When current delay is already at max delay, Let's disable the PM UP
THRESHOLD INTRRUPTS, so that we will not get further interrupts until
current delay is less than max delay, Also request for the PM DOWN
THRESHOLD INTRRUPTS to indicate the decrease in clock freq. and
viceversa for PM DOWN THRESHOLD INTRRUPTS.
v2: Use bool variables (Daniel)
v3: Fix Interrupt masking bit (Deepak)
v4: Use existing symbolic constants in i915_reg.h (Daniel)
v5: Add pm interrupt mask after new_delay calculation (Ville)
Signed-off-by: Deepak S <deepak.s@intel.com>
[danvet: Pass new_delay by value as suggested by Ville. Also appease
checkpatch.]
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We're disabling a boatload of clock gating features on VLV. Maybe these
days we don't need to do that. At least I'm not aware of any workarounds
with this level of paranoia.
This reverts commit 4e8c84a5b1.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
WaDisable4x2SubspanOptimization isn't listed for VLV in the workaround
database, but BSpec says that the relevant bit must be set. Add a
comment to remind people of this.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Call gen7_setup_fixed_func_scheduler() on VLV as well.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
BSpec states that the thread override values set by
gen7_setup_fixed_func_scheduler() are invalid for HSW. So let's not
muck around with them.
Since gen7_setup_fixed_func_scheduler() now has two totally independent
parts, one for IVB and one for HSW, move the HSW part directly into
haswell_init_clock_gating().
Note tht there's another workaround by the name of
WaHSWVSRefCountFullforceMissDisable which basically claims that later
steppings don't need the fix, but since WaVSRefCountFullforceMissDisable
is listed to be needed for all steppings play it safe and keep applying
the workaround.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The current comments indicate that this function implements
WaVSRefCountFullforceMissDisable, which is only true for HSW.
The original purpose of the function is to implement
WaVSThreadDispatchOverride (and a bit more). Fix up the comments
to match reality.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
WaDisableTDLUnitClockGating is only relevant for early steppings of VLV.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
WaDisableVDSUtnitClockGating was only relevant for early steepings of
VLV.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Only early VLV steppings needed thist. Should no longer be relevant.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
WaApplyL3ControlAndL3ChickenMode is only relevant to early HSW
steppings..
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
WaDisableRCZUnitClockGating was needed with early HSW steppings only.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Someone copy pasted the comment from the SNB code w/o reading it.
We never actually implemented the workaround to disable RCPB unit
clock gating on IVB. It would have been needed for early steppings,
but we don't care about those anymore, so just remove the stale
comment.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
WaDisableRCCUnitClockGating is only relevant for SNB.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
WaDisableRCCUnitClockGating is only relevant for SNB.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Can't find any mention of WaDisableVDSUnitClockGating ever being
relevant for SNB. Remove it.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
A lot of the WM functions are only reading from that structure and are
already using const. While converting the code to use dev_priv instead
of dev, I noticed a few places where we can give that hint.
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
With 20+ module parameters, I think referring to them via a struct
improves clarity over just having a bunch of globals. While at it, move
the parameter initialization and definitions into a new file
i915_params.c to reduce clutter in i915_drv.c.
Apart from the ill-named i915_enable_rc6, i915_enable_fbc and
i915_enable_ppgtt parameters, for which we lose the "i915_" prefix
internally, the module parameters now look the same both on the kernel
command line and in code. For example, "i915.modeset".
The downsides of the change are losing static on a couple of variables
and not having the initialization and module_param_named() right next to
each other. On the other hand, all module parameters are now defined in
one place at i915_params.c. Plus you can do this to find all module
parameter references:
$ git grep "i915\." -- drivers/gpu/drm/i915
v2:
- move the definitions into a new file
- s/i915_params/i915/
- make i915_try_reset i915.reset, for consistency
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
WaApplyL3ControlAndL3ChickenMode is only listed for IVB and HSW in
W/A database and BSpec.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The w/a database lists both WaPsdDispatchEnable and
WaDisablePSDDualDispatchEnable for VLV. They appear to be the same
thing, so list both names.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The debug message telling FBC1 has been enabled is missing a newline.
Add it.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
On CTG and IVB+ we don't try to preserve any bits from the
DPFC_CONTROL register. Follow suit on ILK/SNB.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We set up all the bits for DPFC_CONTROL but forgot to actually
write them to the register. Oops.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Make the FBC plane macros take the plane as a parameter.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The ILK/SNB docs don't really mention the the DPFC_HT_MODIFY bit.
CTG docs clearly state that it should be set only when tracking
back buffer modification in persistent mode. The bit is supposed
to be set by software after the first CPU modification to the
back buffer, and it would get automagically cleared by the hardware
on the next page flip.
Since we only track front buffer modification we don't need to set
this bit. GTT modification tracking still appears to work on ILK
and SNB with the bit unset. I don't have a CTG to verify how that
behaves.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The ILK/SNB docs are a bit unclear what the persistent mode does, but
the CTG docs clearly state that it was meant to be used when we're
tracking back buffer modifications. We never do that, so leave it in
non-persistent mode.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We use nuking instead of render tracking on IVB+, so there's
no point in writing IVB_FBC_RT_BASE.
v2: Drop the IVB_FBC_RT_BASE write too
v3: Move the SNB stuff elsewhere, leaving only IVB+ here
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Because whatever.*
* This should contain a fairly long list of issues and still
unresolved resgressions, but I didn't really get a vote.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Those are two distinct concepts. Just use a comment to remind us to
remove that W/A at some point.
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Conflicts are getting out of hand, and now we have to shuffle even
more in -next which was also shuffled in -fixes (the call for
drm_mode_config_reset needs to move yet again).
So do a proper backmerge. I wanted to wait with this for the 3.13
relaese, but alas let's just do this now.
