr)
>>>>>> PP_ASSERT_WITH_CODE(!result, "Attempt to enable AVFS feature
>>>>>> Failed!",
>>>>>> return result);
>>>>>> vega10_acg_enable(hwmgr);
>>>>>> - vega10_save_default_power_profile(hwmgr);
>>>>>>
>>>>>> return 0;
>>>>>> }
>>>>>> @@ -4729,62 +4700,6 @@ static int vega10_power_off_asic(struct pp_hwmgr
>>>>>> *hwmgr)
>>>>>> return result;
>>>>>> }
>>>>>>
>>>>>> -static void vega10_find_min_clock_index(struct pp_hwmgr *hwmgr,
>>>>>> - uint32_t *sclk_idx, uint32_t *mclk_idx,
>>>>>> - uint32_t min_sclk, uint32_t min_mclk)
>>>>>> -{
>>>>>> - struct vega10_hwmgr *data = (struct vega10_hwmgr
>>>>>> *)(hwmgr->backend);
>>>>>> - struct vega10_dpm_table *dpm_table = &(data->dpm_table);
>>>>>> - uint32_t i;
>>>>>> -
>>>>>> - for (i = 0; i < dpm_table->gfx_table.count; i++) {
>>>>>> - if (dpm_table->gfx_table.dpm_levels[i].enabled &&
>>>>>> - dpm_table->gfx_table.dpm_levels[i].value >=
>>>>>> min_sclk) {
>>>>>> - *sclk_idx = i;
>>>>>> - break;
>>>>>> - }
>>>>>> - }
>>>>>> -
>>>>>> - for (i = 0; i < dpm_table->mem_table.count; i++) {
>>>>>> - if (dpm_table->mem_table.dpm_levels[i].enabled &&
>>>>>> - dpm_table->mem_table.dpm_levels[i].value >=
>>>>>> min_mclk) {
>>>>>> - *mclk_idx = i;
>>>>>> - break;
>>>>>> - }
>>>>>> - }
>>>>>> -}
>>>>>> -
>>>>>> -static int vega10_set_power_profile_state(struct pp_hwmgr *hwmgr,
>>>>>> - struct amd_pp_profile *request)
>>>>>> -{
>>>>>> - struct vega10_hwmgr *data = (struct vega10_hwmgr
>>>>>> *)(hwmgr->backend);
>>>>>> - uint32_t sclk_idx = ~0, mclk_idx = ~0;
>>>>>> -
>>>>>> - if (hwmgr->dpm_level != AMD_DPM_FORCED_LEVEL_AUTO)
>>>>>> - return -EINVAL;
>>>>>> -
>>>>>> - vega10_find_min_clock_index(hwmgr, &sclk_idx, &mclk_idx,
>>>>>> - request->min_sclk, request->min_mclk);
>>>>>> -
>>>>>> - if (sclk_idx != ~0) {
>>>>>> - if (!data->registry_data.sclk_dpm_key_disabled)
>>>>>> - smum_send_msg_to_smc_with_parameter(
>>>>>> - hwmgr,
>>>>>> -
>>>>>> PPSMC_MSG_SetSoftMinGfxclkByIndex,
>>>>>> - sclk_idx);
>>>>>> - }
>>>>>> -
>>>>>> - if (mclk_idx != ~0) {
>>>>>> - if (!data->registry_data.mclk_dpm_key_disabled)
>>>>>> - smum_send_msg_to_smc_with_parameter(
>>>>>> - hwmgr,
>>>>>> - PPSMC_MSG_SetSoftMinUclkByIndex,
>>>>>> - mclk_idx);
>>>>>> - }
>>>>>> -
>>>>>> - return 0;
>>>>>> -}
>>>>>> -
>>>>>> static int vega10_get_sclk_od(struct pp_hwmgr *hwmgr)
>>>>>> {
>>>>>> struct vega10_hwmgr *data = (struct vega10_hwmgr
>>>>>> *)(hwmgr->backend);
>>>>>> @@ -5078,7 +4993,6 @@ static int vega10_set_power_profile_mode(struct
>>>>>> pp_hwmgr *hwmgr, long *input, ui
>>>>>>
>>>>>> vega10_check_smc_update_required_for_display_configuration,
>>>>>> .power_off_asic = vega10_power_off_asic,
>>>>>> .disable_smc_firmware_ctf = vega10_thermal_disable_alert,
