Message ID | 1447435781-23416-1-git-send-email-mika.kuoppala@intel.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On Fri, Nov 13, 2015 at 07:29:41PM +0200, Mika Kuoppala wrote: > When we set and later readback a frequency value through > sysfs interface, igt/pm_rpm assumes that we get same value back > if it matches hw granularity. > > On bxt we have found out that this is not always the case. > Currently frequency - hw ratio - frequency conversions round down, > with few exceptions on platforms that have more specific conversions. > On bxt the supported range can be for example from 100Mhz to 650Mhz. > Midpoint is then calculated by test to be 375 which pm_rps uses to find a > closest hw supported frequency. That is 366 (ratio 22), > which it then writes back. But as the rounding down kicks in, > driver actually sets 350 instead of 366, as 366 is 2/3 below 22 * 50/3. > > Fix this by rounding to closest instead of rounding down in > freq-ratio-freq conversions. > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=92768 > Testcase: igt/pm_rps/basic-api > Tested-by: Bob Paauwe <bob.j.paauwe@intel.com> > Cc: Bob Paauwe <bob.j.paauwe@intel.com> > Signed-off-by: Imre Deak <imre.deak@intel.com> > Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com> > --- > drivers/gpu/drm/i915/intel_pm.c | 8 +++++--- > 1 file changed, 5 insertions(+), 3 deletions(-) > > diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c > index 647c0ff..740623f 100644 > --- a/drivers/gpu/drm/i915/intel_pm.c > +++ b/drivers/gpu/drm/i915/intel_pm.c > @@ -7139,7 +7139,8 @@ static int chv_freq_opcode(struct drm_i915_private *dev_priv, int val) > int intel_gpu_freq(struct drm_i915_private *dev_priv, int val) > { > if (IS_GEN9(dev_priv->dev)) > - return (val * GT_FREQUENCY_MULTIPLIER) / GEN9_FREQ_SCALER; > + return DIV_ROUND_CLOSEST(val * GT_FREQUENCY_MULTIPLIER, > + GEN9_FREQ_SCALER); > else if (IS_CHERRYVIEW(dev_priv->dev)) > return chv_gpu_freq(dev_priv, val); > else if (IS_VALLEYVIEW(dev_priv->dev)) > @@ -7151,13 +7152,14 @@ int intel_gpu_freq(struct drm_i915_private *dev_priv, int val) > int intel_freq_opcode(struct drm_i915_private *dev_priv, int val) > { > if (IS_GEN9(dev_priv->dev)) > - return (val * GEN9_FREQ_SCALER) / GT_FREQUENCY_MULTIPLIER; > + return DIV_ROUND_CLOSEST(val * GEN9_FREQ_SCALER, > + GT_FREQUENCY_MULTIPLIER); > else if (IS_CHERRYVIEW(dev_priv->dev)) > return chv_freq_opcode(dev_priv, val); > else if (IS_VALLEYVIEW(dev_priv->dev)) > return byt_freq_opcode(dev_priv, val); > else > - return val / GT_FREQUENCY_MULTIPLIER; > + return DIV_ROUND_CLOSEST(val, GT_FREQUENCY_MULTIPLIER); lgtm Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> > } > > struct request_boost { > -- > 2.5.0 > > _______________________________________________ > Intel-gfx mailing list > Intel-gfx@lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