Conflicts:
drivers/gpu/drm/i915/i915_reg.h
drivers/gpu/drm/i915/intel_ddi.c
drivers/gpu/drm/i915/intel_display.c
drivers/gpu/drm/i915/intel_pm.c
Besides the conflict around the forcewake get/put (where we chaged the
called function in -fixes and added a new parameter in -next) code all
the current conflicts are of the adjacent lines changed type.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
INSTPM is a masked register so use the _MASKED_BIT_{ENABLE,DISABLE}
macros when enabling/disabling self-refresh on 915GM.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
So shuffle the checks around a bit. Also give all the structs and
functions proper prefixes: i830_ for the dual-pipe mobile platforms
and i845_ for the two single-pipe desktop platforms.
Note that the max fifo value isn't actually correct for the i830M, but
since we don't frob the fifo split we don't actually need it. This is
different for some gen3 devices where we need the full fifo for self
refresh mode.
Cc: Thomas Richter <richter@rus.uni-stuttgart.de>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
My OCD just couldn't let this slide. Spotted while reviewing Ville's
patch to only flip planes when we have FBC.
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This was introduced in:
commit 7c4a395ff8
Author: Ville Syrjälä <ville.syrjala@linux.intel.com>
Date: Wed Oct 9 19:17:56 2013 +0300
drm/i915: Don't re-compute pipe watermarks except for the affected pipe
and I missed fixing it in:
commit fec8cba306
Author: Jesse Barnes <jbarnes@virtuousgeek.org>
Date: Wed Nov 27 11:10:26 2013 -0800
drm/i915: use crtc_htotal in watermark calculations to match fastboot v2
It's needed for ILK+ platforms to fastboot without crashing on a divide
by 0 after a DPMS on action.
Note: Ville mentioned in his review that this confusion seems to go
down to the original introduction of this code in
commit 801bcfffbb
Author: Paulo Zanoni <paulo.r.zanoni@intel.com>
Date: Fri May 31 10:08:35 2013 -0300
drm/i915: properly set HSW WM_PIPE registers
So it seems to have been missed both in the fastboot patch and in the
3d mode suppport (where only crtc_htotal reflects the real pipe
width).
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
[danvet: Add note based on Ville's review.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Avoid duplicating the same piece of code several times by separating
the watemark vfunc setup from the init_clock_gating vfunc setup on PCH
platforms.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We forgot to intialize the watermark vfuncs for BDW, and hence the
watermarks were never updated.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Looks like I forgot to update the ILK/SNB/IVB watermark patches to deal
with BDW. Add the relevant BDW checks to make sure we take the HSW
codepaths on BDW as well.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
I need the tricky do_switch fix before I can merge the final piece of
the ppgtt enabling puzzle. Otherwise the conflict will be a real pain
to resolve since the do_switch hunk from -fixes must be placed at the
exact right place within a hunk in the next patch.
Conflicts:
drivers/gpu/drm/i915/i915_gem_context.c
drivers/gpu/drm/i915/i915_gem_execbuffer.c
drivers/gpu/drm/i915/intel_display.c
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We use this hook starting from ILK onwards, so change the prefix
accordingly. Also rename functions/struct names used from
haswell_update_wm that are relevant to ILK already.
No functional change.
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
No functional change.
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We now have a very clear method of disabling LP1+ wartermarks,
and we can actually detect if we actually did disable them, or
if they were already disabled. Use that to clean up the
WaCxSRDisabledForSpriteScaling:ivb handling.
I was hoping to apply the workaround in a way that wouldn't
require a blocking wait, but sadly IVB really does appear to
require LP1+ watermarks to be off for an entire frame before
enabling sprite scaling. Simply disabling LP1+ watermarks
during the previous frame is not enough, no matter how early
in the frame we do it :(
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The new HSW watermark code can now handle ILK/SNB/IVB as well, so
switch them over. Kill the old code.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
ILK doesn't like if we just write the LP1+ watermarks registers with 0.
We need to just disable the watermarks by clearing the enable bit. Use
that method also when disabling LP1+ watermarks in init_clock_gating.
It looks like disabling the sprite LP1 watermarks can cause underruns
even if we just toggle the WM1S_LP_EN bit. So treat that bit like the
actual watermark numbers and avoid setting it to 0 immediately.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Linetime watermarks don't exist on ILK/SNB/IVB, so don't compute them
except on HSW.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
ILK has a bunch of issues with FBC. First of all, BSpec tells us that
FBC WM should never be enabled. Secondly when FBC is enabled
with FBC WM disabled, LP2+ watermarks must be disabled.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Multi-pipe LP1+ watermarks are a HSW+ feature, so let's not do it on
earlier generations.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
On ILK disabling LP1+ watermarks must be done carefully to avoid
underruns. If we just write 0 to the register in the middle of the scan
cycle we often get an underrun. So instead we have to leave the actual
watermark levels in the register intact, and just toggle the enable bit.
Presumably the hardware takes a while to get out of low power mode, and
so the watermark level need to stay valid until that time.
We also have to be careful with the WM1S_LP_EN bit. It seems the
hardware more or less treats it like the actual watermarks numbers, and
so we must not toggle it too soon. Just leave it alone when disabling
the LP1+ watermarks.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
ILK/SNB don't have LP2+ watermarks for sprites. Also the LP1 sprite
watermark register has its own enable bit. Take these differences
into account when programming the LP1+ registers.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
On ILK/SNB only LP0/1 watermarks can be enabled when sprites are
enabled, and on ILK/SNB/IVB sprite scaling is limited to LP0 only.
So we can avoid computing the extra levels we're never going to use.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Add a new function ilk_wm_lp_latency() which will tell us what to write
into the WM_LPx register latency field. HSW is different from erlier
gens in this regard.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
On IVB the display data buffer partitioning control lives in the
DISP_ARB_CTL2 register. Add the relevant defines/code for it.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
In the current code, at haswell_modeset_global_resources, first we
decide if we want to enable/disable the power well, then we decide if
we want to enable/disable PC8. On the case where we're enabling PC8
this works fine, but on the case where we disable PC8 due to a non-eDP
monitor being enabled, we first enable the power well and then disable
PC8. Although wrong, this doesn't seem to be causing any problems now,
and we don't even see anything in dmesg. But the patches for runtime
D3 turn this problem into a real bug, so we need to fix it.
This fixes the "modeset-non-lpsp" subtest from the "pm_pc8" test from
intel-gpu-tools.
v2: - Rebase (i915_disable_power_well).
v3: - More reabase.
v4: - Rebase on top of -fixes instead of -nightly.
This is commit d62292c8f7 in -next, but
we need it in -fixes to address Dave's report.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Reported-by: Dave Jones <davej@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
WaVSRefCountFullforceMissDisable and
WaDSRefCountFullforceMissDisable
VS is a carry-over from HSW, and DS is likely not used by anyone yet.
Cc: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
[danvet: Line of 106 chars is too long. Really.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
I stumbled on to some unimplemented errata. To be honest, I am not
really sure of the impact, just that the docs say to do.
No w/a name for this one.
v2: v1 was a stale thing which should have never seen the light of day.
(Haihao)
Cc: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Don't touch DPFC_RECOMP_CTL on FBC2, use RMW to update
the FBC_CONTROL on FBC1 to make it easier for people to
experiment with different numbers. Also fix the interval
mask for FBC1.
v2: Rebased
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Gen2 and gen3 don't have the FBC_CONTROL2 register, so don't
touch it.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
On gen2 the compressed frame buffer pitch is specified in 32B units
rather than the 64B units used on gen3+.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Fixes regression introduced by:
commit bf51d5e2cd
Author: Paulo Zanoni <paulo.r.zanoni at intel.com>
Date: Wed Jul 3 17:12:13 2013 -0300
drm/i915: switch disable_power_well default value to 1
The bug I'm seeing can be reproduced with:
- Have vgacon configured/enabled
- Make sure the power well gets disabled, then enabled. You can
check this by seeing the messages print by hsw_set_power_well
- Stop your display manager
- echo 0 > /sys/class/vtconsole/vtcon1/bind
I can easily reproduce this by blacklising snd_hda_intel and booting
with eDP+HDMI.
If you do this and then look at dmesg, you'll see we're printing
infinite "Unclaimed register" messages. This is happening because
we're stuck on an infinite loop inside console_unlock(), which is
calling many functions from vgacon.c. And the code that's triggering
the error messages is from vgacon_set_cursor_size().
After we re-enable the power well, every time we read/write the VGA
address 0x3d5 we get an "unclaimed register" interrupt (ERR_INT) and
print error messages. If we write anything to the VGA MSR register (it
doesn't really matter which value you write to bit 0), any
reads/writes to 0x3d5 _don't_ trigger the "unclaimed register" errors
anymore (even if MSR bit 0 is zero). So what happens with the current
code is that when we unbind i915 and bind vgacon, we call
console_unlock(). Function console_unlock() is responsible for
printing any messages that were supposed to be print when the console
was locked, so it calls the TTY layer, which calls the console layer,
which calls vgacon to print the messages. At this point, vgacon
eventually calls vgacon_set_cursor_size(), which touches 0x3d5, which
triggers unclaimed register interrupts. The problem is that when we
get these interrupts, we print the error messages, so we add more work
to console_unlock(), which will try to print it again, and then call
vgacon again, trigger a new interrupt, which will put more stuff to
the buffer, and then we'll be stuck at console_unlock() forever.
If you patch intel_uncore.c to not print anything when we detect
unclaimed registers, we won't get into the console_unlock() infinite
loop and the driver unbind will work just fine. We will still be
getting interrupts every time vgacon touches those registers, but we
will survive. This is a valid experiment, but IMHO it's not the real
fix: if we don't print any error messages we will still keep getting
the interrupts, and if we disable ERR_INT we won't get the interrupt
anymore, but we will also stop getting all the other error interrupts.
I talked about this problem with the HW engineer and his
recommendation is "So don't do any VGA I/O or memory access while the
power well is disabled, and make to re-program MSR after enabling the
power well and before using VGA I/O or memory accesses.".
Notice that this is just a partial fix to fd.o #67813. This fixes the
case where the power well is already enabled when we unbind, not when
it's disabled when we unbind.
V2: - Rebase (first version was sent in September).
V3: - Complete rewrite of the same fix: smaller implementation,
improved commit message.
Testcase: igt/drv_module_reload
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67813
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
I want to add more code to the post_enable function.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Just a bunch of regression fixes plus a few patches for long-standing
issues in gem corner-cases that we've hunted down in the past weeks. Since
apparently people hit those in the wild (and we also have nice igts for
them) I've opted for -fixes and cc: stable.
There's 1-2 things oustanding on top of this where I'm still waiting on
confirmation from testing, but nothing really scary.
* tag 'drm-intel-fixes-2013-12-11' of git://people.freedesktop.org/~danvet/drm-intel:
drm/i915: don't update the dri1 breadcrumb with modesetting
drm/i915: Repeat eviction search after idling the GPU
drm/i915: Fix use-after-free in do_switch
drm/i915: fix pm init ordering
drm/i915: Hold mutex across i915_gem_release
drm/i915: Skip clock checks on BDW
drm/i915: Do not clobber config status after a forced restore of hw state
drm/i915: Take modeset locks around intel_modeset_setup_hw_state()
As promised bdw fixes come separate for now. Just a few minior things.
* 'bdw-fixes' of git://people.freedesktop.org/~danvet/drm-intel:
drm/i915/bdw: PIPE_[BC] I[ME]R moved to powerwell
drm/i915/bdw: Limit GTT to 2GB
drm/i915/bdw: Add comment about gen8 HWS PGA
drm/i915/bdw: Free correct number of ppgtt pages
drm/i915/bdw: Do gen6 style reset for gen8
drm/i915/bdw: GEN8 backlight support
drm/i915/bdw: Add BDW to ULT macro
This patch adds the initial infrastructure to allow a Runtime PM
implementation that sets the device to its D3 state. The patch just
adds the necessary callbacks and the initial infrastructure.