>>>>>> - .set_power_profile_state = vega10_set_power_profile_state,
>>>>>> .get_sclk_od = vega10_get_sclk_od,
>>>>>> .set_sclk_od = vega10_set_sclk_od,
>>>>>> .get_mclk_od = vega10_get_mclk_od,
>>>>>> diff --git a/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h
>>>>>> b/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h
>>>>>> index 6f528e6..b366a5b 100644
>>>>>> --- a/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h
>>>>>> +++ b/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h
>>>>>> @@ -448,6 +448,5 @@ extern int phm_display_clock_voltage_request(struct
>>>>>> pp_hwmgr *hwmgr,
>>>>>>
>>>>>> extern int phm_get_max_high_clocks(struct pp_hwmgr *hwmgr, struct
>>>>>> amd_pp_simple_clock_info *clocks);
>>>>>> extern int phm_disable_smc_firmware_ctf(struct pp_hwmgr *hwmgr);
>>>>>> -extern int phm_reset_power_profile_state(struct pp_hwmgr *hwmgr);
>>>>>> #endif /* _HARDWARE_MANAGER_H_ */
>>>>>>
>>>>>> diff --git a/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h
>>>>>> b/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h
>>>>>> index 7b19be4..a3beb31 100644
>>>>>> --- a/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h
>>>>>> +++ b/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h
>>>>>> @@ -236,8 +236,6 @@ struct pp_smumgr_func {
>>>>>> uint32_t (*get_offsetof)(uint32_t type, uint32_t member);
>>>>>> uint32_t (*get_mac_definition)(uint32_t value);
>>>>>> bool (*is_dpm_running)(struct pp_hwmgr *hwmgr);
>>>>>> - int (*populate_requested_graphic_levels)(struct pp_hwmgr *hwmgr,
>>>>>> - struct amd_pp_profile *request);
>>>>>> bool (*is_hw_avfs_present)(struct pp_hwmgr *hwmgr);
>>>>>> int (*update_dpm_settings)(struct pp_hwmgr *hwmgr, void
>>>>>> *profile_setting);
>>>>>> };
>>>>>> @@ -329,8 +327,6 @@ struct pp_hwmgr_func {
>>>>>> int (*get_mclk_od)(struct pp_hwmgr *hwmgr);
>>>>>> int (*set_mclk_od)(struct pp_hwmgr *hwmgr, uint32_t value);
>>>>>> int (*read_sensor)(struct pp_hwmgr *hwmgr, int idx, void
>>>>>> *value, int *size);
>>>>>> - int (*set_power_profile_state)(struct pp_hwmgr *hwmgr,
>>>>>> - struct amd_pp_profile *request);
>>>>>> int (*avfs_control)(struct pp_hwmgr *hwmgr, bool enable);
>>>>>> int (*disable_smc_firmware_ctf)(struct pp_hwmgr *hwmgr);
>>>>>> int (*set_active_display_count)(struct pp_hwmgr *hwmgr,
>>>>>> uint32_t count);
>>>>>> @@ -752,11 +748,6 @@ struct pp_hwmgr {
>>>>>> uint32_t feature_mask;
>>>>>>
>>>>>> /* UMD Pstate */
>>>>>> - struct amd_pp_profile gfx_power_profile;
>>>>>> - struct amd_pp_profile compute_power_profile;
>>>>>> - struct amd_pp_profile default_gfx_power_profile;
>>>>>> - struct amd_pp_profile default_compute_power_profile;