On Fri, 13 Nov 2015 19:29:41 +0200 Mika Kuoppala <mika.kuoppala@linux.intel.com> wrote: > When we set and later readback a frequency value through > sysfs interface, igt/pm_rpm assumes that we get same value back > if it matches hw granularity. > > On bxt we have found out that this is not always the case. > Currently frequency - hw ratio - frequency conversions round down, > with few exceptions on platforms that have more specific conversions. > On bxt the supported range can be for example from 100Mhz to 650Mhz. > Midpoint is then calculated by test to be 375 which pm_rps uses to find a > closest hw supported frequency. That is 366 (ratio 22), > which it then writes back. But as the rounding down kicks in, > driver actually sets 350 instead of 366, as 366 is 2/3 below 22 * 50/3. > > Fix this by rounding to closest instead of rounding down in > freq-ratio-freq conversions. > > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=92768 > Testcase: igt/pm_rps/basic-api > Tested-by: Bob Paauwe <bob.j.paauwe@intel.com> > Cc: Bob Paauwe <bob.j.paauwe@intel.com> > Signed-off-by: Imre Deak <imre.deak@intel.com> > Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com> > --- > drivers/gpu/drm/i915/intel_pm.c | 8 +++++--- > 1 file changed, 5 insertions(+), 3 deletions(-) > > diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c > index 647c0ff..740623f 100644 > --- a/drivers/gpu/drm/i915/intel_pm.c > +++ b/drivers/gpu/drm/i915/intel_pm.c > @@ -7139,7 +7139,8 @@ static int chv_freq_opcode(struct drm_i915_private *dev_priv, int val) > int intel_gpu_freq(struct drm_i915_private *dev_priv, int val) > { > if (IS_GEN9(dev_priv->dev)) > - return (val * GT_FREQUENCY_MULTIPLIER) / GEN9_FREQ_SCALER; > + return DIV_ROUND_CLOSEST(val * GT_FREQUENCY_MULTIPLIER, > + GEN9_FREQ_SCALER); > else if (IS_CHERRYVIEW(dev_priv->dev)) > return chv_gpu_freq(dev_priv, val); > else if (IS_VALLEYVIEW(dev_priv->dev)) > @@ -7151,13 +7152,14 @@ int intel_gpu_freq(struct drm_i915_private *dev_priv, int val) > int intel_freq_opcode(struct drm_i915_private *dev_priv, int val) > { > if (IS_GEN9(dev_priv->dev)) > - return (val * GEN9_FREQ_SCALER) / GT_FREQUENCY_MULTIPLIER; > + return DIV_ROUND_CLOSEST(val * GEN9_FREQ_SCALER, > + GT_FREQUENCY_MULTIPLIER); > else if (IS_CHERRYVIEW(dev_priv->dev)) > return chv_freq_opcode(dev_priv, val); > else if (IS_VALLEYVIEW(dev_priv->dev)) > return byt_freq_opcode(dev_priv, val); > else > - return val / GT_FREQUENCY_MULTIPLIER; > + return DIV_ROUND_CLOSEST(val, GT_FREQUENCY_MULTIPLIER); > } > > struct request_boost { Reviewed-by: Bob Paauwe <bob.j.paauwe@intel.com>
On Fri, 13 Nov 2015, Bob Paauwe <bob.j.paauwe@intel.com> wrote: > On Fri, 13 Nov 2015 19:29:41 +0200 > Mika Kuoppala <mika.kuoppala@linux.intel.com> wrote: > >> When we set and later readback a frequency value through >> sysfs interface, igt/pm_rpm assumes that we get same value back >> if it matches hw granularity. >> >> On bxt we have found out that this is not always the case. >> Currently frequency - hw ratio - frequency conversions round down, >> with few exceptions on platforms that have more specific conversions. >> On bxt the supported range can be for example from 100Mhz to 650Mhz. >> Midpoint is then calculated by test to be 375 which pm_rps uses to find a >> closest hw supported frequency. That is 366 (ratio 22), >> which it then writes back. But as the rounding down kicks in, >> driver actually sets 350 