We still don't have any platform that actually uses this
infrastructure, we still don't call get/put in all the places we need
to, and we don't have any function to save/restore the state of the
registers. This is not a problem since no platform uses the code added
by this patch. We have a few people simultaneously working on runtime
PM, so this initial code could help everybody make their plans.
V2: - Move some functions to intel_pm.c
- Remove useless pm_runtime_allow() call at init
- Remove useless pm_runtime_mark_last_busy() call at get
- Use pm_runtime_get_sync() instead of 2 calls
- Add a WARN to check if we're really awake
V3: - Rebase.
V4: - Don't need to call pci_{save,restore}_state and
pci_set_power_sate, since they're already called by the PCI
layer
- Remove wrong pm_runtime_enable() call at init_runtime_pm
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
In the current code, at haswell_modeset_global_resources, first we
decide if we want to enable/disable the power well, then we decide if
we want to enable/disable PC8. On the case where we're enabling PC8
this works fine, but on the case where we disable PC8 due to a non-eDP
monitor being enabled, we first enable the power well and then disable
PC8. Although wrong, this doesn't seem to be causing any problems now,
and we don't even see anything in dmesg. But the patches for runtime
D3 turn this problem into a real bug, so we need to fix it.
This fixes the "modeset-non-lpsp" subtest from the "pm_pc8" test from
intel-gpu-tools.
v2: - Rebase (i915_disable_power_well).
v3: - More reabase.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.15 (GNU/Linux)
iQEcBAABAgAGBQJSogqUAAoJEHm+PkMAQRiGM2MIAJrr5KEXEWuuAR4+JkkWBK7A
+dVT4n1MM4wP/aCIyriSlq7kgT03Wxk4Q4wKsj2wZvDQkNgEQjrctgIihc75jqi5
126nmT3YXJLwgDpFA3RHZUWve3j3vfUG53rRuk7K9Xx1sGWU3Ls7BuInvQZ//+QS
6UB4UuEAalmose5U8ToXQfMqZhjwreZKeb64TEZwFvu2klv4cnka1L/zHbmQGgRg
2Pfv+aUrjsYE8s9lkEKX8MIQsDn28Q5Lsv7XIEQwo2at4rYbJaxX6usuC1OI0MQ5
BLUn1GgtvOidq6FzSg6kXiA/MJYH3J0S+p4uULWAprxA+KeJRbWNRroM94W1qAk=
=1Wcq
-----END PGP SIGNATURE-----
Merge tag 'v3.13-rc3' into drm-intel-next-queued
Linux 3.13-rc3
I need a backmerge for two reasons:
- For merging the ppgtt patches from Ben I need to pull in the bdw
support.
- We now have duplicated calls to intel_uncore_forcewake_reset in the
setup code to due 2 different patches merged into -next and 3.13.
The conflict is silen so I need the merge to be able to apply
Deepak's fixup patch.
Conflicts:
drivers/gpu/drm/i915/intel_display.c
Trivial conflict, it doesn't even show up in the merge diff.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Shovel a bit more of the the code into the setup function, and call
it earlier. Otherwise lockdep is unhappy since we cancel the delayed
resume work before it's initialized.
While at it also shovel the pc8 setup code into the same functions.
I wanted to also ditch the header declaration of the hws pc8 functions,
but for unfathomable reasons that stuff is in intel_display.c instead
of intel_pm.c.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=71980
Tested-by: Guo Jinxian <jinxianx.guo@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Initialize the FBC vfuncs on gen2 and gen3 chipsets. Also make
a clean split for gen7+ vs. gen5+ vfunc initialization.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
On gen2 and gen3 chipsets FBC is supported only on plane A. Fix (and
simplify) the plane checks in intel_update_fbc() accordingly.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Atm we call intel_display_power_enabled() from
i915_capture_error_state() in IRQ context and then take a mutex. To fix
this add a new intel_display_power_enabled_sw() which returns the domain
state based on software tracking as opposed to reading the actual HW
state.
Since we use domain_use_count for this without locking on the reader
side make sure we increase the counter only after enabling all required
power wells and decrease it before disabling any of these power wells.
Regression introduced in
commit 1b02383464b4a915627ef3b8fd0ad7f07168c54c
Author: Imre Deak <imre.deak@intel.com>
Date: Tue Sep 24 16:17:09 2013 +0300
drm/i915: support for multiple power wells
Note that atm we depend on the value returned by
intel_display_power_enabled_sw() in i915_capture_error_state() to avoid
unclaimed register access reports. This was never guaranteed though,
since another thread can disable the power concurrently. If this is a
problem we need another explicit way to disable the reporting during
error captures.
v2:
- remove barriers as the caller can't depend on the value
returned from i915_capture_error_state_sw() anyway (Ville)
- dump the state of pipe/transcoder power domain state (Daniel)
Reported-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Setting this bit restores all ring contexts in parallel rather than
serially. Matches current BWG recommendations.
Tested-by: "Meng, Mengmeng" <mengmeng.meng@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Deepak S <deepak.s@inel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We use timeout mode, and we need to lower the timeout to get good RC6
residency when loads are running. This gets me from 0% residency during
glxgears to 77%, which is a pretty good improvement. This value also
matches the current BWG recommentations.
Tested-by: "Meng, Mengmeng" <mengmeng.meng@intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Deepak S <deepak.s@inel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Split vlv force wake routines to help individually control Media/Render
well based on the register access.
We've seen power savings in the lower sub-1W range on workloads that
only need on of the power wells, e.g. glbenchmark, media playback
Note: The same split isn't there for the forcewake queue, only the
forcwake domains are split.
Signed-off-by: Deepak S <deepak.s@intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
[danvet: Rebase on top of the removed forcewake hack in the ring irq
get/put code and add a note to add Deepak's answer to Chris question.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Added power well arguments to all the force wake routines
to help us individually control power well based on the
scenario.