>>>>>> - enum amd_pp_profile_type current_power_profile;
>>>>>> bool en_umd_pstate;
>>>>>> uint32_t power_profile_mode;
>>>>>> uint32_t default_power_profile_mode;
>>>>>> diff --git a/drivers/gpu/drm/amd/powerplay/inc/smumgr.h
>>>>>> b/drivers/gpu/drm/amd/powerplay/inc/smumgr.h
>>>>>> index e05a57e..e1f6e83 100644
>>>>>> --- a/drivers/gpu/drm/amd/powerplay/inc/smumgr.h
>>>>>> +++ b/drivers/gpu/drm/amd/powerplay/inc/smumgr.h
>>>>>> @@ -129,9 +129,6 @@ extern uint32_t smum_get_offsetof(struct pp_hwmgr
>>>>>> *hwmgr,
>>>>>>
>>>>>> extern bool smum_is_dpm_running(struct pp_hwmgr *hwmgr);
>>>>>>
>>>>>> -extern int smum_populate_requested_graphic_levels(struct pp_hwmgr
>>>>>> *hwmgr,
>>>>>> - struct amd_pp_profile *request);
>>>>>> -
>>>>>> extern bool smum_is_hw_avfs_present(struct pp_hwmgr *hwmgr);
>>>>>>
>>>>>> extern int smum_update_dpm_settings(struct pp_hwmgr *hwmgr, void
>>>>>> *profile_setting);
>>>>>> diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/ci_smumgr.c
>>>>>> b/drivers/gpu/drm/amd/powerplay/smumgr/ci_smumgr.c
>>>>>> index 6dd10ef..6f4bf7e 100644
>>>>>> --- a/drivers/gpu/drm/amd/powerplay/smumgr/ci_smumgr.c
>>>>>> +++ b/drivers/gpu/drm/amd/powerplay/smumgr/ci_smumgr.c
>>>>>> @@ -2772,32 +2772,6 @@ static bool ci_is_dpm_running(struct pp_hwmgr
>>>>>> *hwmgr)
>>>>>> return ci_is_smc_ram_running(hwmgr);
>>>>>> }
>>>>>>
>>>>>> -static int ci_populate_requested_graphic_levels(struct pp_hwmgr *hwmgr,
>>>>>> - struct amd_pp_profile
>>>>>> *request)
>>>>>> -{
>>>>>> - struct ci_smumgr *smu_data = (struct ci_smumgr *)
>>>>>> - (hwmgr->smu_backend);
>>>>>> - struct SMU7_Discrete_GraphicsLevel *levels =
>>>>>> - smu_data->smc_state_table.GraphicsLevel;
>>>>>> - uint32_t array = smu_data->dpm_table_start +
>>>>>> - offsetof(SMU7_Discrete_DpmTable, GraphicsLevel);
>>>>>> - uint32_t array_size = sizeof(struct SMU7_Discrete_GraphicsLevel)
>>>>>> *
>>>>>> - SMU7_MAX_LEVELS_GRAPHICS;
>>>>>> - uint32_t i;
>>>>>> -
>>>>>> - for (i = 0; i < smu_data->smc_state_table.GraphicsDpmLevelCount;
>>>>>> i++) {
>>>>>> - levels[i].ActivityLevel =
>>>>>> -
>>>>>> cpu_to_be16(request->activity_threshold);
>>>>>> - levels[i].EnabledForActivity = 1;
>>>>>> - levels[i].UpH = request->up_hyst;
>>>>>> - levels[i].DownH = request->down_hyst;
>>>>>> - }
>>>>>> -
>>>>>> - return ci_copy_bytes_to_smc(hwmgr, array, (uint8_t *)levels,
>>>>>> - array_size, SMC_RAM_END);
>>>>>> -}
>>>>>> -
>>>>>> -
>>>>>> static int ci_smu_init(struct pp_hwmgr *hwmgr)
>>>>>> {
>>>>>> struct ci_smumgr *ci_priv = NULL;
>>>>>> @@ -2942,6 +2916,5 @@ static int ci_update_dpm_settings(struct pp_hwmgr
>>>>>> *hwmgr,
>>>>>> .get_mac_definition = ci_get_mac_definition,
>>>>>> .initialize_mc_reg_table = ci_initialize_mc_reg_table,
