cpufreq: change min_sample_time
satya popuri [Fri, 9 Mar 2012 18:41:33 +0000 (10:41 -0800)]
We prefer a min_sample_time of 30ms. This is tied to our cluster switch
and auto-hotplug artificial delays.

Signed-off-by: satya popuri <spopuri@nvidia.com>

original commit message:

cpufreq: reduce min_sample_time

Reduce the minimum amount of time the interactive governor spends at a
frequency before ramping down. This parameter (min_sample_time) was 80ms
now reduced to 30 ms based on perf and power sweep numbers.

Reviewed-on: http://git-master/r/86900 (cherry picked from commit
581f20a5b7b8c3e4ee2c211b28dbc81510af2ae7).
Signed-off-by: satya popuri <spopuri@nvidia.com>

Change-Id: I24f4cd52737950fd4e78a36b4ee34a84551e0e12
Reviewed-on: http://git-master/r/89217
Reviewed-by: Satya Popuri <spopuri@nvidia.com>
Tested-by: Satya Popuri <spopuri@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Diwakar Tundlam <dtundlam@nvidia.com>

drivers/cpufreq/cpufreq_interactive.c

index 870c988..5b7770c 100644 (file)
@@ -77,7 +77,7 @@ static unsigned long sustain_load;
 /*
  * The minimum amount of time to spend at a frequency before we can ramp down.
  */
-#define DEFAULT_MIN_SAMPLE_TIME 20000;
+#define DEFAULT_MIN_SAMPLE_TIME 30000;
 static unsigned long min_sample_time;
 
 /*