Added a new HAS_CORE_RING_FREQ macro, currently used in
gen6_update_ring_freq & i915_ring_freq_table debugfs function.
The programming & read of ring frequency table is needed for newer
GEN(>=6) platforms, except VLV/CHV.
Issue: VIZ-5144
Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Akash Goel <akash.goel@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Ring frequency table programming is not required on BXT. Added separate
checks to enable the programming only for SKL & skip for BXT.
v2: Removed the BXT check from gen6_update_ring_freq function
Issue: VIZ-5144
Signed-off-by: Akash Goel <akash.goel@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi at intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Currently we update the freq before masking the interrupts, which can
allow new interrupts to occur before the frequency has changed. These
extra interrupts might waste some cpu cycles. This patch corrects
this by masking interrupts prior to updating the frequency.
Note from Chris:
"Well it won't waste CPU cycles as the interrupt is also masked by the
threshold limits, but there should be no harm at all in reordering the
patch so, and it does make a certain amount of sense."
Signed-off-by: Deepak S <deepak.s@intel.com>
Signed-off-by: Praveen Paneri <praveen.paneri@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: Add note from Chris.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Ring frequency table programming changes for SKL. No need for a
floor on ring frequency, as the issue of performance impact with
ring running below DDR frequency, is believed to be fixed on SKL
v2: Removed the check for avoiding ring frequency programming for BXT (Rodrigo)
Issue: VIZ-5144
Signed-off-by: Akash Goel <akash.goel@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Read the efficient frequency (aka RPe) value through the the mailbox
command (0x1A) from the pcode, as done on Haswell and Broadwell.
The turbo minimum frequency softlimit is not revised as per the
efficient frequency value.
v2: Replaced the conditional expression operator with 'if' statement (Tom)
v3: Corrected the derivation of efficient frequency & shifted the
GEN9_FREQ_SCALER multiplications downwards (Ville)
Issue: VIZ-5143
Signed-off-by: Akash Goel <akash.goel@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Because the cool kids use dev_priv and FBC wants to be cool too.
We've been historically using struct drm_device on the FBC function
arguments, but we only really need it for intel_vgpu_active(): we can
use dev_priv everywhere else. So let's fully switch to dev_priv since
I'm getting tired of adding "struct drm_device *dev = dev_priv->dev"
everywhere.
If I get a NACK here I'll propose the opposite: convert all the
functions that currently take dev_priv to take dev.
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Burning cpu cycles isn't awesome, so use sleeps instead.
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The hardware supposedly ignores the WM1 watermarks while the PND
deadline mode is enabled, but clear out the register just in case.
This is what the other OS does, and it does make register dumps look
more consistent when we don't have partial WM1 values lingering in
the registers (some WM1 watermarks already get zeroed when the actually
used DSPFW registers get written).
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Clint Taylor <Clinton.A.Taylor@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Enabling PM5/DDR DVFS with multiple active pipes isn't a validated
configuration. It does seem to work most of the time at least, but
there is clearly an additional risk of underruns, so let's not play
with fire.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Clint Taylor <Clinton.A.Taylor@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
CxSR (or maxfifo on VLV/CHV) blocks somne changes to the plane control
register (enable bit at least, not quite sure about the rest). So in
order to have the plane enable/disable when we want we need to first
kick the hardware out of cxsr.
Unfortunateloy this requires some extra vblank waits. For the CxSR
enable after the plane update we should eventually use an async
vblank worker, but since we don't have that just do sync vblank
waits. For the disable case we have no choice but to do it
synchronously.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Clint Taylor <Clinton.A.Taylor@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
In order to get decnet memory self refresh residency on VLV, flip it
over to the new CHV way of doing things. VLV doesn't do PM5 or DDR DVFS
so it's a bit simpler.