Signed-off-by: Deepak S <deepak.s@intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
[danvet: Resolve conflict with the removed forcewake hack and drop one
spurious hunk Jesse noticed.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This value is more correct, and matches what we read out in the fastboot
code. Without this, the watermark code will panic after the first mode
setting activity after a fastboot.
v2: fix up HSW ->clock usage too (Ville)
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Add a debugfs entry showing the use-count for all power domains of each
power well.
v3: address comments from Paulo:
- simplify power_domain_str() by using a switch table
- move power_well::domain_count to power_domains
- WARN_ON decrementing a 0 refcount
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
So far we distinguished platforms without a dynamic power well with
the HAS_POWER_WELL macro and for such platforms we didn't call any power
domain functions. Instead of doing this check we can add an always-on
power well for these platforms and call the power domain functions
unconditionally. For always-on power wells we only increase/decrease
their refcounts, otherwise they are nop.
This makes high level driver code more readable and as a bonus provides
some idea of the current power domains state for all platforms (once
the relevant debugfs entry is added).
v3: rename intel_power_wells to i9xx_always_on_power_well (Paulo)
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Instead of using a separate function to check whether a power domain is
is always on, add an always-on power well covering all these power
domains and do the usual get/put on these unconditionally. Since we
don't assign a .set handler for these the get/put won't have any effect
besides the adjusted refcount.
This makes the code more readable and provides debug info also on the
use of always-on power wells (once the relevant debugfs entry is added.)
v3: make is_always_on to be bool instead of a bit field (Paulo)
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
HW generations so far had only one always-on power well and optionally
one dynamic power well. Upcoming HW gens may have multiple dynamic power
wells, so add some infrastructure to support them.
The idea is to keep the existing power domain API used by the rest of
the driver and create a mapping between these power domains and the
underlying power wells. This mapping can differ from one HW to another
but high level driver code doesn't need to know about this. Through the
existing get/put API it would just ask for a given power domain and the
power domain framework would make sure the relevant power wells get
enabled in the right order.
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This way the code is simpler and can also be used for other platforms
where the audio power domain->power well mapping is different.
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
SNB has another register where the actual FBC CPU fence number is
stored. The documenation explicitly states that the fence number
in DPFC_CTL must be 0 on SNB. And in fact when it's not zero,
the GTT tracking simply doesn't work.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
All the other .enable_fbc() funcs use plane_name(). Make
gen7_enable_fbc() do the same.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This regression has been introduced in
commit 4fe8590a92
Author: Ville Syrjälä <ville.syrjala@linux.intel.com>
Date: Wed Sep 4 18:25:22 2013 +0300
drm/i915: Use adjusted_mode appropriately when computing watermarks
I guess we should renable the enabled local variable into something a
notch more descriptive, but that's something for -next.
The effect on my i945gme netbook is pretty severe amounts of underruns
- usually the very first pixel gets used for the entire screeen.
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This reverts commit 351aa5666d.
It breaks rc6 on at least one snb machine. Since we don't yet have a
report for ivb let's keep it there for now.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=71656
Cc: Stéphane Marchesin <marcheu@chromium.org>
Cc: erik@vontaene.de
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Pull in Jani's backlight rework branch. This was merged through a
separate branch to be able to sort out the Broadwell conflicts
properly before pulling it into the main development branch.
Conflicts:
drivers/gpu/drm/i915/intel_display.c
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The pipe B and pipe C interrupt mask and enable registers are now part
of the pipe, so disabling the pipe power wells will lost the contests of
the registers.
Art totally debugged this one!
v2: Use the irq_lock to clarify code, and prevent future bugs (Daniel)
Cc: Art Runyan <arthur.j.runyan@intel.com>
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Make sparse happy.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Hold vertex data in cache until last reference
BDW-A workaround
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
BDW-A workaround
BDW Bug #1899532
v2: WARN on when not using preliminary HW support
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Jesse Barnes <jbarnes@virtuosugeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This implements a workaround for PSR dealing with some vblank issue.
WaPsrDPAMaskVBlankInSRD && WaPsrDPRSUnmaskVBlankInSRD
v2: forgot to git add bogus whitespace fix
v3: Update with workaround names.
Use for_each_pipe() and CHICKEN_PIPESL_1(pipe) macro (Ville)
Cc: Art Runyan <arthur.j.runyan@intel.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Jesse Barnes <jbarnes@virtuosugeek.org>
[danvet: Kill redundant IS_BDW check and remove the copious amount of
uneeded lines added.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This is mostly what we have for HSW with the exceptions of:
no writes:
GEN6_RC1_WAKE_RATE_LIMIT
GEN6_RC6pp_WAKE_RATE_LIMIT
GEN6_RC1e_THRESHOLD
GEN6_RC6p_THRESHOLD
GEN6_RC6pp_THRESHOLD
GEN6_RP_DOWN_TIMEOUT - use 1s instead of 1.28s
Don't try to overclock, or program ring/IA frequency tables since we
don't quite have sufficient docs yet.
NOTE: These values do not reflect the changes made recently by Chris.
Since we have no evidence yet what the proper way to tweak for this
platform is, I think it is good to go, and can be optimized by Chris, or
whomever, later.
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Drop spurious hunk and drop TODO - having per-platform rps
register frobbing code is in my opinion preferred, now that all the
infrastructure functions are extracted.]
Reviewed-by: Jesse Barnes <jbarnes@virtuosugeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The current formula we use for HSW is not what is in current docs.
However, changing to the HSW formula on my HSW does not improve power
usage, and decreases performance by about 5% in limited xonotic testing.
For gen8, until we know otherwise, or run experiments, let's use
the HSW formula - which should be the same used in the Windows driver
(and thus help make an apples-applies comparison) on gen8.
v2: Use >= 8 instead of > 7 to be consistent with all other gen
checks.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net> (v1)
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Broadwell has bigger display FIFOs than Haswell. Otherwise the
two are very similar.
v2: Fix FBC WM_LP shift for BDW
v3: Rebase on top of the big Haswell wm rework.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com> (v2)
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
GEN8 also needs this workaround.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
[danvet: Add a generic comment that we need to recheck all these w/a.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Just like Haswell, but with the small twist that the panel fitter for pipe A is
now also in the always-on power well.
v2: Use the new HAS_POWER_WELL macro.
v3: Rebase on top of intel_using_power_well patches.
v4: This time actually update the PFIT check correctly so that the
pipe A pfit is in the always-on domain.
v5: Rebase on top of the VGA power domain addition.
v6: Rebase on top of the new power domain infrastructure. Also pimp the commit
message a bit while at it.
v7: Use IS_BROADWELL instead of IS_GEN8 (Ville).
