public inbox for intel-gfx@lists.freedesktop.org
 help / color / mirror / Atom feed
From: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
To: Daniel Vetter <daniel@ffwll.ch>
Cc: Ander Conselvan de Oliveira
	<ander.conselvan.de.oliveira@intel.com>,
	intel-gfx@lists.freedesktop.org
Subject: Re: [PATCH 03/42] drm/i915: Only update required power domains.
Date: Tue, 12 May 2015 14:05:52 +0200	[thread overview]
Message-ID: <5551ECA0.3070208@linux.intel.com> (raw)
In-Reply-To: <20150511170021.GH15256@phenom.ffwll.local>

Op 11-05-15 om 19:00 schreef Daniel Vetter:
> On Mon, May 11, 2015 at 04:24:39PM +0200, Maarten Lankhorst wrote:
>> This prevents unnecessarily updating power domains, while still
>> enabling all power domains on initial setup.
>>
>> Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
>> ---
>>  drivers/gpu/drm/i915/intel_display.c | 52 ++++++++++++++++++++++++++++--------
>>  1 file changed, 41 insertions(+), 11 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
>> index af96d686aae2..42d0cc329b37 100644
>> --- a/drivers/gpu/drm/i915/intel_display.c
>> +++ b/drivers/gpu/drm/i915/intel_display.c
>> @@ -5298,36 +5298,72 @@ static unsigned long get_crtc_power_domains(struct drm_crtc *crtc)
>>  	return mask;
>>  }
>>  
>> +static bool
>> +needs_modeset(struct drm_crtc_state *state)
> I think we should extract this from drm_atomic_helper.c. But that can be
> done as a follow-up - if you track it ;-)
>> +{
>> +	return state->mode_changed || state->active_changed;
>> +}
>> +
>>  static void modeset_update_crtc_power_domains(struct drm_atomic_state *state)
>>  {
>>  	struct drm_device *dev = state->dev;
>>  	struct drm_i915_private *dev_priv = dev->dev_private;
>>  	unsigned long pipe_domains[I915_MAX_PIPES] = { 0, };
>>  	struct intel_crtc *crtc;
>> +	bool init_power = dev_priv->power_domains.init_power_on;
>> +	bool any_power = init_power, any_modeset = false;
>> +	unsigned long domains;
>>  
>>  	/*
>>  	 * First get all needed power domains, then put all unneeded, to avoid
>>  	 * any unnecessary toggling of the power wells.
>>  	 */
>>  	for_each_intel_crtc(dev, crtc) {
>> +		int idx = drm_crtc_index(&crtc->base);
>> +		struct drm_crtc_state *crtc_state = state->crtc_states[idx];
>>  		enum intel_display_power_domain domain;
>>  
>> -		if (!crtc->base.state->enable)
>> +		if (!init_power && !crtc_state)
>> +			continue;
> 		if (!needs_modeset)
> 			continue;
>
> while you're optimizing this?
>
>> +
>> +		if (needs_modeset(crtc->base.state))
>> +			any_modeset = true;
>> +
>> +		if (crtc->base.state->enable)
>> +			pipe_domains[crtc->pipe] =
>> +				get_crtc_power_domains(&crtc->base);
>> +
>> +		if (pipe_domains[crtc->pipe] == crtc->enabled_power_domains)
>>  			continue;
>>  
>> -		pipe_domains[crtc->pipe] = get_crtc_power_domains(&crtc->base);
>> +		WARN_ON(!init_power && !needs_modeset(crtc->base.state));
>> +
>> +		any_power = true;
>> +		domains = pipe_domains[crtc->pipe] &
>> +			  ~crtc->enabled_power_domains;
>>  
>> -		for_each_power_domain(domain, pipe_domains[crtc->pipe])
>> +		for_each_power_domain(domain, domains)
>>  			intel_display_power_get(dev_priv, domain);
>>  	}
>>  
>> -	if (dev_priv->display.modeset_global_resources)
>> +	if (any_modeset && dev_priv->display.modeset_global_resources)
>>  		dev_priv->display.modeset_global_resources(state);
>>  
>> +	if (!any_power)
>> +		return;
>> +
>>  	for_each_intel_crtc(dev, crtc) {
>> +		int idx = drm_crtc_index(&crtc->base);
>> +		struct drm_crtc_state *crtc_state = state->crtc_states[idx];
>>  		enum intel_display_power_domain domain;
>>  
>> -		for_each_power_domain(domain, crtc->enabled_power_domains)
>> +		if (!init_power && !crtc_state)
>> +			continue;
> Same shortcut here?
It's not an optimization, the same path can be used for modeset and updating planes.
If it's a plane update the code might run with irqs disabled, in which case we can't call power_get or power_put because that requires a mutex.

