[PATCH 2/3] drm/amd/pp: Refine get_gpu_power for VI

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Sampling period is too short. The power reading value will be not 
aligned with AGT's. It will confuse user that why AMD provides two 
different power results.

Regards,
Eric

On 2018-04-04 04:25 AM, Rex Zhu wrote:
> 1. On polaris10/11/12, Sending smu message PPSMC_MSG_GetCurrPkgPwr to
>     read currentpkgpwr directly.
> 2. On Fiji/tonga/bonaire/hawwii, no PPSMC_MSG_GetCurrPkgPwr support.
>     Send PPSMC_MSG_PmStatusLogStart to let smu write currentpkgpwr
>     to ixSMU_PM_STATUS_94. this is asynchronous. need to delay no less
>     than 1ms.
> 3. Clean ixSMU_PM_STATUS_94 immediately when send PPSMC_MSG_PmStatusLogStart
>     to avoid read old value.
> 4. delay 10 ms instand of 20 ms. so the result will more similar to
>     the output of PPSMC_MSG_GetCurrPkgPwr.
> 5. remove max power/vddc/vddci output to keep consistent with vega.
> 6. for vddc/vddci power, we can calculate the average value per
>     [10ms, 4s] in other interface if needed.
>
> Signed-off-by: Rex Zhu <Rex.Zhu at amd.com>
> ---
>   drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c | 45 +++++++++++-------------
>   1 file changed, 21 insertions(+), 24 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c
> index 40f2f87..c0ce672 100644
> --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c
> +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c
> @@ -3363,30 +3363,27 @@ static int smu7_get_pp_table_entry(struct pp_hwmgr *hwmgr,
>   static int smu7_get_gpu_power(struct pp_hwmgr *hwmgr,
>   		struct pp_gpu_power *query)
>   {
> -	PP_ASSERT_WITH_CODE(!smum_send_msg_to_smc(hwmgr,
> -			PPSMC_MSG_PmStatusLogStart),
> -			"Failed to start pm status log!",
> -			return -1);
> -
> -	msleep_interruptible(20);
> -
> -	PP_ASSERT_WITH_CODE(!smum_send_msg_to_smc(hwmgr,
> -			PPSMC_MSG_PmStatusLogSample),
> -			"Failed to sample pm status log!",
> -			return -1);
> -
> -	query->vddc_power = cgs_read_ind_register(hwmgr->device,
> -			CGS_IND_REG__SMC,
> -			ixSMU_PM_STATUS_40);
> -	query->vddci_power = cgs_read_ind_register(hwmgr->device,
> -			CGS_IND_REG__SMC,
> -			ixSMU_PM_STATUS_49);
> -	query->max_gpu_power = cgs_read_ind_register(hwmgr->device,
> -			CGS_IND_REG__SMC,
> -			ixSMU_PM_STATUS_94);
> -	query->average_gpu_power = cgs_read_ind_register(hwmgr->device,
> -			CGS_IND_REG__SMC,
> -			ixSMU_PM_STATUS_95);
> +	if (!query)
> +		return -EINVAL;
> +
> +	memset(query, 0, sizeof *query);
> +
> +	if (hwmgr->chip_id >= CHIP_POLARIS10) {
> +		smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetCurrPkgPwr);
> +		query->average_gpu_power = cgs_read_register(hwmgr->device, mmSMC_MSG_ARG_0);
> +	} else {
> +		smum_send_msg_to_smc(hwmgr, PPSMC_MSG_PmStatusLogStart);
> +		cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC,
> +					ixSMU_PM_STATUS_94, 0);
> +
> +		msleep_interruptible(10);
> +
> +		smum_send_msg_to_smc(hwmgr, PPSMC_MSG_PmStatusLogSample);
> +
> +		query->average_gpu_power = cgs_read_ind_register(hwmgr->device,
> +							CGS_IND_REG__SMC,
> +							ixSMU_PM_STATUS_94);
> +	}
>   
>   	return 0;
>   }



[Index of Archives]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux