moar sysfs stuff
Subsystems (2)
cover moar sysfs stuff
Not very thoroughly tested yet. I'll be adding igt tests ASAP.
Ben Widawsky (6):
drm/i915: Enable some sysfs stuff without CONFIG_PM drm/i915: variable renames
drm/i915: #define gpu freq multipler drm/i915: Add current GPU freq to sysfs drm/i915: Add setters for min/max frequency drm/i915: Show render P state thresholds in sysfs
Jesse Barnes (1):
drm/i915: expose energy counter on SNB and IVB
drivers/gpu/drm/i915/i915_debugfs.c | 22 ++-- drivers/gpu/drm/i915/i915_drv.h | 2 + drivers/gpu/drm/i915/i915_reg.h | 2 + drivers/gpu/drm/i915/i915_sysfs.c | 237 ++++++++++++++++++++++++++++++++---- 4 files changed, 230 insertions(+), 33 deletions(-)
Signature · 1 line
1.7.12
PATCH 1/7 drm/i915: Enable some sysfs stuff without CONFIG_PM
The original patch was actually incorrect in stubbing out the sysfs for l3 parity.
commit 5ab3633d6907018b0b830a720e877c3884d679c3
Author: Hunt Xu <[email protected]> Date: Sun Jul 1 03:45:07 2012 +0000
drm/i915: make rc6 in sysfs functions conditional
Unfortunately Hunt didn't respond to my review comments, and Daniel sucked in the patch again ignoring. Worst of all, I'm too lazy to write the patch for what I originally wanted, which was to keep rc6 sysfs even without CONFIG_PM. This simpler patch does enough to enable us to add more sysfs entries though.
NOTE: This is already picked up in dinq, but is here in case anyone is looking to test this series on it's own.
--- drivers/gpu/drm/i915/i915_sysfs.c | 15 +++------------ 1 file changed, 3 insertions(+), 12 deletions(-)
drivers/gpu/drm/i915/i915_sysfs.c +3 -12
@@ -93,6 +93,7 @@ static struct attribute_group rc6_attr_group = { .name = power_group_name, .attrs = rc6_attrs }; +#endif static int l3_access_valid (struct drm_device *dev,loff_t offset) {
@@ -206,13 +207,14 @@ void i915_setup_sysfs(struct drm_device *dev) {int ret; +#ifdef CONFIG_PM if (INTEL_INFO (dev)->gen >=6 ) { ret =sysfs_merge_group (&dev->primary->kdev.kobj, &rc6_attr_group);if (ret)DRM_ERROR ("RC6 residency sysfs setup failed\n" ); } - +#endif if (HAS_L3_GPU_CACHE (dev)) { ret =device_create_bin_file (&dev->primary->kdev, &dpf_attrs);if (ret)
@@ -225,14 +227,3 @@ void i915_teardown_sysfs(struct drm_device *dev)device_remove_bin_file (&dev->primary->kdev, &dpf_attrs);sysfs_unmerge_group (&dev->primary->kdev.kobj, &rc6_attr_group); } -#else -void i915_setup_sysfs (struct drm_device *dev) -{ -return ; -} - -void i915_teardown_sysfs (struct drm_device *dev) -{ -return ; -} -#endif /* CONFIG_PM */
Signature · 1 line
1.7.12
PATCH 2/7 drm/i915: variable renames
In reply to Ben WidawskyFrom: Ben Widawsky <[email protected]>
Name variables a bit better for copy-pasters. This got turned up as part of review for upcoming sysfs patches.
--- drivers/gpu/drm/i915/i915_sysfs.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)
drivers/gpu/drm/i915/i915_sysfs.c +8 -8
@@ -46,32 +46,32 @@ static u32 calc_residency(struct drm_device *dev, const u32 reg) }static ssize_t -show_rc6_mask (struct device *dev,struct device_attribute *attr,char *buf) +show_rc6_mask (struct device *kdev,struct device_attribute *attr,char *buf) { - struct drm_minor *dminor = container_of(dev, struct drm_minor, kdev); + struct drm_minor *dminor = container_of(kdev, struct drm_minor, kdev);return snprintf (buf,PAGE_SIZE ,"%x" ,intel_enable_rc6 (dminor->dev)); }static ssize_t -show_rc6_ms (struct device *dev,struct device_attribute *attr,char *buf) +show_rc6_ms (struct device *kdev,struct device_attribute *attr,char *buf) { - struct drm_minor *dminor = container_of(dev, struct drm_minor, kdev); + struct drm_minor *dminor = container_of(kdev, struct drm_minor, kdev);u32 rc6_residency =calc_residency (dminor->dev,GEN6_GT_GFX_RC6 );return snprintf (buf,PAGE_SIZE ,"%u" , rc6_residency); }static ssize_t -show_rc6p_ms (struct device *dev,struct device_attribute *attr,char *buf) +show_rc6p_ms (struct device *kdev,struct device_attribute *attr,char *buf) { - struct drm_minor *dminor = container_of(dev, struct drm_minor, kdev); + struct drm_minor *dminor = container_of(kdev, struct drm_minor, kdev);u32 rc6p_residency =calc_residency (dminor->dev, GEN6_GT_GFX_RC6p);return snprintf (buf,PAGE_SIZE ,"%u" , rc6p_residency); }static ssize_t -show_rc6pp_ms (struct device *dev,struct device_attribute *attr,char *buf) +show_rc6pp_ms (struct device *kdev,struct device_attribute *attr,char *buf) { - struct drm_minor *dminor = container_of(dev, struct drm_minor, kdev); + struct drm_minor *dminor = container_of(kdev, struct drm_minor, kdev);u32 rc6pp_residency =calc_residency (dminor->dev, GEN6_GT_GFX_RC6pp);return snprintf (buf,PAGE_SIZE ,"%u" , rc6pp_residency); }
Signature · 1 line
1.7.12
PATCH 3/7 drm/i915: #define gpu freq multipler
In reply to Ben WidawskyMagic numbers are bad mmmkay. In this case in particular the value is especially weird because the docs say multiple things. We'll need this value for sysfs, so extracting it is useful for that as well.
--- drivers/gpu/drm/i915/i915_debugfs.c | 22 +++++++++++----------- drivers/gpu/drm/i915/i915_drv.h | 2 ++ 2 files changed, 13 insertions(+), 11 deletions(-)
drivers/gpu/drm/i915/i915_debugfs.c +11 -11
@@ -930,7 +930,7 @@ static int i915_cur_delayinfo(struct seq_file *m, void *unused)seq_printf (m,"Render p-state limit: %d\n" , rp_state_limits &0xff );seq_printf (m,"CAGF: %dMHz\n" , ((rpstat &GEN6_CAGF_MASK ) >> -GEN6_CAGF_SHIFT ) *50 ); +GEN6_CAGF_SHIFT ) *GT_FREQUENCY_MULTIPLIER );seq_printf (m,"RP CUR UP EI: %dus\n" , rpupei &GEN6_CURICONT_MASK ); seq_printf(m, "RP CUR UP: %dus\n", rpcurup &
@@ -946,15 +946,15 @@ static int i915_cur_delayinfo(struct seq_file *m, void *unused) max_freq = (rp_state_cap &0xff0000 ) >>16 ;seq_printf (m,"Lowest (RPN) frequency: %dMHz\n" , - max_freq *50 ); + max_freq *GT_FREQUENCY_MULTIPLIER ); max_freq = (rp_state_cap &0xff00 ) >>8 ;seq_printf (m,"Nominal (RP1) frequency: %dMHz\n" , - max_freq *50 ); + max_freq *GT_FREQUENCY_MULTIPLIER ); max_freq = rp_state_cap &0xff ;seq_printf (m,"Max non-overclocked (RP0) frequency: %dMHz\n" , - max_freq *50 ); + max_freq *GT_FREQUENCY_MULTIPLIER ); }else {seq_printf (m,"no P-state info available\n" ); }
@@ -1308,7 +1308,7 @@ static int i915_ring_freq_table(struct seq_file *m, void *unused)continue ; } ia_freq =I915_READ (GEN6_PCODE_DATA ); -seq_printf (m,"%d\t\t%d\n" , gpu_freq *50 , ia_freq *100 ); +seq_printf (m,"%d\t\t%d\n" , gpu_freq *GT_FREQUENCY_MULTIPLIER , ia_freq *100 ); }mutex_unlock (&dev->struct_mutex);
@@ -1735,7 +1735,7 @@ i915_max_freq_read(struct file *filp,return ret; len =snprintf (buf,sizeof (buf), -"max freq: %d\n" , dev_priv->rps.max_delay *50 ); +"max freq: %d\n" , dev_priv->rps.max_delay *GT_FREQUENCY_MULTIPLIER );mutex_unlock (&dev->struct_mutex);if (len >sizeof (buf))
@@ -1778,9 +1778,9 @@ i915_max_freq_write(struct file *filp,/* * Turbo will still be enabled, but won't go above the set value. */ - dev_priv->rps.max_delay = val /50 ; + dev_priv->rps.max_delay = val /GT_FREQUENCY_MULTIPLIER ; -gen6_set_rps (dev, val /50 ); +gen6_set_rps (dev, val /GT_FREQUENCY_MULTIPLIER );mutex_unlock (&dev->struct_mutex);return cnt;
@@ -1811,7 +1811,7 @@ i915_min_freq_read(struct file *filp, char __user *ubuf, size_t max,return ret; len =snprintf (buf,sizeof (buf), -"min freq: %d\n" , dev_priv->rps.min_delay *50 ); +"min freq: %d\n" , dev_priv->rps.min_delay *GT_FREQUENCY_MULTIPLIER );mutex_unlock (&dev->struct_mutex);if (len >sizeof (buf))
@@ -1852,9 +1852,9 @@ i915_min_freq_write(struct file *filp, const char __user *ubuf, size_t cnt,/* * Turbo will still be enabled, but won't go below the set value. */ - dev_priv->rps.min_delay = val /50 ; + dev_priv->rps.min_delay = val /GT_FREQUENCY_MULTIPLIER ; -gen6_set_rps (dev, val /50 ); +gen6_set_rps (dev, val /GT_FREQUENCY_MULTIPLIER );mutex_unlock (&dev->struct_mutex);return cnt;
drivers/gpu/drm/i915/i915_drv.h +2 -0
@@ -1162,6 +1162,8 @@ struct drm_i915_file_private {#define HAS_L3_GPU_CACHE (dev) (IS_IVYBRIDGE(dev) || IS_HASWELL(dev)) +#define GT_FREQUENCY_MULTIPLIER 50 +#include "i915_trace.h" /**
Signature · 1 line
1.7.12
Re: [PATCH 3/7] drm/i915: #define gpu freq multipler
Quoted 3 lines · On Thu, 6 Sep 2012 13:54:06 -0700 Ben Widawsky <[email protected]> wrote:
Magic numbers are bad mmmkay. In this case in particular the value is especially weird because the docs say multiple things. We'll need this value for sysfs, so extracting it is useful for that as well.
Signature · 1 line
Jesse Barnes, Intel Open Source Technology Center
PATCH v2 4/7 drm/i915: Add current GPU freq to sysfs
In reply to Ben WidawskyUserspace applications such as PowerTOP are interesting in being able to read the current GPU frequency. The patch itself sets up a generic array for gen6 attributes so we can easily add other items in the future (and it also happens to be just about the cleanest way to do this).
The patch is a nice addition to
commit 1ac02185dff3afac146d745ba220dc6672d1d162
Author: Daniel Vetter <[email protected]> Date: Thu Aug 30 13:26:48 2012 +0200
drm/i915: add a tracepoint for gpu frequency changes
Reading the GPU frequncy can be done by reading a file like:/sys/class/drm/card0/render_frequency_mhz
v2: rename the sysfs file to gt_cur_freq_mhz (Daniel) Use kdev naming (Chris)
Use DEVICE_ATTR over __ATTR (Ben)
Add min/max freq (Ben/Daniel)
user the new #define for frequency multiplier
--- drivers/gpu/drm/i915/i915_sysfs.c | 71 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 71 insertions(+)
drivers/gpu/drm/i915/i915_sysfs.c +71 -0
@@ -203,6 +203,70 @@ static struct bin_attribute dpf_attrs = { .mmap = NULL }; +static ssize_t gt_cur_freq_mhz_show (struct device *kdev, +struct device_attribute *attr,char *buf) +{ + struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev); +struct drm_device *dev = minor->dev; +struct drm_i915_private *dev_priv = dev->dev_private; +int ret; + + ret =i915_mutex_lock_interruptible (dev); +if (ret) +return ret; + + ret = dev_priv->rps.cur_delay *GT_FREQUENCY_MULTIPLIER ; +mutex_unlock (&dev->struct_mutex); + +return snprintf (buf,PAGE_SIZE ,"%d" , ret); +} + +static ssize_t gt_max_freq_mhz_show (struct device *kdev,struct device_attribute *attr,char *buf) +{ + struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev); +struct drm_device *dev = minor->dev; +struct drm_i915_private *dev_priv = dev->dev_private; +int ret; + + ret =i915_mutex_lock_interruptible (dev); +if (ret) +return ret; + + ret = dev_priv->rps.max_delay *GT_FREQUENCY_MULTIPLIER ; +mutex_unlock (&dev->struct_mutex); + +return snprintf (buf,PAGE_SIZE ,"%d" , ret); +} + +static ssize_t gt_min_freq_mhz_show (struct device *kdev,struct device_attribute *attr,char *buf) +{ + struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev); +struct drm_device *dev = minor->dev; +struct drm_i915_private *dev_priv = dev->dev_private; +int ret; + + ret =i915_mutex_lock_interruptible (dev); +if (ret) +return ret; + + ret = dev_priv->rps.min_delay *GT_FREQUENCY_MULTIPLIER ; +mutex_unlock (&dev->struct_mutex); + +return snprintf (buf,PAGE_SIZE ,"%d" , ret); +} + +static DEVICE_ATTR(gt_cur_freq_mhz, S_IRUGO, gt_cur_freq_mhz_show, NULL); +static DEVICE_ATTR(gt_max_freq_mhz, S_IRUGO | S_IWUSR, gt_max_freq_mhz_show, NULL); +static DEVICE_ATTR(gt_min_freq_mhz, S_IRUGO | S_IWUSR, gt_min_freq_mhz_show, NULL); + +static const struct attribute *gen6_attrs[] = { + &dev_attr_gt_cur_freq_mhz.attr, + &dev_attr_gt_max_freq_mhz.attr, + &dev_attr_gt_min_freq_mhz.attr, +NULL , +}; + +void i915_setup_sysfs (struct drm_device *dev) {int ret;
@@ -220,10 +284,17 @@ void i915_setup_sysfs(struct drm_device *dev)if (ret)DRM_ERROR ("l3 parity sysfs setup failed\n" ); } + +if (INTEL_INFO (dev)->gen >=6 ) { + ret =sysfs_create_files (&dev->primary->kdev.kobj, gen6_attrs); +if (ret) +DRM_ERROR ("gen6 sysfs setup failed\n" ); + } }void i915_teardown_sysfs (struct drm_device *dev) { +sysfs_remove_files (&dev->primary->kdev.kobj, gen6_attrs);device_remove_bin_file (&dev->primary->kdev, &dpf_attrs);sysfs_unmerge_group (&dev->primary->kdev.kobj, &rc6_attr_group); }
Signature · 1 line
1.7.12
A message that is not in the archive (sent off-list or not yet received): [email protected]
Re: [PATCH 4/7 v2] drm/i915: Add current GPU freq to sysfs
Quoted 9 lines · Michael Larabel
Have you discussed this with Radeon/Nouveau to see if they would adopt a common sysfs interface? In the past, Eugeni Dodonov and I had talked about this with a desire to have the open-source drivers expose the useful information for monitoring in a similar manner for helping userspace (current/max/min clock speeds, voltage, etc). [My personal interest is from monitoring this information in a more standardized way from the Phoronix Test Suite system monitoring components rather than writing driver-specific paths like it is now.]
Quoted 4 lines
Martin (CCed) has also expressed interest from the Nouveau side in calling for some standardization.
-- Michael
We can... I did very briefly look at what Nouveau does with the hwmon interface, and what DRM provided, and thought it'd be more work than I wanted to sign up for (which is just about nothing sysfs other than display). Thanks for bringing him into the thread though.
Quoted 130 lines · Ben Widawsky
Quoted 128 lines
Userspace applications such as PowerTOP are interesting in being able to read the current GPU frequency. The patch itself sets up a generic array for gen6 attributes so we can easily add other items in the future (and it also happens to be just about the cleanest way to do this).
The patch is a nice addition to
commit1ac02185dff3afac146d745ba220dc6672d1d162Author: Daniel Vetter <[email protected]> Date: Thu Aug 30 13:26:48 2012 +0200drm/i915: add a tracepoint for gpu frequency changes
Reading the GPU frequncy can be done by reading a file like:
/sys/class/drm/card0/render_frequency_mhzv2: rename the sysfs file to
gt_cur_freq_mhz(Daniel) Use kdev naming (Chris)
UseDEVICE_ATTRover__ATTR(Ben)
Add min/max freq (Ben/Daniel)
user the new #define for frequency multiplierCC Arjan van de Ven <[email protected]>CC Jacob Pan <[email protected]>Signed-off-by Ben Widawsky <[email protected]>--- drivers/gpu/drm/i915/i915_sysfs.c | 71 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 71 insertions(+)drivers/gpu/drm/i915/i915_sysfs.c +35
-0@@ -203,6 +203,70 @@ static struct bin_attribute dpf_attrs = { .mmap = NULL }; +static ssize_t gt_cur_freq_mhz_show(struct device *kdev, +struct device_attribute *attr,char *buf) +{ + struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev); +struct drm_device *dev = minor->dev; +struct drm_i915_private *dev_priv = dev->dev_private; +int ret; + + ret =i915_mutex_lock_interruptible (dev); +if (ret) +return ret; + + ret = dev_priv->rps.cur_delay *GT_FREQUENCY_MULTIPLIER ; +mutex_unlock (&dev->struct_mutex); + +return snprintf (buf,PAGE_SIZE ,"%d" , ret); +} + +static ssize_t gt_max_freq_mhz_show (struct device *kdev,struct device_attribute *attr,char *buf) +{ + struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev); +struct drm_device *dev = minor->dev; +struct drm_i915_private *dev_priv = dev->dev_private; +int ret; + + ret =i915_mutex_lock_interruptible (dev); +if (ret) +return ret; + + ret = dev_priv->rps.max_delay *GT_FREQUENCY_MULTIPLIER ; +mutex_unlock (&dev->struct_mutex); + +return snprintf (buf,PAGE_SIZE ,"%d" , ret); +} + +static ssize_t gt_min_freq_mhz_show(struct device *kdev, struct
device_attribute*attr, char *buf)drivers/gpu/drm/i915/i915_sysfs.c +2
-0+{ + struct drm_minor *minor = container_of(kdev, struct drm_minor,kdev);
drivers/gpu/drm/i915/i915_sysfs.c +15
-0+struct drm_device *dev = minor->dev; +struct drm_i915_private *dev_priv = dev->dev_private; +int ret; + + ret =i915_mutex_lock_interruptible (dev); +if (ret) +return ret; + + ret = dev_priv->rps.min_delay *GT_FREQUENCY_MULTIPLIER ; +mutex_unlock (&dev->struct_mutex); + +return snprintf (buf,PAGE_SIZE ,"%d" , ret); +} + +static DEVICE_ATTR(gt_cur_freq_mhz, S_IRUGO, gt_cur_freq_mhz_show,NULL);
+staticDEVICE_ATTR(gt_max_freq_mhz, S_IRUGO | S_IWUSR, gt_max_freq_mhz_show, NULL);
+staticDEVICE_ATTR(gt_min_freq_mhz, S_IRUGO | S_IWUSR, gt_min_freq_mhz_show, NULL);drivers/gpu/drm/i915/i915_sysfs.c +16
-0+ +static const struct attribute *gen6_attrs[] = { + &dev_attr_gt_cur_freq_mhz.attr, + &dev_attr_gt_max_freq_mhz.attr, + &dev_attr_gt_min_freq_mhz.attr, +NULL , +}; + +void i915_setup_sysfs (struct drm_device *dev) {int ret;@@ -220,10 +284,17 @@ void i915_setup_sysfs(struct drm_device *dev)if (ret)DRM_ERROR ("l3 parity sysfs setup failed\n" ); } + +if (INTEL_INFO (dev)->gen >=6 ) { + ret =sysfs_create_files (&dev->primary->kdev.kobj, gen6_attrs); +if (ret) +DRM_ERROR ("gen6 sysfs setup failed\n" ); + } }void i915_teardown_sysfs (struct drm_device *dev) { +sysfs_remove_files (&dev->primary->kdev.kobj, gen6_attrs);device_remove_bin_file (&dev->primary->kdev, &dpf_attrs);sysfs_unmerge_group (&dev->primary->kdev.kobj, &rc6_attr_group); }
Signature · 1 line
Ben Widawsky, Intel Open Source Technology Center
PATCH 5/7 drm/i915: Add setters for min/max frequency
In reply to Ben WidawskyProvide a standardized sysfs interface for setting min, and max frequencies. The code which reads the limits were lifted from the debugfs files. As a brief explanation, the limits are similar to the CPU p-states. We have 3 states:
RP0 - ie. max frequency
RP1 - ie. "local min" frequency
RPn - seriously lowest frequency
Initially Daniel asked me to clamp the writes to supported values, but in conforming to the way the cpufreq drivers seem to work, instead return -EINVAL (noticed by Jesse in discussion).
The values can be used by userspace wishing to control the limits of the GPU (see the CC list for people who care).
--- drivers/gpu/drm/i915/i915_sysfs.c | 78 ++++++++++++++++++++++++++++++++++++++- 1 file changed, 76 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/i915_sysfs.c +76 -2
@@ -238,6 +238,43 @@ static ssize_t gt_max_freq_mhz_show(struct device *kdev, struct device_attributereturn snprintf (buf,PAGE_SIZE ,"%d" , ret); } +static ssize_t gt_max_freq_mhz_store (struct device *kdev, +struct device_attribute *attr, +const char *buf,size_t count) +{ + struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev); +struct drm_device *dev = minor->dev; +struct drm_i915_private *dev_priv = dev->dev_private; +u32 val, rp_state_cap, hw_max; +ssize_t ret; + + ret =kstrtou32 (buf,0 , &val); +if (ret) +return ret; + + ret =mutex_lock_interruptible (&dev->struct_mutex); +if (ret) +return ret; + + rp_state_cap =I915_READ (GEN6_RP_STATE_CAP ); + hw_max = (rp_state_cap &0xff ) *GT_FREQUENCY_MULTIPLIER ; + +if (val > hw_max) +return -EINVAL ; + +if (val < dev_priv->rps.min_delay) + val = dev_priv->rps.min_delay; + +if (dev_priv->rps.cur_delay > val) +gen6_set_rps (dev_priv->dev, val); + + dev_priv->rps.max_delay = val /GT_FREQUENCY_MULTIPLIER ; + +mutex_unlock (&dev->struct_mutex); + +return count; +} +static ssize_t gt_min_freq_mhz_show (struct device *kdev,struct device_attribute *attr,char *buf) { struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev);
@@ -255,9 +292,46 @@ static ssize_t gt_min_freq_mhz_show(struct device *kdev, struct device_attributereturn snprintf (buf,PAGE_SIZE ,"%d" , ret); } +static ssize_t gt_min_freq_mhz_store (struct device *kdev, +struct device_attribute *attr, +const char *buf,size_t count) +{ + struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev); +struct drm_device *dev = minor->dev; +struct drm_i915_private *dev_priv = dev->dev_private; +u32 val, rp_state_cap, hw_min; +ssize_t ret; + + ret =kstrtou32 (buf,0 , &val); +if (ret) +return ret; + + ret =mutex_lock_interruptible (&dev->struct_mutex); +if (ret) +return ret; + + rp_state_cap =I915_READ (GEN6_RP_STATE_CAP ); + hw_min = ((rp_state_cap &0xff0000 ) >>16 ) *GT_FREQUENCY_MULTIPLIER ; + +if (val < hw_min) +return -EINVAL ; + +if (val > dev_priv->rps.max_delay) + val = dev_priv->rps.max_delay; + +if (dev_priv->rps.cur_delay < val) +gen6_set_rps (dev_priv->dev, val); + + dev_priv->rps.min_delay = val /GT_FREQUENCY_MULTIPLIER ; + +mutex_unlock (&dev->struct_mutex); + +return count; +} + static DEVICE_ATTR(gt_cur_freq_mhz, S_IRUGO, gt_cur_freq_mhz_show, NULL); -static DEVICE_ATTR(gt_max_freq_mhz, S_IRUGO | S_IWUSR, gt_max_freq_mhz_show, NULL); -static DEVICE_ATTR(gt_min_freq_mhz, S_IRUGO | S_IWUSR, gt_min_freq_mhz_show, NULL); +static DEVICE_ATTR(gt_max_freq_mhz, S_IRUGO | S_IWUSR, gt_max_freq_mhz_show, gt_max_freq_mhz_store); +static DEVICE_ATTR(gt_min_freq_mhz, S_IRUGO | S_IWUSR, gt_min_freq_mhz_show, gt_min_freq_mhz_store);static const struct attribute *gen6_attrs[] = { &dev_attr_gt_cur_freq_mhz.attr,
Signature · 1 line
1.7.12
Re: [PATCH 5/7] drm/i915: Add setters for min/max frequency
Quoted 7 lines · On Thu, 6 Sep 2012 13:54:08 -0700 Ben Widawsky <[email protected]> wrote:
Provide a standardized sysfs interface for setting min, and max frequencies. The code which reads the limits were lifted from the debugfs files. As a brief explanation, the limits are similar to the CPU p-states. We have 3 states:
RP0 - ie. max frequency
RP1 - ie. "local min" frequency
I'd say "preferred min" or "vmin" here instead.
Quoted 8 lines
RPn - seriously lowest frequency
Initially Daniel asked me to clamp the writes to supported values, but in conforming to the way the cpufreq drivers seem to work, instead return
-EINVAL(noticed by Jesse in discussion).The values can be used by userspace wishing to control the limits of the GPU (see the CC list for people who care).
Can you add this and any other i915 bits that are missing to the sysfs documentation under Documentation/?
Thanks for moving to -EINVAL, it makes a lot more sense to me.
Signature · 1 line
Jesse Barnes, Intel Open Source Technology Center
PATCH v2 5/7 drm/i915: Add setters for min/max frequency
In reply to Ben WidawskyProvide a standardized sysfs interface for setting min, and max frequencies. The code which reads the limits were lifted from the debugfs files. As a brief explanation, the limits are similar to the CPU p-states. We have 3 states:
RP0 - ie. max frequency
RP1 - ie. "preferred min" frequency
RPn - seriously lowest frequency
Initially Daniel asked me to clamp the writes to supported values, but in conforming to the way the cpufreq drivers seem to work, instead return -EINVAL (noticed by Jesse in discussion).
The values can be used by userspace wishing to control the limits of the GPU (see the CC list for people who care).
v2: add the dropped mutex_unlock in error cases (Chris) EINVAL on both too min, or too max (Daniel)
--- drivers/gpu/drm/i915/i915_sysfs.c | 84 ++++++++++++++++++++++++++++++++++++++- 1 file changed, 82 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/i915_sysfs.c +82 -2
@@ -238,6 +238,46 @@ static ssize_t gt_max_freq_mhz_show(struct device *kdev, struct device_attributereturn snprintf (buf,PAGE_SIZE ,"%d" , ret); } +static ssize_t gt_max_freq_mhz_store (struct device *kdev, +struct device_attribute *attr, +const char *buf,size_t count) +{ + struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev); +struct drm_device *dev = minor->dev; +struct drm_i915_private *dev_priv = dev->dev_private; +u32 val, rp_state_cap, hw_max, hw_min; +ssize_t ret; + + ret =kstrtou32 (buf,0 , &val); +if (ret) +return ret; + + ret =mutex_lock_interruptible (&dev->struct_mutex); +if (ret) +return ret; + + rp_state_cap =I915_READ (GEN6_RP_STATE_CAP ); + hw_max = (rp_state_cap &0xff ) *GT_FREQUENCY_MULTIPLIER ; + hw_min = ((rp_state_cap &0xff0000 ) >>16 ) *GT_FREQUENCY_MULTIPLIER ; + +if (val < hw_min || val > hw_max) { +mutex_unlock (&dev->struct_mutex); +return -EINVAL ; + } + +if (val < dev_priv->rps.min_delay) + val = dev_priv->rps.min_delay; + +if (dev_priv->rps.cur_delay > val) +gen6_set_rps (dev_priv->dev, val); + + dev_priv->rps.max_delay = val /GT_FREQUENCY_MULTIPLIER ; + +mutex_unlock (&dev->struct_mutex); + +return count; +} +static ssize_t gt_min_freq_mhz_show (struct device *kdev,struct device_attribute *attr,char *buf) { struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev);
@@ -255,9 +295,49 @@ static ssize_t gt_min_freq_mhz_show(struct device *kdev, struct device_attributereturn snprintf (buf,PAGE_SIZE ,"%d" , ret); } +static ssize_t gt_min_freq_mhz_store (struct device *kdev, +struct device_attribute *attr, +const char *buf,size_t count) +{ + struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev); +struct drm_device *dev = minor->dev; +struct drm_i915_private *dev_priv = dev->dev_private; +u32 val, rp_state_cap, hw_max, hw_min; +ssize_t ret; + + ret =kstrtou32 (buf,0 , &val); +if (ret) +return ret; + + ret =mutex_lock_interruptible (&dev->struct_mutex); +if (ret) +return ret; + + rp_state_cap =I915_READ (GEN6_RP_STATE_CAP ); + hw_max = (rp_state_cap &0xff ) *GT_FREQUENCY_MULTIPLIER ; + hw_min = ((rp_state_cap &0xff0000 ) >>16 ) *GT_FREQUENCY_MULTIPLIER ; + +if (val < hw_min || val > hw_max) { +mutex_unlock (&dev->struct_mutex); +return -EINVAL ; + } + +if (val > dev_priv->rps.max_delay) + val = dev_priv->rps.max_delay; + +if (dev_priv->rps.cur_delay < val) +gen6_set_rps (dev_priv->dev, val); + + dev_priv->rps.min_delay = val /GT_FREQUENCY_MULTIPLIER ; + +mutex_unlock (&dev->struct_mutex); + +return count; +} + static DEVICE_ATTR(gt_cur_freq_mhz, S_IRUGO, gt_cur_freq_mhz_show, NULL); -static DEVICE_ATTR(gt_max_freq_mhz, S_IRUGO | S_IWUSR, gt_max_freq_mhz_show, NULL); -static DEVICE_ATTR(gt_min_freq_mhz, S_IRUGO | S_IWUSR, gt_min_freq_mhz_show, NULL); +static DEVICE_ATTR(gt_max_freq_mhz, S_IRUGO | S_IWUSR, gt_max_freq_mhz_show, gt_max_freq_mhz_store); +static DEVICE_ATTR(gt_min_freq_mhz, S_IRUGO | S_IWUSR, gt_min_freq_mhz_show, gt_min_freq_mhz_store);static const struct attribute *gen6_attrs[] = { &dev_attr_gt_cur_freq_mhz.attr,
Signature · 1 line
1.7.12
Re: [PATCH 5/7 v2] drm/i915: Add setters for min/max frequency
The only bit of the series I don't like is the conversion to uJ which looks like it can overflow. Otherwise,
-Chris
Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
PATCH 6/7 drm/i915: Show render P state thresholds in sysfs
In reply to Ben WidawskyThis is useful for userspace utilities which wish to use the previous interface, specifically for micromanaging the increase/decrease steps by setting min == max.