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com> (v1)
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Clock gating init is really a catch all function for registers we need
to write early in loading the driver.
Atm just the bare metal stuff we need, more will surely come.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Polling to make sure the current GPU frequency matches the last
requested frequency should not be necessay, and if there's some
throttling involved, the two might not match anyway.
Since we're still seeing this trigger occasionally, and it just
introduces a rather pointless 10 ms delay, it seems like better
to kill it off.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The RPS register writing routines use the current value of min/max to
set certain limits and interrupt gating. If we set those afterwards, we
risk setting up the hw incorrectly and losing power management events,
and worse, trigger some internal assertions.
Reorder the calling sequences to be correct, and remove the then
unrequired clamping from inside set_rps(). And for a bonus, fix the bug
of calling gen6_set_rps() from Valleyview.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
CC: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
For DDR data rate reporting by Punit in PUNIT_GPU_FREQ_STS, the actual
data encoding is 00b=800, 01b=1066, 10b=1333, 11b=1333.
Some premium VLV sku will get the DDR_DATA_RATE set as 11. As a result,
the turbo frequency reporting will be incorrect without this workaround.
Signed-off-by: Chon Ming Lee <chon.ming.lee@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We'll be looking at more than just mem_freq from dev_priv, so
just pass the whole thing.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We're currently miscalculating the VLV graphics clock a little bit.
This is caused by rounding the step to integer MHz, which does not
match reality. Change the formula to match the GUnit HAS to give
more accurate answers.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Either the docs were wrong or the values have changed since the old days
before we had wheels.
Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We don't want it delayed with the RPS work.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
I want to merge in the new Broadwell support as a late hw enabling
pull request. But since the internal branch was based upon our
drm-intel-nightly integration branch I need to resolve all the
oustanding conflicts in drm/i915 with a backmerge to make the 60+
patches apply properly.
We'll propably have some fun because Linus will come up with a
slightly different merge solution.
Conflicts:
drivers/gpu/drm/i915/i915_dma.c
drivers/gpu/drm/i915/i915_drv.c
drivers/gpu/drm/i915/intel_crt.c
drivers/gpu/drm/i915/intel_ddi.c
drivers/gpu/drm/i915/intel_display.c
drivers/gpu/drm/i915/intel_dp.c
drivers/gpu/drm/i915/intel_drv.h
All rather simple adjacent lines changed or partial backports from
-next to -fixes, with the exception of the thaw code in i915_dma.c.
That one needed a bit of shuffling to restore the intent.
Oh and the massive header file reordering in intel_drv.h is a bit
trouble. But not much.
v2: Also don't forget the fixup for the silent conflict that results
in compile fail ...
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
In
commit 6efdf354dd
Author: Imre Deak <imre.deak@intel.com>
Date: Wed Oct 16 17:25:52 2013 +0300
the check for i915_disable_power_well flag was removed by overlook,
so add it back now.
Reported-by: Paulo Zanoni <paulo.zanoni@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Similarly rename the other related functions in the power domain
interface.
Higher level driver code calling these functions knows only about power
domains, not the underlying power wells which may be different on
different platforms. Also these functions really init/cleanup/resume
power domains and only through that all related power wells, so rename
them accordingly.
Note that I left i915_{request,release}_power_well as is, since that
really changes the state only of a single power well (and is HSW
specific). It should also get a better name once we make it more
generic by controlling things through a new audio power domain.
v4:
- use intel prefix instead of i915 everywhere (Paulo)
- use a $prefix_$block_$action format (Daniel)
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The only real need for this field was in
i915_{request,release}_power_well, but there we can get at it by a
container_of magic. Also since in the future we'll have multiple power
wells each with its own power_well struct it makes sense to remove the
field from there where it'd be just redundancy.
Suggested-by: Paulo Zanoni <paulo.zanoni@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Currently we make sure that all power domains are enabled during driver
init and turn off unneded ones only after the first modeset. Similarly
during suspend we enable all power domains, which will remain on through
the following resume until the first modeset.
This logic is supported by intel_set_power_well() in the power domain
framework. It would be nice to simplify the API, so that we only have
get/put functions and make it more explicit on the higher level how this
"power well on during init" logic works. This will make it also easier
if in the future we want to shorten the time the power wells are on.
For this add a new device private flag tracking whether we have the
power wells on because of init/suspend and use only
intel_display_power_get()/put(). As nothing else uses
intel_set_power_well() we can remove it.
This also fixes
commit 6efdf354dd
Author: Imre Deak <imre.deak@intel.com>
Date: Wed Oct 16 17:25:52 2013 +0300
drm/i915: enable only the needed power domains during modeset
where removing intel_set_power_well() resulted in not releasing the
reference on the power well that was taken during init and thus leaving
the power well on all the time. Regression reported by Paulo.
v2:
- move the init_power_on flag to the power_domains struct (Daniel)
v3:
- add note about this being a regression fix too (Paulo)
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
In the future we'll need to support multiple power wells, so prepare for
that here. Create a new power domains struct which contains all
power domain/well specific fields. Since we'll have one lock protecting
all power wells, move power_well->lock to the new struct too.
No functional change.
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Production HSW does not need it. I confirmed this with Art.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Production IVB does not need it. I confirmed this with Art.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
All our registers which are written through the MCHBAR are defined
descriptively as an offset to the MCHBAR. We had 3 outliers here.
Convert these as well so all registers which are offsets are MCHBAR can
be easily identified/found within the code.