>>>>>> .is_dpm_running = ci_is_dpm_running,
>>>>>> - .populate_requested_graphic_levels =
>>>>>> ci_populate_requested_graphic_levels,
>>>>>> .update_dpm_settings = ci_update_dpm_settings,
>>>>>> };
>>>>>> diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/fiji_smumgr.c
>>>>>> b/drivers/gpu/drm/amd/powerplay/smumgr/fiji_smumgr.c
>>>>>> index f7f58f77..220159f 100644
>>>>>> --- a/drivers/gpu/drm/amd/powerplay/smumgr/fiji_smumgr.c
>>>>>> +++ b/drivers/gpu/drm/amd/powerplay/smumgr/fiji_smumgr.c
>>>>>> @@ -1960,44 +1960,6 @@ static int fiji_init_arb_table_index(struct
>>>>>> pp_hwmgr *hwmgr)
>>>>>> smu_data->smu7_data.arb_table_start, tmp,
>>>>>> SMC_RAM_END);
>>>>>> }
>>>>>>
>>>>>> -static int fiji_save_default_power_profile(struct pp_hwmgr *hwmgr)
>>>>>> -{
>>>>>> - struct fiji_smumgr *data = (struct fiji_smumgr
>>>>>> *)(hwmgr->smu_backend);
>>>>>> - struct SMU73_Discrete_GraphicsLevel *levels =
>>>>>> - data->smc_state_table.GraphicsLevel;
>>>>>> - unsigned min_level = 1;
>>>>>> -
>>>>>> - hwmgr->default_gfx_power_profile.activity_threshold =
>>>>>> - be16_to_cpu(levels[0].ActivityLevel);
>>>>>> - hwmgr->default_gfx_power_profile.up_hyst = levels[0].UpHyst;
>>>>>> - hwmgr->default_gfx_power_profile.down_hyst = levels[0].DownHyst;
>>>>>> - hwmgr->default_gfx_power_profile.type = AMD_PP_GFX_PROFILE;
>>>>>> -
>>>>>> - hwmgr->default_compute_power_profile =
>>>>>> hwmgr->default_gfx_power_profile;
>>>>>> - hwmgr->default_compute_power_profile.type =
>>>>>> AMD_PP_COMPUTE_PROFILE;
>>>>>> -
>>>>>> - /* Workaround compute SDMA instability: disable lowest SCLK
>>>>>> - * DPM level. Optimize compute power profile: Use only highest
>>>>>> - * 2 power levels (if more than 2 are available), Hysteresis:
>>>>>> - * 0ms up, 5ms down
>>>>>> - */
>>>>>> - if (data->smc_state_table.GraphicsDpmLevelCount > 2)
>>>>>> - min_level = data->smc_state_table.GraphicsDpmLevelCount
>>>>>> - 2;
>>>>>> - else if (data->smc_state_table.GraphicsDpmLevelCount == 2)
>>>>>> - min_level = 1;
>>>>>> - else
>>>>>> - min_level = 0;
>>>>>> - hwmgr->default_compute_power_profile.min_sclk =
>>>>>> - be32_to_cpu(levels[min_level].SclkFrequency);
>>>>>> - hwmgr->default_compute_power_profile.up_hyst = 0;
>>>>>> - hwmgr->default_compute_power_profile.down_hyst = 5;
>>>>>> -
>>>>>> - hwmgr->gfx_power_profile = hwmgr->default_gfx_power_profile;
>>>>>> - hwmgr->compute_power_profile =
>>>>>> hwmgr->default_compute_power_profile;
>>>>>> -
>>>>>> - return 0;
>>>>>> -}
>>>>>> -
>>>>>> static int fiji_setup_dpm_led_config(struct pp_hwmgr *hwmgr)
>>>>>> {
>>>>>> pp_atomctrl_voltage_table param_led_dpm;
>>>>>> @@ -2238,8 +2200,6 @@ static int fiji_init_smc_table(struct pp_hwmgr