~Maarten

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

  reply	other threads:[~2015-05-12 12:05 UTC|newest]

Thread overview: 98+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-05-11 14:24 [PATCH 00/42] drm/i915: Convert to atomic, part 2 Maarten Lankhorst
2015-05-11 14:24 ` [PATCH 01/42] drm/atomic: Allow drivers to subclass drm_atomic_state Maarten Lankhorst
2015-05-13  5:52   ` Daniel Vetter
2015-05-11 14:24 ` [PATCH 02/42] drm/i915: get rid of intel_crtc_disable and related code, v2 Maarten Lankhorst
2015-05-11 17:08   ` Daniel Vetter
2015-05-11 14:24 ` [PATCH 03/42] drm/i915: Only update required power domains Maarten Lankhorst
2015-05-11 17:00   ` Daniel Vetter
2015-05-12 12:05     ` Maarten Lankhorst [this message]
2015-05-12 13:13       ` Daniel Vetter
2015-05-11 14:24 ` [PATCH 04/42] drm/i915: use intel_crtc_control everywhere Maarten Lankhorst
2015-05-11 17:11   ` Daniel Vetter
2015-05-12 12:06     ` Maarten Lankhorst
2015-05-12 13:16       ` Daniel Vetter
2015-05-12 14:38         ` Daniel Stone
2015-05-11 14:24 ` [PATCH 05/42] drm/i915: Get rid of new_encoder Maarten Lankhorst
2015-05-11 17:17   ` Daniel Vetter
2015-05-11 14:24 ` [PATCH 06/42] drm/i915: get rid of new_crtc Maarten Lankhorst
2015-05-11 17:28   ` Daniel Vetter
2015-05-12 12:07     ` Maarten Lankhorst
2015-05-11 14:24 ` [PATCH 07/42] drm/i915: Get rid of crtc->new_enabled, v2 Maarten Lankhorst
2015-05-11 17:33   ` Daniel Vetter
2015-05-11 17:44     ` Daniel Vetter
2015-05-11 14:24 ` [PATCH 08/42] drm/i915: Implement intel_crtc_toggle using atomic state Maarten Lankhorst
2015-05-11 18:12   ` Daniel Vetter
2015-05-11 14:24 ` [PATCH 09/42] drm/i915: Make intel_modeset_fixup_state similar to the atomic helper Maarten Lankhorst
2015-05-12  6:59   ` Daniel Vetter
2015-05-12 12:41     ` Maarten Lankhorst
2015-05-12 13:18       ` Daniel Vetter
2015-05-11 14:24 ` [PATCH 10/42] drm/i915: make plane helpers fully atomic Maarten Lankhorst
2015-05-12  8:18   ` Daniel Vetter
2015-05-12 13:33     ` Maarten Lankhorst
2015-05-12 13:43       ` Ville Syrjälä
2015-05-12 13:46         ` Ville Syrjälä
2015-05-12 15:31         ` Daniel Vetter
2015-05-12 16:00       ` Daniel Vetter
2015-05-11 14:24 ` [PATCH 11/42] drm/i915: Update less state during modeset Maarten Lankhorst
2015-05-12  8:22   ` Daniel Vetter
2015-05-11 14:24 ` [PATCH 12/42] drm/i915: move swap_state to the right place Maarten Lankhorst
2015-05-12  8:25   ` Daniel Vetter
2015-05-11 14:24 ` [PATCH 13/42] drm/i915: Set mode_changed for audio in intel_modeset_pipe_config() Maarten Lankhorst
2015-05-11 14:24 ` [PATCH 14/42] drm/i915: Make __intel_set_mode() take only atomic state as argument Maarten Lankhorst
2015-05-11 14:24 ` [PATCH 15/42] drm/i915: Use hwmode for vblanks Maarten Lankhorst
2015-05-11 14:24 ` [PATCH 16/42] drm/i915: Remove usage of crtc->config from i915_debugfs.c Maarten Lankhorst
2015-05-12  8:51   ` Daniel Vetter
2015-05-11 14:24 ` [PATCH 17/42] drm/i915: Remove use of crtc->config from intel_pm.c Maarten Lankhorst
2015-05-12  8:54   ` Daniel Vetter
2015-05-11 14:24 ` [PATCH 18/42] drm/i915: Remove use of crtc->config from intel_audio.c Maarten Lankhorst
2015-05-11 14:24 ` [PATCH 19/42] drm/i915: remove use of crtc->config from intel_fbc.c Maarten Lankhorst
2015-05-11 14:24 ` [PATCH 20/42] drm/i915: remove use of crtc->config from intel_atomic.c and intel_sprite.c Maarten Lankhorst
2015-05-12  9:03   ` Daniel Vetter
2015-05-12 13:36     ` Maarten Lankhorst
2015-05-11 14:24 ` [PATCH 21/42] drm/i915: Remove use of crtc->config from intel_overlay.c Maarten Lankhorst