With this, convert DCLK to also follow this format.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
There is no hard need for this to be a spin lock, as we don't take these
locks in irq context from anywhere. An upcoming patch will add calls to
punit read/write functions from within regions protected by this lock
and those functions need a mutex in turn. As a solution for that convert
the spin lock to be a mutex.
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
It is just cleaner this way and makes it easier to add support for
other HW generations with always-on power wells powering a different
set of domains.
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Since we use intel_enable_rc6() now for more than just when we're
enabling RC6, we'll see this message many times, and it is just
confusing.
As an example, calc_residency calls this function whenever poked via
sysfs. This leaves the impression in dmesg that we're constantly
re-enabling RC6.
While at it, move the defines and description from drv.h to intel_pm.c,
since these are only ever used in that code.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Using the 5/6 DDB split make sense only when sprites are enabled.
So check that before we waste any cycles computing the merged
watermarks with the 5/6 DDB split.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Makes the behaviour of the function more clear.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Makes the intention more clear.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This debug print just adds overhead to the watermark merging process,
and doesn't really give enough information to be useful. Just kill
and let's add something much better a bit later.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Fill out the HSW watermark s/w tracking structures with the current
hardware state in intel_modeset_setup_hw_state(). This allows us to skip
the HW state readback during watermark programming and just use the values
we keep around in dev_priv->wm. Reduces the overhead of the watermark
programming quite a bit.
v2: s/init_wm/wm_get_hw_state
Remove stale comment about sprites
Make DDB partitioning readout safer
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
[danvet: Fix whitespace fail.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Currently hsw_write_vm_values() may write to certain watermark
registers needlessly. For instance if only, say, LP3 changes,
the current code will again disable all LP1+ watermarks even
though only LP3 needs to be reconfigured.
Add an easy to read function that will compute the dirtyness of the
watermarks, and use that information to further optimize the watermark
programming.
v2: Disable LP1+ watermarks around changing LP0 watermarks for Paulo
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
To make it easier to check what watermark updates are actually
necessary, keep copies of the relevant bits that match the current
hardware state.
Also add DDB partitioning into hsw_wm_values as that's another piece
of state we want to track.
We don't read out the hardware state on init yet, so we can't really
start using this yet, but it will be used later.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
[danvet: Paulo asked for a comment around the memcmp to say that we
depend upon zero-initializing the entire structures due to padding.
But a later patch in this series removes the memcmp again. So this is
ok as-is.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The fbc_wm_enabled member in intel_wm_config is useless for the time
being. The original idea for it was that we'd pre-compute it and so
that the WM merging process could know whether it needs to worry
about FBC watermarks at all.
But we don't have a convenient way to pre-check for the possibility
of FBC being used. intel_update_fbc() should be split up for that.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
On HSW the LP1,LP2,LP3 levels are either 1,2,3 or 1,3,4. We make the
conversion from LPn to to the level at one point current. Later we're
going to do it in a few places, so move it to a separate function.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Using the 5/6 DDB split make sense only when sprites are enabled.
So check that before we waste any cycles computing the merged
watermarks with the 5/6 DDB split.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Move the watermark max computations into haswell_update_wm(). This
allows keeping the 1/2 vs. 5/6 split code in one place, and avoid having
to pass around so many things. We also save a bit of stack space by only
requiring one copy of struct hsw_wm_maximums.
Also move the intel_wm_config out from hsw_compute_wm_parameters() and
pass it it. We'll have some need for it in haswell_update_wm() later.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Let's try to keep using the intermediate intel_pipe_wm representation
for as long as possible. It avoids subtle knowledge about the
internals of the hardware registers when trying to choose the
best watermark configuration.
While at it replace the memset() w/ zero initialization.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
I want to convert hsw_find_best_result() to use intel_pipe_wm, so we
need to move the merging to happen outside hsw_compute_wm_results().
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
No point in re-computing the watermarks for all pipes, when only one
pipe has changed. The watermarks stored under intel_crtc.wm.active are
still valid for the other pipes. We just need to redo the merging.
We can also skip the merge/update procedure completely if the new
watermarks for the affected pipe come out unchanged.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Introduce a new struct intel_pipe_wm which contains all the
watermarks for a single pipe. Use it to unify the LP0 and LP1+
watermark computations so that we can just iterate through the
watermark levels neatly and call ilk_compute_wm_level() for each.
Also add another tool ilk_wm_merge() that merges the LP1+ watermarks
from all pipes. For that, embed one intel_pipe_wm inside intel_crtc that
contains the currently valid watermarks for each pipe.
This is mainly preparatory work for pre-computing the watermarks for
each pipe and merging them at a later time. For now the merging still
happens immediately.
v2: Add some comments about level 0 DDB split and intel_wm_config
Add WARN_ON for level 0 being disabled
s/lp_wm/merged
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Needed to prevent display corruption in high res panels.
v2: use correct unit names (Rodrigo)
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Tested-by: Ulrich Drepper <drepper@gmail.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
As we delay the initial RPS enabling (upon boot and after resume), there
is a chance that we may start to render and trigger RPS boosts before we
set up the punit. Any changes we make could result in inconsistent
hardware state, with a danger of causing undefined behaviour. However,
as the boosting is a optional tweak to RPS, we can simply ignore it
whilst RPS is not yet enabled.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Let's try to avoid these confusing negated booleans.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The policy's max frequency is not equal to the CPU's max frequency. The
ring frequency is derived from the CPU frequency, and not the policy
frequency.
One example of how this may differ through sysfs. If the sysfs max
frequency is modified, that will be used for the max ring frequency
calculation.
(/sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq). As far as I
know, no current governor uses anything but max as the default, but in
theory, they could. Similarly distributions might set policy as part of
their init process.
It's ideal to use the real frequency because when we're currently scaled
up on the GPU. In this case we likely want to race to idle, and using a
less than max ring frequency is non-optimal for this situation.
AFAIK, this patch should have no impact on a majority of people.