>>>>>> *hwmgr)
>>>>>> PP_ASSERT_WITH_CODE(0 == result,
>>>>>> "Failed to setup dpm led config", return
>>>>>> result);
>>>>>>
>>>>>> - fiji_save_default_power_profile(hwmgr);
>>>>>> -
>>>>>> return 0;
>>>>>> }
>>>>>>
>>>>>> @@ -2694,31 +2654,6 @@ static bool fiji_is_dpm_running(struct pp_hwmgr
>>>>>> *hwmgr)
>>>>>> ? true : false;
>>>>>> }
>>>>>>
>>>>>> -static int fiji_populate_requested_graphic_levels(struct pp_hwmgr
>>>>>> *hwmgr,
>>>>>> - struct amd_pp_profile *request)
>>>>>> -{
>>>>>> - struct fiji_smumgr *smu_data = (struct fiji_smumgr *)
>>>>>> - (hwmgr->smu_backend);
>>>>>> - struct SMU73_Discrete_GraphicsLevel *levels =
>>>>>> - smu_data->smc_state_table.GraphicsLevel;
>>>>>> - uint32_t array = smu_data->smu7_data.dpm_table_start +
>>>>>> - offsetof(SMU73_Discrete_DpmTable,
>>>>>> GraphicsLevel);
>>>>>> - uint32_t array_size = sizeof(struct
>>>>>> SMU73_Discrete_GraphicsLevel) *
>>>>>> - SMU73_MAX_LEVELS_GRAPHICS;
>>>>>> - uint32_t i;
>>>>>> -
>>>>>> - for (i = 0; i < smu_data->smc_state_table.GraphicsDpmLevelCount;
>>>>>> i++) {
>>>>>> - levels[i].ActivityLevel =
>>>>>> -
>>>>>> cpu_to_be16(request->activity_threshold);
>>>>>> - levels[i].EnabledForActivity = 1;
>>>>>> - levels[i].UpHyst = request->up_hyst;
>>>>>> - levels[i].DownHyst = request->down_hyst;
>>>>>> - }
>>>>>> -
>>>>>> - return smu7_copy_bytes_to_smc(hwmgr, array, (uint8_t *)levels,
>>>>>> - array_size, SMC_RAM_END);
>>>>>> -}
>>>>>> -
>>>>>> static int fiji_update_dpm_settings(struct pp_hwmgr *hwmgr,
>>>>>> void *profile_setting)
>>>>>> {
>>>>>> @@ -2838,7 +2773,6 @@ static int fiji_update_dpm_settings(struct
>>>>>> pp_hwmgr *hwmgr,
>>>>>> .get_mac_definition = fiji_get_mac_definition,
>>>>>> .initialize_mc_reg_table = fiji_initialize_mc_reg_table,
>>>>>> .is_dpm_running = fiji_is_dpm_running,
>>>>>> - .populate_requested_graphic_levels =
>>>>>> fiji_populate_requested_graphic_levels,
>>>>>> .is_hw_avfs_present = fiji_is_hw_avfs_present,
>>>>>> .update_dpm_settings = fiji_update_dpm_settings,
>>>>>> };
>>>>>> diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c
>>>>>> b/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c
>>>>>> index ae07b5d..75f46ed 100644
>>>>>> --- a/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c
>>>>>> +++ b/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c
>>>>>> @@ -1840,42 +1840,6 @@ static void
>>>>>> polaris10_initialize_power_tune_defaults(struct pp_hwmgr *hwmgr)
>>>>>>
>>>>>> }
>>>>>>
>>>>>> -static void polaris10_save_default_power_profile(struct pp_hwmgr
>>>>>> *hwmgr)
>>>>>> -{
>>>>>> - struct polaris10_smumgr *data = (struct polaris10_smumgr
>>>>>> *)(hwmgr->smu_backend);
>>>>>> - struct SMU74_Discrete_GraphicsLevel *levels =
