nv-tegra-mirror/linux-nvgpu
1
0
Fork 0
mirror of git://nv-tegra.nvidia.com/linux-nvgpu.git synced 2025-12-22 17:36:20 +03:00
Code Issues Packages Projects Releases Wiki Activity

gpu: nvgpu: Add sw shadow for load value

Browse Source 
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>
This commit is contained in:
Arto Merilainen
2014-08-06 09:30:11 +03:00
committed by Dan Willemsen
parent 273f754cb5
commit b33020008b
4 changed files with 14 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
drivers/gpu/nvgpu/gk20a/gk20a_scale.c
View File

@@ -192,6 +192,9 @@ static void gk20a_scale_notify(struct platform_device *pdev, bool busy)
struct gk20a_scale_profile *profile = g->scale_profile; struct gk20a_scale_profile *profile = g->scale_profile;
struct devfreq *devfreq = g->devfreq; struct devfreq *devfreq = g->devfreq;
/* update the software shadow */
gk20a_pmu_load_update(g);
/* inform edp about new constraint */ /* inform edp about new constraint */
if (platform->prescale) if (platform->prescale)
platform->prescale(pdev); platform->prescale(pdev);

1
drivers/gpu/nvgpu/gk20a/gk20a_sysfs.c
View File

@@ -278,6 +278,7 @@ static ssize_t gk20a_load_show(struct device *dev,
busy_time = 0; busy_time = 0;
} else { } else {
gk20a_busy(g->dev); gk20a_busy(g->dev);
gk20a_pmu_load_update(g);
gk20a_pmu_load_norm(g, &busy_time); gk20a_pmu_load_norm(g, &busy_time);
gk20a_idle(g->dev); gk20a_idle(g->dev);
} }

10
drivers/gpu/nvgpu/gk20a/pmu_gk20a.c
View File

@@ -3676,17 +3676,23 @@ int gk20a_pmu_destroy(struct gk20a *g)
} }
int gk20a_pmu_load_norm(struct gk20a *g, u32 *load) 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; struct pmu_gk20a *pmu = &g->pmu;
u16 _load = 0; u16 _load = 0;
if (!pmu->perfmon_ready) { if (!pmu->perfmon_ready) {
*load = 0; pmu->load_shadow = 0;
return 0; return 0;
} }
pmu_copy_from_dmem(pmu, pmu->sample_buffer, (u8 *)&_load, 2, 0); pmu_copy_from_dmem(pmu, pmu->sample_buffer, (u8 *)&_load, 2, 0);
*load = _load / 10; pmu->load_shadow = _load / 10;
return 0; return 0;
} }

2
drivers/gpu/nvgpu/gk20a/pmu_gk20a.h
View File

@@ -1079,6 +1079,7 @@ struct pmu_gk20a {
bool perfmon_ready; bool perfmon_ready;
u32 sample_buffer; u32 sample_buffer;
u32 load_shadow;
struct mutex isr_mutex; struct mutex isr_mutex;
struct mutex isr_enable_lock; 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 pmu_mutex_release(struct pmu_gk20a *pmu, u32 id, u32 *token);
int gk20a_pmu_destroy(struct gk20a *g); int gk20a_pmu_destroy(struct gk20a *g);
int gk20a_pmu_load_norm(struct gk20a *g, u32 *load); 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); int gk20a_pmu_debugfs_init(struct platform_device *dev);
void gk20a_pmu_reset_load_counters(struct gk20a *g); void gk20a_pmu_reset_load_counters(struct gk20a *g);
void gk20a_pmu_get_load_counters(struct gk20a *g, u32 *busy_cycles, void gk20a_pmu_get_load_counters(struct gk20a *g, u32 *busy_cycles,