From: Daniel Vetter <daniel@ffwll.ch>
To: deepak.s@intel.com
Cc: intel-gfx@lists.freedesktop.org
Subject: Re: [PATCH 4/4] drm/i915: WA to fix Voltage is not getting dropped to Vmin when Gfx is power gated.
Date: Mon, 9 Dec 2013 09:35:46 +0100 [thread overview]
Message-ID: <20131209083546.GC9804@phenom.ffwll.local> (raw)
In-Reply-To: <1386492406-24716-5-git-send-email-deepak.s@intel.com>
On Sun, Dec 08, 2013 at 02:16:46PM +0530, deepak.s@intel.com wrote:
> From: Deepak S <deepak.s@intel.com>
>
> Voltage is not getting dropped to Vmin when GFX enters RC6 and running
> in/out of turbo frequency. When we enter RC6 and GFX Clocks are off, the
> voltage remains higher than Vmin. As GFX does not request lower
> frequencies when it is power gated. Ideally the Voltage should be
> dropped to Vmin, To fix this we check for Gfx Idle, and set the freq to
> RPe.
>
> Signed-off-by: Deepak S <deepak.s@intel.com>
Do we really still need this patch on upstream?
- It reads the synchronous punit wait loops which we've killed in
commit 4c7915616a7a09be0c1e5b76c26381df15fe671b
Author: Ville Syrjälä <ville.syrjala@linux.intel.com>
Date: Thu Nov 7 19:57:48 2013 +0200
drm/i915: Kill vlv_update_rps_cur_delay()
- It seems to only set the gpu to the lowest freq (RPe which is RPe after
your previous patch), which we already do since Chris sw-controlled
throttle/de-throttle work.
Iirc both of these aren't in the current vlv android tree. If we indeed
need a special dance to whack the gpu into Vmin then I think we need a bit
more comments (and I'd prefer if we have an explicit check for Vmin
instead of going through the rps delay and hoping for the best).
-Daniel
> ---
> drivers/gpu/drm/i915/i915_reg.h | 4 +++
> drivers/gpu/drm/i915/intel_pm.c | 78 ++++++++++++++++++++++++++++++++++++++++-
> 2 files changed, 81 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h
> index 3be449d..e48ed32 100644
> --- a/drivers/gpu/drm/i915/i915_reg.h
> +++ b/drivers/gpu/drm/i915/i915_reg.h
> @@ -4967,6 +4967,10 @@
> GEN6_PM_RP_DOWN_THRESHOLD | \
> GEN6_PM_RP_DOWN_TIMEOUT)
>
> +#define VLV_GTLC_SURVIVABILITY_REG 0x130098
> +#define VLV_GFX_CLK_STATUS_BIT (1<<3)
> +#define VLV_GFX_CLK_FORCE_ON_BIT (1<<2)
> +
> #define GEN6_GT_GFX_RC6_LOCKED 0x138104
> #define VLV_COUNTER_CONTROL 0x138104
> #define VLV_COUNT_RANGE_HIGH (1<<15)
> diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
> index 6b80ec4..347d77b 100644
> --- a/drivers/gpu/drm/i915/intel_pm.c
> +++ b/drivers/gpu/drm/i915/intel_pm.c
> @@ -3581,12 +3581,87 @@ void gen6_set_rps(struct drm_device *dev, u8 val)
> trace_intel_gpu_freq_change(val * 50);
> }
>
> +/* vlv_set_rps_idle: Set the frequency to Rpe if Gfx clocks are down
> + *
> + * If Gfx clock is UP, then reset the timer as there is a possibility
> + * that normal Turbo logic can bring down the freq to Rpe.
> + * If Gfx clock is Down, then
> + * 1. Mask Turbo interrupts
> + * 2. Bring up Gfx clock
> + * 3. Change the freq to Rpe and wait till P-Unit updates freq
> + * 4. Clear the Force GFX CLK ON bit so that Gfx can down
> + * 5. Unmask Turbo interrupts
> +*/
> +static void vlv_set_rps_idle(struct drm_i915_private *dev_priv)
> +{
> +
> + /*
> + * When we are idle. Drop to min voltage state.
> + * Note: we use RPe here since it should match the
> + * Vmin we were shooting for. That should give us better
> + * perf when we come back out of RC6 than if we used the
> + * min freq available.
> + */
> +
> + if (dev_priv->rps.cur_delay <= dev_priv->rps.rpe_delay)
> + return;
> +
> + if (I915_READ(VLV_GTLC_SURVIVABILITY_REG) & VLV_GFX_CLK_STATUS_BIT) {
> + /* GT is not power gated. Let's wait for normal
> + * Turbo Logic can bring down the freq to RPe.