>>>>>> - data->smc_state_table.GraphicsLevel;
>>>>>> - unsigned min_level = 1;
>>>>>> -
>>>>>> - hwmgr->default_gfx_power_profile.activity_threshold =
>>>>>> - be16_to_cpu(levels[0].ActivityLevel);
>>>>>> - hwmgr->default_gfx_power_profile.up_hyst = levels[0].UpHyst;
>>>>>> - hwmgr->default_gfx_power_profile.down_hyst = levels[0].DownHyst;
>>>>>> - hwmgr->default_gfx_power_profile.type = AMD_PP_GFX_PROFILE;
>>>>>> -
>>>>>> - hwmgr->default_compute_power_profile =
>>>>>> hwmgr->default_gfx_power_profile;
>>>>>> - hwmgr->default_compute_power_profile.type =
>>>>>> AMD_PP_COMPUTE_PROFILE;
>>>>>> -
>>>>>> - /* Workaround compute SDMA instability: disable lowest SCLK
>>>>>> - * DPM level. Optimize compute power profile: Use only highest
>>>>>> - * 2 power levels (if more than 2 are available), Hysteresis:
>>>>>> - * 0ms up, 5ms down
>>>>>> - */
>>>>>> - if (data->smc_state_table.GraphicsDpmLevelCount > 2)
>>>>>> - min_level = data->smc_state_table.GraphicsDpmLevelCount
>>>>>> - 2;
>>>>>> - else if (data->smc_state_table.GraphicsDpmLevelCount == 2)
>>>>>> - min_level = 1;
>>>>>> - else
>>>>>> - min_level = 0;
>>>>>> - hwmgr->default_compute_power_profile.min_sclk =
>>>>>> -
>>>>>> be32_to_cpu(levels[min_level].SclkSetting.SclkFrequency);
>>>>>> - hwmgr->default_compute_power_profile.up_hyst = 0;
>>>>>> - hwmgr->default_compute_power_profile.down_hyst = 5;
>>>>>> -
>>>>>> - hwmgr->gfx_power_profile = hwmgr->default_gfx_power_profile;
>>>>>> - hwmgr->compute_power_profile =
>>>>>> hwmgr->default_compute_power_profile;
>>>>>> -}
>>>>>> -
>>>>>> static int polaris10_init_smc_table(struct pp_hwmgr *hwmgr)
>>>>>> {
>>>>>> int result;
>>>>>> @@ -2090,8 +2054,6 @@ static int polaris10_init_smc_table(struct
>>>>>> pp_hwmgr *hwmgr)
>>>>>> PP_ASSERT_WITH_CODE(0 == result,
>>>>>> "Failed to populate PM fuses to SMC memory!",
>>>>>> return result);
>>>>>>
>>>>>> - polaris10_save_default_power_profile(hwmgr);
>>>>>> -
>>>>>> return 0;
>>>>>> }
>>>>>>
>>>>>> @@ -2550,31 +2512,6 @@ static bool polaris10_is_dpm_running(struct
>>>>>> pp_hwmgr *hwmgr)
>>>>>> ? true : false;
>>>>>> }
>>>>>>
>>>>>> -static int polaris10_populate_requested_graphic_levels(struct pp_hwmgr
>>>>>> *hwmgr,
>>>>>> - struct amd_pp_profile *request)
>>>>>> -{
>>>>>> - struct polaris10_smumgr *smu_data = (struct polaris10_smumgr *)
>>>>>> - (hwmgr->smu_backend);
>>>>>> - struct SMU74_Discrete_GraphicsLevel *levels =
>>>>>> - smu_data->smc_state_table.GraphicsLevel;
>>>>>> - uint32_t array = smu_data->smu7_data.dpm_table_start +
>>>>>> - offsetof(SMU74_Discrete_DpmTable,
>>>>>> GraphicsLevel);
>>>>>> - uint32_t array_size = sizeof(struct
>>>>>> SMU74_Discrete_GraphicsLevel) *