I'm not sure the currently memory latency used for CHV is really
appropriate for VLV. Some further testing will probably be needed to
figure that out.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Clint Taylor <Clinton.A.Taylor@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Consider which planes are active and compute the FIFO split based on the
relative data rates. Since we only consider the pipe src width rather
than the plane width when computing watermarks it seems best to do the
same when computing the FIFO split as well. This means the only thing we
actually have to consider for the FIFO splut is the bpp, and we can
ignore the rest.
I've just stuffed the logic into the watermark code for now. Eventually
it'll need to move into the atomic update for the crtc.
There's also one extra complication I've not yet considered; Some of the
DSPARB registers contain bits related to multiple pipes. The registers
are double buffered but apparently they update on the vblank of any
active pipe. So doing the FIFO reconfiguration properly when multiple
pipes are active is not going to be fun. But let's ignore that mess for
now.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Clint Taylor <Clinton.A.Taylor@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Turns out the VLV/CHV system agent doesn't understand memory
latencies, so trying to rely on the PND deadline mechanism is not
going to fly especially when DDR DVFS is enabled. Currently we try to
avoid the problems by lying to the system agent about the deadlines
and setting the FIFO watermarks to 8 cachelines. This however leads to
bad memory self refresh residency.
So in order to satosfy everyone we'll just give up on the deadline
scheme and program the watermarks old school based on the worst case
memory latency.
I've modelled this a bit on the ILK+ approach where we compute multiple
sets of watermarks for each pipe (PM2,PM5,DDR DVFS) and when merge thet
appropriate one later with the watermarks from other pipes. There isn't
too much to merge actually since each pipe has a totally independent
FIFO (well apart from the mess with the partially shared DSPARB
registers), but still decopuling the pipes from each other seems like a
good idea.
Eventually we'll want to perform the watermark update in two phases
around the plane update to avoid underruns due to the single buffered
watermark registers. But that's still in limbo for ILK+ too, so I've not
gone that far yet for VLV/CHV either.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Clint Taylor <Clinton.A.Taylor@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Read out the current watermark settings from the hardware at driver init
time. This will allow us to compare the newly calculated values against
the currrent ones and potentially avoid needless WM updates.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Clint Taylor <Clinton.A.Taylor@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We want cxsr exit to happen ASAP, so toss in some POSTING_READ()s to
make sure things are really kicked off.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Clint Taylor <Clinton.A.Taylor@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Broxton is using a different register and different bit ordering
for rps status capabilities.
Also GT perf freqency register is different for Broxton so update
that.
Signed-off-by: Bob Paauwe <bob.j.paauwe@intel.com>
Reviewed-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Leftover from the big purge
commit a561165493
Author: John Harrison <John.C.Harrison@Intel.com>
Date: Thu Mar 5 14:03:03 2015 +0000
drm/i915: Remove ironlake rc6 support
Cc: John Harrison <John.C.Harrison@Intel.com>
Cc: Rob Clark <robdclark@gmail.com>
Reported-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Bspec says we shouldn't enable IPS on BDW when the pipe pixel rate
exceeds 95% of the core display clock. Apparently this can cause
underruns.
There's no similar restriction listed for HSW, so leave that one alone
for now.
v2: Add pipe_config_supports_ips() (Chris)
v3: Compare against the max cdclk insted of the current cdclk
v4: Rebased to the latest
v5: Rebased to the latest
v6: Fix for patch style problems
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83497
Tested-by: Timo Aaltonen <tjaalton@ubuntu.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Rather than reading out the current cdclk value use the cached value we
have tucked away in dev_priv.