> + * */
> + vlv_punit_write(dev_priv, PUNIT_REG_GPU_FREQ_REQ,
> + dev_priv->rps.rpe_delay);
> +
> + /* Make sure Rpe is set by P-Unit*/
> + if (wait_for((vlv_update_rps_cur_delay(dev_priv) &&
> + (dev_priv->rps.cur_delay ==
> + dev_priv->rps.rpe_delay)), 100))
> + DRM_DEBUG_DRIVER("Not able to set Rpe\n");
> +
> + } else {
> + /* When, Gfx clock is DOWN */
> +
> + /* Mask turbo interrupt so that they will not come in between */
> + I915_WRITE(GEN6_PMINTRMSK, 0xffffffff);
> +
> + /* Bring up the Gfx clock */
> + I915_WRITE(VLV_GTLC_SURVIVABILITY_REG,
> + I915_READ(VLV_GTLC_SURVIVABILITY_REG) |
> + VLV_GFX_CLK_FORCE_ON_BIT);
> +
> + if (wait_for(((VLV_GFX_CLK_STATUS_BIT &
> + I915_READ(VLV_GTLC_SURVIVABILITY_REG)) != 0), 500)) {
> + DRM_ERROR("GFX_CLK_ON request timed out\n");
> + return;
> + }
> +
> + vlv_punit_write(dev_priv, PUNIT_REG_GPU_FREQ_REQ,
> + dev_priv->rps.rpe_delay);
> +
> + /* Make sure Rpe is set by P-Unit*/
> + if (wait_for((vlv_update_rps_cur_delay(dev_priv) &&
> + (dev_priv->rps.cur_delay ==
> + dev_priv->rps.rpe_delay)), 100))
> + DRM_DEBUG_DRIVER("Not able to set Rpe\n");
> +
> + /* Release the Gfx clock */
> + I915_WRITE(VLV_GTLC_SURVIVABILITY_REG,
> + I915_READ(VLV_GTLC_SURVIVABILITY_REG) &
> + ~VLV_GFX_CLK_FORCE_ON_BIT);
> +
> + /* Unmask Turbo interrupts */
> + I915_WRITE(GEN6_PMINTRMSK, ~GEN6_PM_RPS_EVENTS);
> + }
> +}
> +
> +
> void gen6_rps_idle(struct drm_i915_private *dev_priv)
> {
> mutex_lock(&dev_priv->rps.hw_lock);
> if (dev_priv->rps.enabled) {
> if (dev_priv->info->is_valleyview)
> - valleyview_set_rps(dev_priv->dev, dev_priv->rps.min_delay);
> + vlv_set_rps_idle(dev_priv);
> else
> gen6_set_rps(dev_priv->dev, dev_priv->rps.min_delay);
> dev_priv->rps.last_adj = 0;
> @@ -4846,6 +4921,7 @@ void intel_gpu_ips_teardown(void)
> i915_mch_dev = NULL;
> spin_unlock_irq(&mchdev_lock);
> }
> +
> static void intel_init_emon(struct drm_device *dev)
> {
> struct drm_i915_private *dev_priv = dev->dev_private;
> --
> 1.8.4.2
>
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx
--
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
prev parent reply other threads:[~2013-12-09 8:34 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-12-08 8:46 [PATCH 0/4] Fixes for vlv turbo deepak.s
2013-12-08 8:46 ` [PATCH 1/4] drm/i915: update current freq properly before requesting new freq deepak.s
2013-12-09 10:03 ` Ville Syrjälä
2013-12-09 14:22 ` S, Deepak
2013-12-08 8:46 ` [PATCH 2/4] drm/i915: set min delay to rpe delay (Efficient frequency) for better performace deepak.s
2013-12-08 16:07 ` Chris Wilson
2013-12-09 13:58 ` S, Deepak
2013-12-08 8:46 ` [PATCH 3/4] drm/i915: Disable/Enable PM Intrrupts based on the current freq deepak.s
2013-12-09 8:11 ` Daniel Vetter
2013-12-09 10:11 ` Ville Syrjälä
2013-12-10 9:44 ` S, Deepak
2013-12-08 8:46 ` [PATCH 4/4] drm/i915: WA to fix Voltage is not getting dropped to Vmin when Gfx is power gated deepak.s
2013-12-09 8:35 ` Daniel Vetter [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20131209083546.GC9804@phenom.ffwll.local \
--to=daniel@ffwll.ch \
--cc=deepak.s@intel.com \
--cc=intel-gfx@lists.freedesktop.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.