--- drivers/gpu/drm/i915/i915_sysfs.c | 36 ++++++++++++++++++++++++++++++++++++ 1 file changed, 36 insertions(+)
drivers/gpu/drm/i915/i915_sysfs.c +36 -0
@@ -333,10 +333,46 @@ static DEVICE_ATTR(gt_cur_freq_mhz, S_IRUGO, gt_cur_freq_mhz_show, NULL); static DEVICE_ATTR(gt_max_freq_mhz, S_IRUGO | S_IWUSR, gt_max_freq_mhz_show, gt_max_freq_mhz_store); static DEVICE_ATTR(gt_min_freq_mhz, S_IRUGO | S_IWUSR, gt_min_freq_mhz_show, gt_min_freq_mhz_store); + +static ssize_t gt_rp_mhz_show (struct device *kdev,struct device_attribute *attr,char *buf); +static DEVICE_ATTR(gt_RP0_freq_mhz, S_IRUGO, gt_rp_mhz_show, NULL); +static DEVICE_ATTR(gt_RP1_freq_mhz, S_IRUGO, gt_rp_mhz_show, NULL); +static DEVICE_ATTR(gt_RPn_freq_mhz, S_IRUGO, gt_rp_mhz_show, NULL); + +/* For now we have a static number of RP states */ +static ssize_t gt_rp_mhz_show (struct device *kdev,struct device_attribute *attr,char *buf) +{ + struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev); +struct drm_device *dev = minor->dev; +struct drm_i915_private *dev_priv = dev->dev_private; +u32 val, rp_state_cap; +ssize_t ret; + + ret =mutex_lock_interruptible (&dev->struct_mutex); +if (ret) +return ret; + rp_state_cap =I915_READ (GEN6_RP_STATE_CAP ); +mutex_unlock (&dev->struct_mutex); + +if (attr == &dev_attr_gt_RP0_freq_mhz) { + val = ((rp_state_cap &0x0000ff ) >>0 ) *GT_FREQUENCY_MULTIPLIER ; + }else if (attr == &dev_attr_gt_RP1_freq_mhz) { + val = ((rp_state_cap &0x00ff00 ) >>8 ) *GT_FREQUENCY_MULTIPLIER ; + }else if (attr == &dev_attr_gt_RPn_freq_mhz) { + val = ((rp_state_cap &0xff0000 ) >>16 ) *GT_FREQUENCY_MULTIPLIER ; + }else { +BUG (); + } +return snprintf (buf,PAGE_SIZE ,"%d" , val); +} +static const struct attribute *gen6_attrs[] = { &dev_attr_gt_cur_freq_mhz.attr, &dev_attr_gt_max_freq_mhz.attr, &dev_attr_gt_min_freq_mhz.attr, + &dev_attr_gt_RP0_freq_mhz.attr, + &dev_attr_gt_RP1_freq_mhz.attr, + &dev_attr_gt_RPn_freq_mhz.attr,NULL , };
Signature · 1 line
1.7.12
Re: [PATCH 6/7] drm/i915: Show render P state thresholds in sysfs
Quoted 52 lines · On Thu, 6 Sep 2012 13:54:09 -0700 Ben Widawsky <[email protected]> wrote:
This is useful for userspace utilities which wish to use the previous interface, specifically for micromanaging the increase/decrease steps by setting min == max.
CC Jacob Pan <[email protected]>CC Jesse Barnes <[email protected]>Signed-off-by Ben Widawsky <[email protected]>--- drivers/gpu/drm/i915/i915_sysfs.c | 36 ++++++++++++++++++++++++++++++++++++ 1 file changed, 36 insertions(+)drivers/gpu/drm/i915/i915_sysfs.c +33
-0@@ -333,10 +333,46 @@ static DEVICE_ATTR(gt_cur_freq_mhz, S_IRUGO, gt_cur_freq_mhz_show, NULL); static DEVICE_ATTR(gt_max_freq_mhz, S_IRUGO | S_IWUSR, gt_max_freq_mhz_show, gt_max_freq_mhz_store); static DEVICE_ATTR(gt_min_freq_mhz, S_IRUGO | S_IWUSR, gt_min_freq_mhz_show, gt_min_freq_mhz_store); + +static ssize_t gt_rp_mhz_show (struct device *kdev,struct device_attribute *attr,char *buf); +static DEVICE_ATTR(gt_RP0_freq_mhz, S_IRUGO, gt_rp_mhz_show, NULL); +static DEVICE_ATTR(gt_RP1_freq_mhz, S_IRUGO, gt_rp_mhz_show, NULL); +static DEVICE_ATTR(gt_RPn_freq_mhz, S_IRUGO, gt_rp_mhz_show, NULL); + +/* For now we have a static number of RP states */ +static ssize_t gt_rp_mhz_show (struct device *kdev,struct device_attribute *attr,char *buf) +{ + struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev); +struct drm_device *dev = minor->dev; +struct drm_i915_private *dev_priv = dev->dev_private; +u32 val, rp_state_cap; +ssize_t ret; + + ret =mutex_lock_interruptible (&dev->struct_mutex); +if (ret) +return ret; + rp_state_cap =I915_READ (GEN6_RP_STATE_CAP ); +mutex_unlock (&dev->struct_mutex); + +if (attr == &dev_attr_gt_RP0_freq_mhz) { + val = ((rp_state_cap &0x0000ff ) >>0 ) *GT_FREQUENCY_MULTIPLIER ; + }else if (attr == &dev_attr_gt_RP1_freq_mhz) { + val = ((rp_state_cap &0x00ff00 ) >>8 ) *GT_FREQUENCY_MULTIPLIER ; + }else if (attr == &dev_attr_gt_RPn_freq_mhz) { + val = ((rp_state_cap &0xff0000 ) >>16 ) *GT_FREQUENCY_MULTIPLIER ; + }else { +BUG (); + } +return snprintf (buf,PAGE_SIZE ,"%d" , val); +} +
drivers/gpu/drm/i915/i915_sysfs.c static DEVICE_ATTR(gt_cur_freq_mhz, S_IRUGO, gt_cur_freq_mhz_show, NULL); · line 378
Quoted 10 lines
drivers/gpu/drm/i915/i915_sysfs.c +3
-0static const struct attribute *gen6_attrs[] = { &dev_attr_gt_cur_freq_mhz.attr, &dev_attr_gt_max_freq_mhz.attr, &dev_attr_gt_min_freq_mhz.attr, + &dev_attr_gt_RP0_freq_mhz.attr, + &dev_attr_gt_RP1_freq_mhz.attr, + &dev_attr_gt_RPn_freq_mhz.attr,NULL , };
If you add documentation:
Signature · 1 line
Jesse Barnes, Intel Open Source Technology Center
PATCH 7/7 drm/i915: expose energy counter on SNB and IVB
In reply to Ben WidawskyFrom: Jesse Barnes <[email protected]>
On SNB and IVB, there's an MSR (also exposed through MCHBAR) we can use to read out the amount of energy used over time. Expose this in sysfs to make it easy to do power comparisons with different configurations.
If the platform supports it, the file will show up under the drm/card0/power subdirectory of the PCI device in sysfs as gt_energy_uJ. The value in the file is a running total of energy (in microjoules) consumed by the graphics device.
v2: move to sysfs (Ben, Daniel)
expose a simple value (Chris)
drop unrelated hunk (Ben)
v3: by Ben
Tied it into existing rc6 sysfs entries and named that a more generic "power attrs." Fixed rebase conflicts.