>>>>>> - SMU74_MAX_LEVELS_GRAPHICS;
>>>>>> - uint32_t i;
>>>>>> -
>>>>>> - for (i = 0; i < smu_data->smc_state_table.GraphicsDpmLevelCount;
>>>>>> i++) {
>>>>>> - levels[i].ActivityLevel =
>>>>>> -
>>>>>> cpu_to_be16(request->activity_threshold);
>>>>>> - levels[i].EnabledForActivity = 1;
>>>>>> - levels[i].UpHyst = request->up_hyst;
>>>>>> - levels[i].DownHyst = request->down_hyst;
>>>>>> - }
>>>>>> -
>>>>>> - return smu7_copy_bytes_to_smc(hwmgr, array, (uint8_t *)levels,
>>>>>> - array_size, SMC_RAM_END);
>>>>>> -}
>>>>>> -
>>>>>> static int polaris10_update_dpm_settings(struct pp_hwmgr *hwmgr,
>>>>>> void *profile_setting)
>>>>>> {
>>>>>> @@ -2693,7 +2630,6 @@ static int polaris10_update_dpm_settings(struct
>>>>>> pp_hwmgr *hwmgr,
>>>>>> .populate_all_memory_levels =
>>>>>> polaris10_populate_all_memory_levels,
>>>>>> .get_mac_definition = polaris10_get_mac_definition,
>>>>>> .is_dpm_running = polaris10_is_dpm_running,
>>>>>> - .populate_requested_graphic_levels =
>>>>>> polaris10_populate_requested_graphic_levels,
>>>>>> .is_hw_avfs_present = polaris10_is_hw_avfs_present,
>>>>>> .update_dpm_settings = polaris10_update_dpm_settings,
>>>>>> };
>>>>>> diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c
>>>>>> b/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c
>>>>>> index 1ce4959..43b1010 100644
>>>>>> --- a/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c
>>>>>> +++ b/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c
>>>>>> @@ -236,16 +236,6 @@ bool smum_is_dpm_running(struct pp_hwmgr *hwmgr)
>>>>>> return true;
>>>>>> }
>>>>>>
>>>>>> -int smum_populate_requested_graphic_levels(struct pp_hwmgr *hwmgr,
>>>>>> - struct amd_pp_profile *request)
>>>>>> -{
>>>>>> - if (hwmgr->smumgr_funcs->populate_requested_graphic_levels)
>>>>>> - return
>>>>>> hwmgr->smumgr_funcs->populate_requested_graphic_levels(
>>>>>> - hwmgr, request);
>>>>>> -
>>>>>> - return 0;
>>>>>> -}
>>>>>> -
>>>>>> bool smum_is_hw_avfs_present(struct pp_hwmgr *hwmgr)
>>>>>> {
>>>>>> if (hwmgr->smumgr_funcs->is_hw_avfs_present)
>>>>>> diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/tonga_smumgr.c
>>>>>> b/drivers/gpu/drm/amd/powerplay/smumgr/tonga_smumgr.c
>>>>>> index 1fe5b77..8856cc1 100644
>>>>>> --- a/drivers/gpu/drm/amd/powerplay/smumgr/tonga_smumgr.c
>>>>>> +++ b/drivers/gpu/drm/amd/powerplay/smumgr/tonga_smumgr.c
>>>>>> @@ -2263,42 +2263,6 @@ static void
>>>>>> tonga_initialize_power_tune_defaults(struct pp_hwmgr *hwmgr)
>>>>>> smu_data->power_tune_defaults =
>>>>>> &tonga_power_tune_data_set_array[0];
>>>>>> }
>>>>>>
>>>>>> -static void tonga_save_default_power_profile(struct pp_hwmgr *hwmgr)
>>>>>> -{
>>>>>> - struct tonga_smumgr *data = (struct tonga_smumgr