2015-05-12  9:06   ` Daniel Vetter
2015-05-11 14:24 ` [PATCH 22/42] drm/i915: Pass old state to crtc_disable and use it Maarten Lankhorst
2015-05-12  9:13   ` Daniel Vetter
2015-05-11 14:24 ` [PATCH 23/42] drm/i915: Pass old state to encoder->(post_)disable Maarten Lankhorst
2015-05-12  9:16   ` Daniel Vetter
2015-05-11 14:25 ` [PATCH 24/42] drm/i915: Remove use of crtc->config from intel_fbdev.c Maarten Lankhorst
2015-05-11 14:25 ` [PATCH 25/42] drm/i915: Remove use of crtc->config from intel_psr.c Maarten Lankhorst
2015-05-12  9:20   ` Daniel Vetter
2015-05-12 13:41     ` Maarten Lankhorst
2015-05-11 14:25 ` [PATCH 26/42] drm/i915: Remove use of crtc->config from intel_ddi.c Maarten Lankhorst
2015-05-11 14:25 ` [PATCH 27/42] drm/i915: Remove use of crtc->config from intel_dp.c Maarten Lankhorst
2015-05-12  9:22   ` Daniel Vetter
2015-05-12 13:43     ` Maarten Lankhorst
2015-05-11 14:25 ` [PATCH 28/42] drm/i915: Remove use of crtc->config from intel_dp_mst.c Maarten Lankhorst
2015-05-11 14:25 ` [PATCH 29/42] drm/i915: Remove use of crtc->config from intel_dsi.c Maarten Lankhorst
2015-05-11 14:25 ` [PATCH 30/42] drm/i915: Remove use of crtc->config in intel_hdmi.c Maarten Lankhorst
2015-05-12  9:26   ` Daniel Vetter
2015-05-11 14:25 ` [PATCH 31/42] drm/i915: Remove use of crtc->config in intel_sdvo.c Maarten Lankhorst
2015-05-11 14:25 ` [PATCH 32/42] drm/i915: Calculate haswell plane workaround Maarten Lankhorst
2015-05-12  9:43   ` Daniel Vetter
2015-05-12 14:05     ` Maarten Lankhorst
2015-05-12 16:54       ` Daniel Vetter
2015-05-11 14:25 ` [PATCH 33/42] drm/i915: remove crtc->active tracking completely Maarten Lankhorst
2015-05-12  9:55   ` Daniel Vetter
2015-05-12 10:03   ` Daniel Vetter
2015-05-12 14:07     ` Maarten Lankhorst
2015-05-12 16:57       ` Daniel Vetter
2015-05-12 17:01         ` Daniel Stone
2015-05-12 17:08           ` Daniel Vetter
2015-05-11 14:25 ` [PATCH 34/42] drm/i915: get rid of crtc->config in intel_display.c, part 1 Maarten Lankhorst
2015-05-12 10:11   ` Daniel Vetter
2015-05-12 14:13     ` Maarten Lankhorst
2015-05-12 17:01       ` Daniel Vetter
2015-05-11 14:25 ` [PATCH 35/42] drm/i915: get rid of crtc->config in intel_display.c, part 2 Maarten Lankhorst
2015-05-12 10:17   ` Daniel Vetter
2015-05-11 14:25 ` [PATCH 36/42] drm/i915: get rid of crtc->config Maarten Lankhorst
2015-05-11 14:25 ` [PATCH 37/42] drm/i915: swap state correctly in intel_atomic_commit Maarten Lankhorst
2015-05-12 13:03   ` Daniel Vetter
2015-05-12 14:16     ` Maarten Lankhorst
2015-05-12 17:03       ` Daniel Vetter
2015-05-11 14:25 ` [PATCH 38/42] drm/i915: Use global atomic state for staged pll config Maarten Lankhorst
2015-05-11 14:25 ` [PATCH 39/42] drm/i915: Support modeset across multiple pipes Maarten Lankhorst
2015-05-11 14:25 ` [PATCH 40/42] drm/i915: Move cdclk and pll setup to intel_modeset_compute_config() Maarten Lankhorst
2015-05-11 14:25 ` [PATCH 41/42] drm/i915: Read hw state into an atomic state struct Maarten Lankhorst
2015-05-11 14:25 ` [PATCH 42/42] drm/i915: return early in __intel_set_mode_setup_plls without modeset Maarten Lankhorst
2015-05-13  7:04 ` [PATCH 00/42] drm/i915: Convert to atomic, part 2 Daniel Vetter

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=5551ECA0.3070208@linux.intel.com \
    --to=maarten.lankhorst@linux.intel.com \
    --cc=ander.conselvan.de.oliveira@intel.com \
    --cc=daniel@ffwll.ch \
    --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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox