On Mon, 15 Oct 2018, Manasi Navare <manasi.d.navare@xxxxxxxxx> wrote:
> On Mon, Oct 15, 2018 at 02:50:32PM -0700, Anusha Srivatsa wrote:
>> Similar to DSC DPCD registers, let us cache
>> FEC_CAPABLE register to avoid using stale
>> values. With this we can avoid aux reads
>> everytime and instead read the cached values.
>>
>> Suggested-by: Jani Nikula <jani.nikula@xxxxxxxxxxxxxxx>
>> Cc: Jani Nikula <jani.nikula@xxxxxxxxxxxxxxx>
>> Cc: Ville Syrjala <ville.syrjala@xxxxxxxxxxxxxxx>
>> Cc: Manasi Navare <manasi.d.navare@xxxxxxxxx>
>> Signed-off-by: Anusha Srivatsa <anusha.srivatsa@xxxxxxxxx>
>> ---
>> drivers/gpu/drm/i915/intel_dp.c | 8 ++++++++
>> drivers/gpu/drm/i915/intel_drv.h | 1 +
>> 2 files changed, 9 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/i915/intel_dp.c b/drivers/gpu/drm/i915/intel_dp.c
>> index dcc5a207fcbd..8a0e0a0b26f6 100644
>> --- a/drivers/gpu/drm/i915/intel_dp.c
>> +++ b/drivers/gpu/drm/i915/intel_dp.c
>> @@ -4163,8 +4163,10 @@ static void intel_dp_get_dsc_sink_cap(struct intel_dp *intel_dp)
>> /*
>> *Clear the cached register set to avoid using stale values
>> * for the sinks that do not support DSC.
>> + * Similarly, clear the cached FEC register.
>> */
>> memset(intel_dp->dsc_dpcd, 0, sizeof(intel_dp->dsc_dpcd));
>> + memset(intel_dp->fec_dpcd, 0, sizeof(intel_dp->fec_dpcd));
>
> Memset is an expensive opertaion for just a single value,
> just set that to 0
Agreed on making it a single byte and setting it to 0, but the memset
impact on performance is noise compared to e.g. the native aux
transactions on this path.
BR,
Jani.
>
>>
>> /* Cache the DSC DPCD if eDP or DP rev >= 1.4 */
>> if (intel_dp->dpcd[DP_DPCD_REV] >= 0x14 ||
>> @@ -4179,6 +4181,12 @@ static void intel_dp_get_dsc_sink_cap(struct intel_dp *intel_dp)
>> (int)sizeof(intel_dp->dsc_dpcd),
>> intel_dp->dsc_dpcd);
>> }
>> + /* FEC is supported only on DP 1.4 */
>> + if (!intel_dp_is_edp(intel_dp) && intel_dp->dpcd[DP_DPCD_REV] >= 0x14) {
>> + if (drm_dp_dpcd_readb(&intel_dp->aux, DP_FEC_CAPABILITY,
>> + intel_dp->fec_dpcd) < 0)
>> + DRM_ERROR("Failed to read FEC DPCD register\n");
>> + }
>> }
>>
>> static bool
>> diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h
>> index 7b4af8cba279..b87ea052c9ca 100644
>> --- a/drivers/gpu/drm/i915/intel_drv.h
>> +++ b/drivers/gpu/drm/i915/intel_drv.h
>> @@ -1086,6 +1086,7 @@ struct intel_dp {
>> uint8_t downstream_ports[DP_MAX_DOWNSTREAM_PORTS];
>> uint8_t edp_dpcd[EDP_DISPLAY_CTL_CAP_SIZE];
>> u8 dsc_dpcd[DP_DSC_RECEIVER_CAP_SIZE];
>> + u8 fec_dpcd[1];
>
> Why create an array for just 1 value, I think just a u8 fec_capable field
> will suffice
>
> Manasi
>
>> /* source rates */
>> int num_source_rates;
>> const int *source_rates;
>> --
>> 2.17.1
>>
--
Jani Nikula, Intel Open Source Graphics Center
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