>>>>>> *)(hwmgr->smu_backend);
>>>>>> - struct SMU72_Discrete_GraphicsLevel *levels =
>>>>>> - data->smc_state_table.GraphicsLevel;
>>>>>> - unsigned min_level = 1;
>>>>>> -
>>>>>> - hwmgr->default_gfx_power_profile.activity_threshold =
>>>>>> - be16_to_cpu(levels[0].ActivityLevel);
>>>>>> - hwmgr->default_gfx_power_profile.up_hyst = levels[0].UpHyst;
>>>>>> - hwmgr->default_gfx_power_profile.down_hyst = levels[0].DownHyst;
>>>>>> - hwmgr->default_gfx_power_profile.type = AMD_PP_GFX_PROFILE;
>>>>>> -
>>>>>> - hwmgr->default_compute_power_profile =
>>>>>> hwmgr->default_gfx_power_profile;
>>>>>> - hwmgr->default_compute_power_profile.type =
>>>>>> AMD_PP_COMPUTE_PROFILE;
>>>>>> -
>>>>>> - /* Workaround compute SDMA instability: disable lowest SCLK
>>>>>> - * DPM level. Optimize compute power profile: Use only highest
>>>>>> - * 2 power levels (if more than 2 are available), Hysteresis:
>>>>>> - * 0ms up, 5ms down
>>>>>> - */
>>>>>> - if (data->smc_state_table.GraphicsDpmLevelCount > 2)
>>>>>> - min_level = data->smc_state_table.GraphicsDpmLevelCount
>>>>>> - 2;
>>>>>> - else if (data->smc_state_table.GraphicsDpmLevelCount == 2)
>>>>>> - min_level = 1;
>>>>>> - else
>>>>>> - min_level = 0;
>>>>>> - hwmgr->default_compute_power_profile.min_sclk =
>>>>>> - be32_to_cpu(levels[min_level].SclkFrequency);
>>>>>> - hwmgr->default_compute_power_profile.up_hyst = 0;
>>>>>> - hwmgr->default_compute_power_profile.down_hyst = 5;
>>>>>> -
>>>>>> - hwmgr->gfx_power_profile = hwmgr->default_gfx_power_profile;
>>>>>> - hwmgr->compute_power_profile =
>>>>>> hwmgr->default_compute_power_profile;
>>>>>> -}
>>>>>> -
>>>>>> static int tonga_init_smc_table(struct pp_hwmgr *hwmgr)
>>>>>> {
>>>>>> int result;
>>>>>> @@ -2540,8 +2504,6 @@ static int tonga_init_smc_table(struct pp_hwmgr
>>>>>> *hwmgr)
>>>>>> PP_ASSERT_WITH_CODE((!result),
>>>>>> "Failed to populate initialize MC Reg table !", return
>>>>>> result);
>>>>>>
>>>>>> - tonga_save_default_power_profile(hwmgr);
>>>>>> -
>>>>>> return 0;
>>>>>> }
>>>>>>
>>>>>> @@ -3259,31 +3221,6 @@ static bool tonga_is_dpm_running(struct pp_hwmgr
>>>>>> *hwmgr)
>>>>>> ? true : false;
>>>>>> }
>>>>>>
>>>>>> -static int tonga_populate_requested_graphic_levels(struct pp_hwmgr
>>>>>> *hwmgr,
>>>>>> - struct amd_pp_profile *request)
>>>>>> -{
>>>>>> - struct tonga_smumgr *smu_data = (struct tonga_smumgr *)
>>>>>> - (hwmgr->smu_backend);
>>>>>> - struct SMU72_Discrete_GraphicsLevel *levels =
>>>>>> - smu_data->smc_state_table.GraphicsLevel;
>>>>>> - uint32_t array = smu_data->smu7_data.dpm_table_start +
>>>>>> - offsetof(SMU72_Discrete_DpmTable,
>>>>>> GraphicsLevel);
More information about the amd-gfx
mailing list