v2: Rebased to the latest
v3: Rebased to the latest
v4: Fix for patch style problems
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAABAgAGBQJVa7zvAAoJEHm+PkMAQRiGtfMIAILs3sxFtrC1hApgcfRLF/7z
K34bwTRqErzqUO/orTwakEr9kSIpIL0zIPSryTCOTPZLfMGkQjhHXO3KR/DSbbTV
MZ8y/BM/yelFA/Np+1LjbiYjTNRnTRvCoaQihkIH8Rn02g7ob9HyL4gIGKpuGFcZ
04GacL2cgChqsRSACdNef948jCoJXKgcuDpe39DXphDWZnBKNZ3HFuJ6bryGJf9A
1/eCI4is85BNwKPemQUYR0xx83UIzDfrghatZP2mOCDDSA2MNg8HNxLTd12LGoQD
tfgX4B7aftzW9Y7GSEDfZ0IKm2NRzgPmCVj6PjVR/iI0lIK4Aq0Z/lDJxxEq3XQ=
=AJM5
-----END PGP SIGNATURE-----
Merge tag 'v4.1-rc6' into drm-next
Linux 4.1-rc6
backmerge 4.1-rc6 as some of the later pull reqs are based on newer bases
and I'd prefer to do the fixup myself.
Rename dpio_lock to sb_lock to inform the reader that its primary
purpose is to protect the sideband mailbox rather than some DPIO
state.
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 primary plane frobbing was removed from the sprite code in
commit ecce87ea3a
Author: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Date: Tue Apr 21 17:12:50 2015 +0300
drm/i915: Remove implicitly disabling primary plane for now
but the intel_flush_primary_plane() calls were left behind. Replace them
with straight forward POSTING_READ() of the sprite surface address
register.
The other user of intel_flush_primary_plane() is g4x_disable_trickle_feed()
where we can just inline the steps directly.
This allows intel_flush_primary_plane() to be killed off.
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
In commit 1854d5ca0d
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Tue Apr 7 16:20:32 2015 +0100
drm/i915: Deminish contribution of wait-boosting from clients
we removed an atomic timer based check for allowing waitboosting and
moved it below the mutex taken during RPS. However, that mutex can be
held for long periods of time on Vallyview/Cherryview as communication
with the PCU is slow. As clients may frequently wait for results (e.g.
such as tranform feedback) we introduced contention between the client
and the RPS worker. We can take advantage of the RPS worker, by
switching the wait boost decision to use spin locks and defer the
actual reclocking to the worker.
Fixes a regression of up to 45% on Baytrail and Baswell!
v2 (Daniel):
- Use max_freq_softlimit instead of the not-yet-merged boost
frequency.
- Don't inject a fake irq into the boost work, instead treat
client_boost as just another legit waker.
v3: Drop the now unused mask (Chris).
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90112
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> (v1)
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
As Daniel commented on
commit b7ffe1362c5f468b853223acc9268804aa92afc8
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Mon Apr 27 13:41:24 2015 +0100
drm/i915: Free RPS boosts for all laggards
it is better to be explicit when sharing hardcoded values such as
throttle/boost timeouts. Make it so!
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
GTT caching was disabled by default on gen8 due to not working with
big pages. Some information suggests that it got fixed, but still
GTT caching has been left disabled by default. Or could be it just
meant that the default was changed to off, and hence the problem
got solved.
Enable GTT caching in the hopes of some performance increase.
Whether or not the big pages issue has been fixed is irrelevant
at this stage since we don't use big pages.
This gives me a 1-2% improvement in xonotic on my BSW. Haven't tried
BDW, but supposedly it has larger TLBs so might not benefit as much.
On HSW GTT caching is enabled by default.
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>
GEN8_L3SQCREG1 isn't saved in the context (verified by going through
a context dump), and so we shouldn't be using the ring w/a code to
initialize it. Also Bspec explicitly talks about MMIO and writing it
with the CPU.
Additionally there's another w/a WaTempDisableDOPClkGating:bdw which
tells us to disable DOP clock gating around the GEN8_L3SQCREG1 write
to make sure everyone notices the change. So let's do that as well.
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
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 not using ilk_init_lp_watermarks() on BDW for some reason.
Probably due to the BDW patches and the relevant WM patches landing
roughlly at the same time. Fix it up.
