I see a couple of new performance tunings for PTL, can we have them in the same patch and then rewrite the commit message, instead of it being only applicable to L3SQCREG5?

-shekhar

On 9/7/2024 5:25 AM, Matt Roper wrote:
On Fri, Sep 06, 2024 at 02:51:52PM -0700, Matt Atwood wrote:
From: Matt Roper <matthew.d.roper@intel.com>

The L3SQCREG5 tuning suggestion remains the same on Xe3 as on Xe2.

Bspec: 72161
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Matt Atwood <matthew.s.atwood@intel.com>
---
 drivers/gpu/drm/xe/xe_tuning.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/xe/xe_tuning.c b/drivers/gpu/drm/xe/xe_tuning.c
index faa1bf42e50e..9bb6afaa560a 100644
--- a/drivers/gpu/drm/xe/xe_tuning.c
+++ b/drivers/gpu/drm/xe/xe_tuning.c
@@ -25,15 +25,15 @@ static const struct xe_rtp_entry_sr gt_tunings[] = {
 	  XE_RTP_ACTIONS(SET(XEHP_SQCM, EN_32B_ACCESS))
 	},
 
-	/* Xe2 */
+	/* Xe2 / Xe3 */
 
 	{ XE_RTP_NAME("Tuning: L3 cache"),
-	  XE_RTP_RULES(GRAPHICS_VERSION_RANGE(2001, XE_RTP_END_VERSION_UNDEFINED)),
+	  XE_RTP_RULES(GRAPHICS_VERSION_RANGE(2001, 3000)),
We should just leave this one as upper bound undefined (i.e., applies to
Xe2 and all future platforms until we come back here to change it).

 	  XE_RTP_ACTIONS(FIELD_SET(XEHP_L3SQCREG5, L3_PWM_TIMER_INIT_VAL_MASK,
 				   REG_FIELD_PREP(L3_PWM_TIMER_INIT_VAL_MASK, 0x7f)))
 	},
 	{ XE_RTP_NAME("Tuning: L3 cache - media"),
-	  XE_RTP_RULES(MEDIA_VERSION(2000)),
+	  XE_RTP_RULES(MEDIA_VERSION_RANGE(2000, 3000)),
And we should probably use an unbound upper range on this one as well.
In general we should probably assume that tuning settings will continue
to hold true to future platforms unless/until we know otherwise, since
that also matches how these things are now being documented in the
bspec.


Matt

 	  XE_RTP_ACTIONS(FIELD_SET(XE2LPM_L3SQCREG5, L3_PWM_TIMER_INIT_VAL_MASK,
 				   REG_FIELD_PREP(L3_PWM_TIMER_INIT_VAL_MASK, 0x7f)))
 	},
-- 
2.44.0


    
-- 
-shekhar