This behavior hasn't been changed since it was first introduced:
commit 23b2f8bb92
Author: Jesse Barnes <jbarnes@virtuousgeek.org>
Date: Tue Jun 28 13:04:16 2011 -0700
drm/i915: load a ring frequency scaling table v3
CC: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The intel_flush_primary_plane name actually tells us which plane
we're talking about.
Also reorganize the internals a bit and add a missing POSTING_READ()
to make sure the hardware has seen the changes by the time we
return from the function.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The conflict in intel_drv.h tripped me up a bit since a patch in dinq
moves all the functions around, but another one in drm-next removes a
single function. So I'ev figured backing this into a backmerge would
be good.
i915_dma.c is just adjacent lines changed, nothing nefarious there.
Conflicts:
drivers/gpu/drm/i915/i915_dma.c
drivers/gpu/drm/i915/intel_drv.h
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
drm_vblank_init() is too ugly. Make it a bit easier on the eye by
collecting all the per-crtc vblank counters, timestamps etc. to
a structure and just allocate an array of those.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
After applying wait-boost we often find ourselves stuck at higher clocks
than required. The current threshold value requires the GPU to be
continuously and completely idle for 313ms before it is dropped by one
bin. Conversely, we require the GPU to be busy for an average of 90% over
a 84ms period before we upclock. So the current thresholds almost never
downclock the GPU, and respond very slowly to sudden demands for more
power. It is easy to observe that we currently lock into the wrong bin
and both underperform in benchmarks and consume more power than optimal
(just by repeating the task and measuring the different results).
An alternative approach, as discussed in the bspec, is to use a
continuous threshold for upclocking, and an average value for downclocking.
This is good for quickly detecting and reacting to state changes within a
frame, however it fails with the common throttling method of waiting
upon the outstanding frame - at least it is difficult to choose a
threshold that works well at 15,000fps and at 60fps. So continue to use
average busy/idle loads to determine frequency change.
v2: Use 3 power zones to keep frequencies low in steady-state mostly
idle (e.g. scrolling, interactive 2D drawing), and frequencies high
for demanding games. In between those end-states, we use a
fast-reclocking algorithm to converge more quickly on the desired bin.
v3: Bug fixes - make sure we reset adj after switching power zones.
v4: Tune - drop the continuous busy thresholds as it prevents us from
choosing the right frequency for glxgears style swap benchmarks. Instead
the goal is to be able to find the right clocks irrespective of the
wait-boost.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Kenneth Graunke <kenneth@whitecape.org>
Cc: Stéphane Marchesin <stephane.marchesin@gmail.com>
Cc: Owen Taylor <otaylor@redhat.com>
Cc: "Meng, Mengmeng" <mengmeng.meng@intel.com>
Cc: "Zhuang, Lena" <lena.zhuang@intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
If we encounter a situation where the CPU blocks waiting for results
from the GPU, give the GPU a kick to boost its the frequency.
This should work to reduce user interface stalls and to quickly promote
mesa to high frequencies - but the cost is that our requested frequency
stalls high (as we do not idle for long enough before rc6 to start
reducing frequencies, nor are we aggressive at down clocking an
underused GPU). However, this should be mitigated by rc6 itself powering
off the GPU when idle, and that energy use is dependent upon the workload
of the GPU in addition to its frequency (e.g. the math or sampler
functions only consume power when used). Still, this is likely to
adversely affect light workloads.
In particular, this nearly eliminates the highly noticeable wake-up lag
in animations from idle. For example, expose or workspace transitions.
(However, given the situation where we fail to downclock, our requested
frequency is almost always the maximum, except for Baytrail where we
manually downclock upon idling. This often masks the latency of
upclocking after being idle, so animations are typically smooth - at the
cost of increased power consumption.)
Stéphane raised the concern that this will punish good applications and
reward bad applications - but due to the nature of how mesa performs its
client throttling, I believe all mesa applications will be roughly
equally affected. To address this concern, and to prevent applications
like compositors from permanently boosting the RPS state, we ratelimit the
frequency of the wait-boosts each client recieves.
Unfortunately, this techinique is ineffective with Ironlake - which also
has dynamic render power states and suffers just as dramatically. For
Ironlake, the thermal/power headroom is shared with the CPU through
Intelligent Power Sharing and the intel-ips module. This leaves us with
no GPU boost frequencies available when coming out of idle, and due to
hardware limitations we cannot change the arbitration between the CPU and
GPU quickly enough to be effective.
v2: Limit each client to receiving a single boost for each active period.
Tested by QA to only marginally increase power, and to demonstrably
increase throughput in games. No latency measurements yet.
v3: Cater for front-buffer rendering with manual throttling.
v4: Tidy up.
v5: Sadly the compositor needs frequent boosts as it may never idle, but
due to its picking mechanism (using ReadPixels) may require frequent
waits. Those waits, along with the waits for the vrefresh swap, conspire
to keep the GPU at low frequencies despite the interactive latency. To
overcome this we ditch the one-boost-per-active-period and just ratelimit
the number of wait-boosts each client can receive.
Reported-and-tested-by: Paul Neumann <paul104x@yahoo.de>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68716
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Kenneth Graunke <kenneth@whitecape.org>
Cc: Stéphane Marchesin <stephane.marchesin@gmail.com>
Cc: Owen Taylor <otaylor@redhat.com>
Cc: "Meng, Mengmeng" <mengmeng.meng@intel.com>
Cc: "Zhuang, Lena" <lena.zhuang@intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
[danvet: No extern for function prototypes in headers.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This patch attempts to clean up the ring/IA scaling programming in the
following ways.
1. Fix the comment about the DDR frequency. The math is 266MHz, not
133MHz. Formula was right, docs are wrong.
2. Mask the DCLK register since I don't know how it is defined on future
platforms.
3. use mult_frac instead of magic math.
This helps for future platform enabling.
v2: Actually use the right patch. The v1 was a mix of things, none of
which was right. Note that due to rounding, we actually get different
values (slightly higher) for the effective ring frequency.
v3: Use 1.25 instead of 1.33 as the original code did. (Jesse)
CC: Jesse Barnes <jbarnes@virtuousgeek.org>
CC: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
