gpu: nvgpu: Add sw shadow for load value

Reading the load value may increase CPU power consumption
temprorarily. In most cases we are ok with a value that
was read a moment earlier.

This patch introduces a software shadow for gpu load. The shadow
is updated before starting scaling and all scaling code paths use
the sw shadow.

Change-Id: I53d2ccb8e7f83147f411a14d3104d890dd9af9a3
Signed-off-by: Arto Merilainen <amerilainen@nvidia.com>
Reviewed-on: http://git-master/r/453347
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Terje Bergstrom <tbergstrom@nvidia.com>

4 files changed, 14 insertions, 2 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/gk20a_scale.c b/drivers/gpu/nvgpu/gk20a/gk20a_scale.c
index 8a92828f..d09a18f8 100644
--- a/drivers/gpu/nvgpu/gk20a/gk20a_scale.c
+++ b/drivers/gpu/nvgpu/gk20a/gk20a_scale.c
@@ -192,6 +192,9 @@ static void gk20a_scale_notify(struct platform_device *pdev, bool busy)
         struct gk20a_scale_profile *profile = g->scale_profile;
         struct devfreq *devfreq = g->devfreq;
 
+        /* update the software shadow */
+        gk20a_pmu_load_update(g);
+
         /* inform edp about new constraint */
         if (platform->prescale)
                 platform->prescale(pdev);
diff --git a/drivers/gpu/nvgpu/gk20a/gk20a_sysfs.c b/drivers/gpu/nvgpu/gk20a/gk20a_sysfs.c
index 994c9cd2..97a0452c 100644
--- a/drivers/gpu/nvgpu/gk20a/gk20a_sysfs.c
+++ b/drivers/gpu/nvgpu/gk20a/gk20a_sysfs.c
@@ -278,6 +278,7 @@ static ssize_t gk20a_load_show(struct device *dev,
                 busy_time = 0;
         } else {
                 gk20a_busy(g->dev);
+                gk20a_pmu_load_update(g);
                 gk20a_pmu_load_norm(g, &busy_time);
                 gk20a_idle(g->dev);
         }
diff --git a/drivers/gpu/nvgpu/gk20a/pmu_gk20a.c b/drivers/gpu/nvgpu/gk20a/pmu_gk20a.c
index 7c441f53..06e7a4e6 100644
--- a/drivers/gpu/nvgpu/gk20a/pmu_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/pmu_gk20a.c
@@ -3677,16 +3677,22 @@ int gk20a_pmu_destroy(struct gk20a *g)
 
 int gk20a_pmu_load_norm(struct gk20a *g, u32 *load)
 {
+        *load = g->pmu.load_shadow;
+        return 0;
+}
+
+int gk20a_pmu_load_update(struct gk20a *g)
+{
         struct pmu_gk20a *pmu = &g->pmu;
         u16 _load = 0;
 
         if (!pmu->perfmon_ready) {
-                *load = 0;
+                pmu->load_shadow = 0;
                 return 0;
         }
 
         pmu_copy_from_dmem(pmu, pmu->sample_buffer, (u8 *)&_load, 2, 0);
-        *load = _load / 10;
+        pmu->load_shadow = _load / 10;
 
         return 0;
 }
diff --git a/drivers/gpu/nvgpu/gk20a/pmu_gk20a.h b/drivers/gpu/nvgpu/gk20a/pmu_gk20a.h
index 292aabb0..694e0288 100644
--- a/drivers/gpu/nvgpu/gk20a/pmu_gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/pmu_gk20a.h
@@ -1079,6 +1079,7 @@ struct pmu_gk20a {
         bool perfmon_ready;
 
         u32 sample_buffer;
+        u32 load_shadow;
 
         struct mutex isr_mutex;
         struct mutex isr_enable_lock;
@@ -1119,6 +1120,7 @@ int pmu_mutex_acquire(struct pmu_gk20a *pmu, u32 id, u32 *token);
 int pmu_mutex_release(struct pmu_gk20a *pmu, u32 id, u32 *token);
 int gk20a_pmu_destroy(struct gk20a *g);
 int gk20a_pmu_load_norm(struct gk20a *g, u32 *load);
+int gk20a_pmu_load_update(struct gk20a *g);
 int gk20a_pmu_debugfs_init(struct platform_device *dev);
 void gk20a_pmu_reset_load_counters(struct gk20a *g);
 void gk20a_pmu_get_load_counters(struct gk20a *g, u32 *busy_cycles,