drm/i915/gen10: fix the gen 10 SAGV block time
authorPaulo Zanoni <paulo.r.zanoni@intel.com>
Wed, 9 Aug 2017 20:52:45 +0000 (13:52 -0700)
committerRodrigo Vivi <rodrigo.vivi@intel.com>
Thu, 10 Aug 2017 20:59:28 +0000 (13:59 -0700)
A previous commit added CNL to intel_has_sagv(), but forgot to adjust
the SAGV block time to gen 10 platforms.

Cc: Mahesh Kumar <mahesh1.kumar@intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170809205248.11917-3-rodrigo.vivi@intel.com
drivers/gpu/drm/i915/intel_pm.c

index 52bf62b6e38c4d09e0b08fd2e09e38d068e3b66a..f64fdae88b7dc8acbe2d53b3698e225e1eff2e3c 100644 (file)
@@ -3552,8 +3552,6 @@ bool ilk_disable_lp_wm(struct drm_device *dev)
        return _ilk_disable_lp_wm(dev_priv, WM_DIRTY_LP_ALL);
 }
 
-#define SKL_SAGV_BLOCK_TIME    30 /* µs */
-
 /*
  * FIXME: We still don't have the proper code detect if we need to apply the WA,
  * so assume we'll always need it in order to avoid underruns.
@@ -3678,12 +3676,13 @@ bool intel_can_enable_sagv(struct drm_atomic_state *state)
        struct intel_crtc_state *cstate;
        enum pipe pipe;
        int level, latency;
+       int sagv_block_time_us = IS_GEN9(dev_priv) ? 30 : 20;
 
        if (!intel_has_sagv(dev_priv))
                return false;
 
        /*
-        * SKL workaround: bspec recommends we disable the SAGV when we have
+        * SKL+ workaround: bspec recommends we disable the SAGV when we have
         * more then one pipe enabled
         *
         * If there are no active CRTCs, no additional checks need be performed
@@ -3722,11 +3721,11 @@ bool intel_can_enable_sagv(struct drm_atomic_state *state)
                        latency += 15;
 
                /*
-                * If any of the planes on this pipe don't enable wm levels
-                * that incur memory latencies higher then 30µs we can't enable
-                * the SAGV
+                * If any of the planes on this pipe don't enable wm levels that
+                * incur memory latencies higher than sagv_block_time_us we
+                * can't enable the SAGV.
                 */
-               if (latency < SKL_SAGV_BLOCK_TIME)
+               if (latency < sagv_block_time_us)
                        return false;
        }