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>
If the client stalls on a congested request, chosen to be 20ms old to
match throttling, allow the client a free RPS boost.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: s/rq/req/]
[danvet: s/0/NULL/ reported by 0-day build]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Now that we have internal clients, rather than faking a whole
drm_i915_file_private just for tracking RPS boosts, create a new struct
intel_rps_client and pass it along when waiting.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: s/rq/req/]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Since we will often pageflip to an active surface, we will often have to
wait for the surface to be written before issuing the flip. Also we are
likely to wait on that surface in plenty of time before the vblank.
Since we have a mechanism for boosting when a flip misses the expected
vblank, curtain the number of times we RPS boost when simply waiting for
mmioflip.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: s/rq/req/]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Ring switches can occur many times per frame, and are often out of
control, causing frequent RPS boosting for no practical benefit. Treat
the sw semaphore synchronisation as a separate client and only allow it
to boost once per busy/idle cycle.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: s/rq/req/]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The merged seqno->request conversion from John called request
variables req, but some (not all) of Chris' recent patches changed
those to just rq. We've had a lenghty (and inconclusive) discussion on
irc which is the more meaningful name with maybe at most a slight bias
towards req.
Given that the "don't change names without good reason to avoid
conflicts" rule applies, so lets go back to a req everywhere for
consistency. I'll sed any patches for which this will cause conflicts
before applying.
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: John Harrison <John.C.Harrison@Intel.com>
[danvet: s/origina/merged/ as pointed out by Chris - the first
mass-conversion patch was from Chris, the merged one from John.]
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
After feedback from the hardware team, now we set the GPU min/idel freq to RPe.
Punit is expecting us to operate GPU between Rpe & Rp0. If we drop the
frequency to RPn, punit is failing to change the vgg input voltage to
minimum :(
Since Punit validates the rps range [RPe, RP0]. This patch
removes unused cherryview_rps_min_freq function.
v2: Change commit message
v3: set min_freq before idle_freq (chris)
v4: Squash 'Remove unused rps min function' patch
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
It is observed on BSW that requesting a new frequency from Punit
does nothing when the GPU is in rc6, and if we let GPU enter rc6
with a high frequency, Vnn remains slightly higher than at minimum
frequency. Extending vlv_set_rps_idle() workaround on CHV/BSW.
v2: Update commit msg (Ville)
suggested-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
When GPU is idle on VLV, Request freq to punit should be good enough to
get the voltage back to VNN. Also, make sure gfx clock force applies
before requesting the freq fot vlv.
v2: Do forcewake before setting idle frequency (ville)
Update function comments to match the code (Deepak)
v3: Fix get/put across idle frequency Request. (Ville)
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=75244
suggested-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
As we're doing throughout the code, being optimistic that platform n + 1
will mostly reuse the same things as platform n allows us to minimize
the enabling work needed.
This time, it's about the number of WM levels.
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Skylake nv12 format requires dbuf (aka. ddb) calculations
and programming for each of y and uv sub-planes. Made minor
changes to reuse current dbuf calculations and programming
for uv plane. i.e., with this change, existing computation
is used for either packed format or uv portion of nv12
depending on incoming format. Added new code for dbuf
computation and programming for y plane.
This patch is a pre-requisite for adding NV12 format support.
Actual nv12 support is coming in later patches.
Signed-off-by: Chandra Konduru <chandra.konduru@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Commit c9f038a1a5 ("drm/i915: Don't assume primary & cursor are
always on for wm calculation (v4)") fixes a null pointer dereference.
Setting the primary and cursor panes to false in
ilk_compute_wm_parameters to false does however give the following
errors in the kernel log and causes the screen to flicker.
[ 101.133716] [drm:intel_set_cpu_fifo_underrun_reporting [i915]]
*ERROR* uncleared fifo underrun on pipe A
[ 101.133725] [drm:intel_cpu_fifo_underrun_irq_handler [i915]]
*ERROR* CPU pipe A FIFO underrun
Always setting the panes to enabled fixes this error.