instead of 366, as 366 is 2/3 below 22 * 50/3. >> >> Fix this by rounding to closest instead of rounding down in >> freq-ratio-freq conversions. >> >> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=92768 >> Testcase: igt/pm_rps/basic-api >> Tested-by: Bob Paauwe <bob.j.paauwe@intel.com> >> Cc: Bob Paauwe <bob.j.paauwe@intel.com> >> Signed-off-by: Imre Deak <imre.deak@intel.com> >> Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com> >> --- >> drivers/gpu/drm/i915/intel_pm.c | 8 +++++--- >> 1 file changed, 5 insertions(+), 3 deletions(-) >> >> diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c >> index 647c0ff..740623f 100644 >> --- a/drivers/gpu/drm/i915/intel_pm.c >> +++ b/drivers/gpu/drm/i915/intel_pm.c >> @@ -7139,7 +7139,8 @@ static int chv_freq_opcode(struct drm_i915_private *dev_priv, int val) >> int intel_gpu_freq(struct drm_i915_private *dev_priv, int val) >> { >> if (IS_GEN9(dev_priv->dev)) >> - return (val * GT_FREQUENCY_MULTIPLIER) / GEN9_FREQ_SCALER; >> + return DIV_ROUND_CLOSEST(val * GT_FREQUENCY_MULTIPLIER, >> + GEN9_FREQ_SCALER); >> else if (IS_CHERRYVIEW(dev_priv->dev)) >> return chv_gpu_freq(dev_priv, val); >> else if (IS_VALLEYVIEW(dev_priv->dev)) >> @@ -7151,13 +7152,14 @@ int intel_gpu_freq(struct drm_i915_private *dev_priv, int val) >> int intel_freq_opcode(struct drm_i915_private *dev_priv, int val) >> { >> if (IS_GEN9(dev_priv->dev)) >> - return (val * GEN9_FREQ_SCALER) / GT_FREQUENCY_MULTIPLIER; >> + return DIV_ROUND_CLOSEST(val * GEN9_FREQ_SCALER, >> + GT_FREQUENCY_MULTIPLIER); >> else if (IS_CHERRYVIEW(dev_priv->dev)) >> return chv_freq_opcode(dev_priv, val); >> else if (IS_VALLEYVIEW(dev_priv->dev)) >> return byt_freq_opcode(dev_priv, val); >> else >> - return val / GT_FREQUENCY_MULTIPLIER; >> + return DIV_ROUND_CLOSEST(val, GT_FREQUENCY_MULTIPLIER); >> } >> >> struct request_boost { > > Reviewed-by: Bob Paauwe <bob.j.paauwe@intel.com> Pushed to drm-intel-fixes, thanks for the patch and review. BR, Jani.
diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 647c0ff..740623f 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -7139,7 +7139,8 @@ static int chv_freq_opcode(struct drm_i915_private *dev_priv, int val) int intel_gpu_freq(struct drm_i915_private *dev_priv, int val) { if (IS_GEN9(dev_priv->dev)) - return (val * GT_FREQUENCY_MULTIPLIER) / GEN9_FREQ_SCALER; + return DIV_ROUND_CLOSEST(val * GT_FREQUENCY_MULTIPLIER, + GEN9_FREQ_SCALER); else if (IS_CHERRYVIEW(dev_priv->dev)) return chv_gpu_freq(dev_priv, val); else if (IS_VALLEYVIEW(dev_priv->dev)) @@ -7151,13 +7152,14 @@ int intel_gpu_freq(struct drm_i915_private *dev_priv, int val) int intel_freq_opcode(struct drm_i915_private *dev_priv, int val) { if (IS_GEN9(dev_priv->dev)) - return (val * GEN9_FREQ_SCALER) / GT_FREQUENCY_MULTIPLIER; + return DIV_ROUND_CLOSEST(val * GEN9_FREQ_SCALER, + GT_FREQUENCY_MULTIPLIER); else if (IS_CHERRYVIEW(dev_priv->dev)) return chv_freq_opcode(dev_priv, val); else if (IS_VALLEYVIEW(dev_priv->dev)) return byt_freq_opcode(dev_priv, val); else - return val / GT_FREQUENCY_MULTIPLIER; + return DIV_ROUND_CLOSEST(val, GT_FREQUENCY_MULTIPLIER); } struct request_boost {