--- drivers/gpu/drm/i915/i915_reg.h | 2 ++ drivers/gpu/drm/i915/i915_sysfs.c | 25 +++++++++++++++++++++++-- 2 files changed, 25 insertions(+), 2 deletions(-)
drivers/gpu/drm/i915/i915_reg.h +2 -0
@@ -1242,6 +1242,8 @@#define CLKCFG_MEM_800 (3 << 4)#define CLKCFG_MEM_MASK (7 << 4) +#define SECP_NRG_STTS (MCHBAR_MIRROR_BASE_SNB + 0x592c) +#define TSC1 0x11001#define TSE (1<<0)#define TR1 0x11006
drivers/gpu/drm/i915/i915_sysfs.c +23 -2
@@ -76,22 +76,43 @@ show_rc6pp_ms(struct device *kdev, struct device_attribute *attr, char *buf)return snprintf (buf,PAGE_SIZE ,"%u" , rc6pp_residency); } +#define MSR_IA32_PACKAGE_POWER_SKU_UNIT 0x00000606 + +static ssize_t +show_gt_energy_uJ (struct device *dev,struct device_attribute *attr,char *buf) +{ + struct drm_minor *dminor = container_of(dev, struct drm_minor, kdev); +struct drm_i915_private *dev_priv = dminor->dev->dev_private; +u64 ppsu; +u32 val, units; + +rdmsrl (MSR_IA32_PACKAGE_POWER_SKU_UNIT , ppsu); + + ppsu = (ppsu &0x1f00 ) >>8 ; + units =1000000 / (1 << ppsu);/* convert to uJ */ + val =I915_READ (SECP_NRG_STTS ); + +return snprintf (buf,PAGE_SIZE ,"%u" , val * units); +} + static DEVICE_ATTR(rc6_enable, S_IRUGO, show_rc6_mask, NULL); static DEVICE_ATTR(rc6_residency_ms, S_IRUGO, show_rc6_ms, NULL); static DEVICE_ATTR(rc6p_residency_ms, S_IRUGO, show_rc6p_ms, NULL); static DEVICE_ATTR(rc6pp_residency_ms, S_IRUGO, show_rc6pp_ms, NULL); +static DEVICE_ATTR(gt_energy_uJ, S_IRUGO, show_gt_energy_uJ, NULL); -static struct attribute *rc6_attrs[] = { +static struct attribute *power_attrs[] = { &dev_attr_rc6_enable.attr, &dev_attr_rc6_residency_ms.attr, &dev_attr_rc6p_residency_ms.attr, &dev_attr_rc6pp_residency_ms.attr, + &dev_attr_gt_energy_uJ.attr,NULL };static struct attribute_group rc6_attr_group = { .name = power_group_name, - .attrs = rc6_attrs + .attrs = power_attrs };#endif
Signature · 1 line
1.7.12
Re: [PATCH 7/7] drm/i915: expose energy counter on SNB and IVB
Quoted 55 lines · Ben Widawsky
From: Jesse Barnes <[email protected]>
On SNB and IVB, there's an MSR (also exposed through MCHBAR) we can use to read out the amount of energy used over time. Expose this in sysfs to make it easy to do power comparisons with different configurations.
If the platform supports it, the file will show up under the drm/card0/power subdirectory of the PCI device in sysfs as
gt_energy_uJ. The value in the file is a running total of energy (in microjoules) consumed by the graphics device.v2: move to sysfs (Ben, Daniel)
expose a simple value (Chris)
drop unrelated hunk (Ben)Signed-off-by Jesse Barnes <[email protected]>v3: by Ben
Tied it into existing rc6 sysfs entries and named that a more generic "power attrs." Fixed rebase conflicts.Signed-off-by Ben Widawsky <[email protected]>--- drivers/gpu/drm/i915/i915_reg.h | 2 ++ drivers/gpu/drm/i915/i915_sysfs.c | 25 +++++++++++++++++++++++-- 2 files changed, 25 insertions(+), 2 deletions(-)drivers/gpu/drm/i915/i915_reg.h +2
-0@@ -1242,6 +1242,8 @@#define CLKCFG_MEM_800 (3 << 4)#define CLKCFG_MEM_MASK (7 << 4) +#define SECP_NRG_STTS (MCHBAR_MIRROR_BASE_SNB + 0x592c) +#define TSC1 0x11001#define TSE (1<<0)#define TR1 0x11006drivers/gpu/drm/i915/i915_sysfs.c +7
-0@@ -76,22 +76,43 @@ show_rc6pp_ms(struct device *kdev, struct device_attribute *attr, char *buf)return snprintf (buf,PAGE_SIZE ,"%u" , rc6pp_residency); } +#define MSR_IA32_PACKAGE_POWER_SKU_UNIT 0x00000606 + +static ssize_t +show_gt_energy_uJ (struct device *dev,struct device_attribute *attr,char *buf) +{ + struct drm_minor *dminor = container_of(dev, struct drm_minor, kdev); +struct drm_i915_private *dev_priv = dminor->dev->dev_private;
drivers/gpu/drm/i915/i915_sysfs.c show_rc6pp_ms(struct device *kdev, struct device_attribute *attr, char *buf) · line 95
Quoted 10 lines
drivers/gpu/drm/i915/i915_sysfs.c +10
-0+u64 ppsu; +u32 val, units; + +rdmsrl (MSR_IA32_PACKAGE_POWER_SKU_UNIT , ppsu); + + ppsu = (ppsu &0x1f00 ) >>8 ; + units =1000000 / (1 << ppsu);/* convert to uJ */ + val =I915_READ (SECP_NRG_STTS ); + +return snprintf (buf,PAGE_SIZE ,"%u" , val * units);
Besides that this could overflow it also introduces rounding errors that accumulate. I guess we need a 64 mul+div here, like in the residency counters.
-Daniel
Quoted 32 lines
drivers/gpu/drm/i915/i915_sysfs.c +6
-2+} + static DEVICE_ATTR(rc6_enable, S_IRUGO, show_rc6_mask, NULL); static DEVICE_ATTR(rc6_residency_ms, S_IRUGO, show_rc6_ms, NULL); static DEVICE_ATTR(rc6p_residency_ms, S_IRUGO, show_rc6p_ms, NULL); static DEVICE_ATTR(rc6pp_residency_ms, S_IRUGO, show_rc6pp_ms, NULL); +static DEVICE_ATTR(gt_energy_uJ, S_IRUGO, show_gt_energy_uJ, NULL); -static struct attribute *rc6_attrs[] = { +static struct attribute *power_attrs[] = { &dev_attr_rc6_enable.attr, &dev_attr_rc6_residency_ms.attr, &dev_attr_rc6p_residency_ms.attr, &dev_attr_rc6pp_residency_ms.attr, + &dev_attr_gt_energy_uJ.attr,NULL };static struct attribute_group rc6_attr_group = { .name = power_group_name, - .attrs = rc6_attrs + .attrs = power_attrs };#endif Signature · 6 lines
1.7.12 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch
Re: [PATCH 7/7] drm/i915: expose energy counter on SNB and IVB
Quoted 63 lines · On Fri, 7 Sep 2012 12:03:37 +0200 Daniel Vetter <[email protected]> wrote:
Quoted 62 lines · Ben Widawsky
From: Jesse Barnes <[email protected]>
On SNB and IVB, there's an MSR (also exposed through MCHBAR) we can use to read out the amount of energy used over time. Expose this in sysfs to make it easy to do power comparisons with different configurations.