Helped-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Thomas Gummerer <t.gummerer@gmail.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Tested-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Robert noticed that the FF_SLICE_CS_CHICKEN2 offset was wrong. Ooops.
Ville noticed that the write was wrong since FF_SLICE_CS_CHICKEN2 is a
masked register. Re-oops.
A wonder if went through 2 people while having roughly a bug per line...
The problem was introduced in the original patch:
commit 2caa3b260a
Author: Damien Lespiau <damien.lespiau@intel.com>
Date: Mon Feb 9 19:33:20 2015 +0000
drm/i915/skl: Implement WaDisableChickenBitTSGBarrierAckForFFSliceCS
v2: Also fix the register write (Ville)
Reported-by: Robert Beckett <robert.beckett@intel.com>
Reported-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Robert Beckett <robert.beckett@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Replace the hardcoded 9 with a call to intel_freq_opcode(450).
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Based on the spec, Setting up static BIAS for GPU to improve the
rps performace.
v2: rename reg defn to match spec. (Ville)
v3: Updated bias setting for chv (Deepak)
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
After feedback from the hardware team we are changing the RC6
promotional timer to increase the power saving without
changing performance.
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Separate topic branch for bxt didn't work out since we needed to
refactor the gmbus code a bit to make it look decent. So backmerge.
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Set TLBPF in TILECTL. This fixes an issue with BXT HW seeing
corrupted pte entries.
v2:
- move the workaround to bxt_init_clock_gating (imre)
Signed-off-by: Robert Beckett <robert.beckett@intel.com> (v1)
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
v2:
- Make the condition to select between SKL and BXT consistent with the
corresponding condition in init_workarounds_ring (Nick)
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
When RC6 along with Render power gating is enabled, GPU hang
happens due to lack of synchronization between GTI and Render
power gating.
v2: Updated commit message and WA name (Damien)
Change-Id: If1614206341eb52a21eadae8c5ebb2655029b50c
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Sagar Kamble <sagar.a.kamble@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
With boosting for missed pageflips, we have a much stronger indication
of when we need to (temporarily) boost GPU frequency to ensure smooth
delivery of frames. So now only allow each client to perform one RPS boost
in each period of GPU activity due to stalling on results.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Deepak S <deepak.s@linux.intel.com>
Reviewed-by: Deepak S <deepak.s@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
If we hit a vblank and see that have a pageflip queue but not yet
processed, ensure that the GPU is running at maximum in order to clear
the backlog. Pageflips are only queued for the following vblank, if we
miss it, there will be a visible stutter. Boosting the GPU frequency
doesn't prevent us from missing the target vblank, but it should help
the subsequent frames hitting theirs.
v2: Reorder vblank vs flip-complete so that we only check for a missed
flip after processing the completion events, and avoid spurious boosts.
v3: Rename missed_vblank
v4: Rebase
v5: Cancel the outstanding work in runtime suspend
v6: Rebase
v7: Rebase required fixing
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Deepak S<deepak.s@linux.intel.com>
Reviewed-by: Deepak S<deepak.s@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reuse the same reclocking strategy for Baytail as on its bigger brethren,
Sandybridge and Ivybridge. In particular, this makes the device quicker
to reclock (both up and down) though the tendency now is to downclock
more aggressively to compensate for the RPS boosts.
v2: Rebase
v3: Exclude Cherrytrail as Deepak was concerned that the increased
number of register writes would wake the common powerwell too often.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Deepak S <deepak.s@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Deepak S <deepak.s@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Both WaDisableSDEUnitClockGating and WaSetGAPSunitClckGateDisable are
needed on B0 as well.