If the platform supports it, the file will show up under the drm/card0/power subdirectory of the PCI device in sysfs as
gt_energy_uJ. The value in the file is a running total of energy (in microjoules) consumed by the graphics device.v2: move to sysfs (Ben, Daniel)
expose a simple value (Chris)
drop unrelated hunk (Ben)Signed-off-by Jesse Barnes <[email protected]>v3: by Ben
Tied it into existing rc6 sysfs entries and named that a more generic "power attrs." Fixed rebase conflicts.Signed-off-by Ben Widawsky <[email protected]>--- drivers/gpu/drm/i915/i915_reg.h | 2 ++ drivers/gpu/drm/i915/i915_sysfs.c | 25 +++++++++++++++++++++++-- 2 files changed, 25 insertions(+), 2 deletions(-)drivers/gpu/drm/i915/i915_reg.h +2
-0@@ -1242,6 +1242,8 @@#define CLKCFG_MEM_800 (3 << 4)#define CLKCFG_MEM_MASK (7 << 4) +#define SECP_NRG_STTS (MCHBAR_MIRROR_BASE_SNB + 0x592c) +#define TSC1 0x11001#define TSE (1<<0)#define TR1 0x11006drivers/gpu/drm/i915/i915_sysfs.c +14
-0@@ -76,22 +76,43 @@ show_rc6pp_ms(struct device *kdev, struct device_attribute *attr, char *buf)return snprintf (buf,PAGE_SIZE ,"%u" , rc6pp_residency); } +#define MSR_IA32_PACKAGE_POWER_SKU_UNIT 0x00000606 + +static ssize_t +show_gt_energy_uJ (struct device *dev,struct device_attribute *attr,char *buf) +{ + struct drm_minor *dminor = container_of(dev, struct drm_minor, kdev); +struct drm_i915_private *dev_priv = dminor->dev->dev_private; +u64 ppsu; +u32 val, units; + +rdmsrl (MSR_IA32_PACKAGE_POWER_SKU_UNIT , ppsu); + + ppsu = (ppsu &0x1f00 ) >>8 ; + units =1000000 / (1 << ppsu);/* convert to uJ */
Quoted 9 lines
drivers/gpu/drm/i915/i915_sysfs.c show_rc6pp_ms(struct device *kdev, struct device_attribute *attr, char *buf) · line 95
Quoted 3 lines
drivers/gpu/drm/i915/i915_sysfs.c +3
-0+ val =I915_READ (SECP_NRG_STTS ); + +return snprintf (buf,PAGE_SIZE ,"%u" , val * units);Besides that this could overflow it also introduces rounding errors that accumulate. I guess we need a 64 mul+div here, like in the residency counters.
-Daniel
Jesse, I'm not fixing this one for you since I screwed up the residency math enough times before.
Quoted 33 lines
Quoted 32 lines
drivers/gpu/drm/i915/i915_sysfs.c +6
-2+} + static DEVICE_ATTR(rc6_enable, S_IRUGO, show_rc6_mask, NULL); static DEVICE_ATTR(rc6_residency_ms, S_IRUGO, show_rc6_ms, NULL); static DEVICE_ATTR(rc6p_residency_ms, S_IRUGO, show_rc6p_ms, NULL); static DEVICE_ATTR(rc6pp_residency_ms, S_IRUGO, show_rc6pp_ms, NULL); +static DEVICE_ATTR(gt_energy_uJ, S_IRUGO, show_gt_energy_uJ, NULL); -static struct attribute *rc6_attrs[] = { +static struct attribute *power_attrs[] = { &dev_attr_rc6_enable.attr, &dev_attr_rc6_residency_ms.attr, &dev_attr_rc6p_residency_ms.attr, &dev_attr_rc6pp_residency_ms.attr, + &dev_attr_gt_energy_uJ.attr,NULL };static struct attribute_group rc6_attr_group = { .name = power_group_name, - .attrs = rc6_attrs + .attrs = power_attrs };#endif Signature · 6 lines
1.7.12 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Signature · 1 line
Ben Widawsky, Intel Open Source Technology Center
Re: [PATCH 7/7] drm/i915: expose energy counter on SNB and IVB
Quoted 19 lines · Ben Widawsky
Quoted 16 lines · On Fri, 7 Sep 2012 12:03:37 +0200 Daniel Vetter <[email protected]> wrote:
Quoted 15 lines · Ben Widawsky
From: Jesse Barnes <[email protected]>
On SNB and IVB, there's an MSR (also exposed through MCHBAR) we can use to read out the amount of energy used over time. Expose this in sysfs to make it easy to do power comparisons with different configurations.
If the platform supports it, the file will show up under the drm/card0/power subdirectory of the PCI device in sysfs as
gt_energy_uJ. The value in the file is a running total of energy (in microjoules) consumed by the graphics device.v2: move to sysfs (Ben, Daniel)
expose a simple value (Chris)
drop unrelated hunk (Ben)
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Signed-off-by Jesse Barnes <[email protected]>v3: by Ben
Tied it into existing rc6 sysfs entries and named that a more generic "power attrs." Fixed rebase conflicts.Signed-off-by Ben Widawsky <[email protected]>
A reminder that we never finished this patch.
-Chris
Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
Re: [PATCH 7/7] drm/i915: expose energy counter on SNB and IVB
Quoted 22 lines · Chris Wilson
Quoted 21 lines · Ben Widawsky
Quoted 18 lines · On Fri, 7 Sep 2012 12:03:37 +0200 Daniel Vetter <[email protected]> wrote:
Quoted 17 lines · Ben Widawsky
From: Jesse Barnes <[email protected]>
On SNB and IVB, there's an MSR (also exposed through MCHBAR) we can use to read out the amount of energy used over time. Expose this in sysfs to make it easy to do power comparisons with different configurations.
If the platform supports it, the file will show up under the drm/card0/power subdirectory of the PCI device in sysfs as
gt_energy_uJ. The value in the file is a running total of energy (in microjoules) consumed by the graphics device.v2: move to sysfs (Ben, Daniel)
expose a simple value (Chris)
drop unrelated hunk (Ben)Signed-off-by Jesse Barnes <[email protected]>
Quoted 11 lines
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
v3: by Ben
Tied it into existing rc6 sysfs entries and named that a more generic "power attrs." Fixed rebase conflicts.Signed-off-by Ben Widawsky <[email protected]>A reminder that we never finished this patch.
-ChrisSignature · 1 line
Chris Wilson, Intel Open Source Technology Centre
This will be supported in the RAPL driver, so I think we don't need it. The latest version of that can be found:
https://lkml.org/lkml/2013/4/9/812
Signature · 1 line
Ben Widawsky, Intel Open Source Technology Center