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
../drivers/gpu/drm/i915/intel_pm.c:3185:45: warning: Initializer entry defined twice
../drivers/gpu/drm/i915/intel_pm.c:3185:52: also defined here
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Unify the HSW/BDW/SKL cdclk extraction code to conform to the same
.get_display_clock_speed() mold that all the other platforms
use.
v2: Update due to SKL code getting added
v3: Rebase on top of -nightly (introduction of intel_audio.c) (Mika Kahola)
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Mika Kahola <mika.kahola@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
[danvet: Add v3 note as suggested by Damien.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
v2: Pass in rotation info to sprite plane updates as well.
v3: Use helper to determine 90/270 rotation. (Michel Thierry)
v4: Rebased for fb modifiers and atomic changes.
For: VIZ-4546
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Michel Thierry <michel.thierry@intel.com> (v3)
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Use both up/down manual ei calcuations for symmetry and greater
flexibility for reclocking, instead of faking the down interrupt based
on a fixed integer number of up interrupts.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Deepak S<deepak.s@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Rewrite commit 31685c258e
Author: Deepak S <deepak.s@linux.intel.com>
Date: Thu Jul 3 17:33:01 2014 -0400
drm/i915/vlv: WA for Turbo and RC6 to work together.
Other than code clarity, the major improvement is to disable the extra
interrupts generated when idle. However, the reclocking remains rather
slow under the new manual regime, in particular it fails to downclock as
quickly as desired. The second major improvement is that for certain
workloads, like games, we need to combine render+media activity counters
as the work of displaying the frame is split across the engines and both
need to be taken into account when deciding the global GPU frequency as
memory cycles are shared.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Deepak S <deepak.s@linux.intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Deepak S<deepak.s@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
When we idle, we set the GPU frequency to the hardware minimum (not user
minimum). We introduce a new variable to distinguish between the
different roles, and to allow easy tuning of the idle frequency without
impacting over aspects of RPS. Setting the minimum frequency should be a
safety blanket as the pcu on the GPU should be power gating itself
anyway. However, in order for us to do set the absolute minimum
frequency, we need to relax a few of our assertions that we do not
exceed the user limits.
v2: Add idle_freq
v3: Init idle_freq for vlv and add a bunch of WARNs
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Deepak S <deepak.s@linux.intel.com>
Reviewed-by: Deepak S<deepak.s@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Enable the RPS interrupts programming(enable/disable/reset) for GEN9,
as missing changes to enable the RPS support on GEN9 have been added.
Signed-off-by: Akash Goel <akash.goel@intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
On SKL, GT frequency is programmed in units of 16.66 MHZ units compared
to 50 MHZ for older platforms. Also the time value specified for Up/Down EI &
Up/Down thresholds are expressed in units of 1.33 us, compared to 1.28
us for older platforms. So updated the gen9_enable_rps function as per that.
v2: Updated to use new macro GT_INTERVAL_FROM_US
v3: Removed the initial setup of certain registers, from gen9_enable_rps,
which gets overridden later from gen6_set_rps (Damien)
v4: Removed the enabling of rps interrupts, from gen9_enable_rps.
To be done from intel_gen6_powersave_work only, as done for other
platforms also.
Signed-off-by: Akash Goel <akash.goel@intel.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
RP Interrupt Up/Down Frequency Limits register (A014) definition
has changed for SKL. Updated the gen6_rps_limits function as per that
v2: Renamed the function to intel_rps_limits (Chris)
Signed-off-by: Akash Goel <akash.goel@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Prior to SKL, the time period programmed in Up/Down EI & Up/Down
threshold registers was in units of 1.28 micro seconds. But for
SKL, the units have changed (1.333 micro seconds).
Have generalized the implementation of gen6_set_rps_thresholds function,
by removing the hard coding done in it as per 1.28 micro seconds.
v2: Renamed the local variables & removed superfluous comments (Chris)
Signed-off-by: Akash Goel <akash.goel@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
On SKL, the frequency is programmed differently in RPNSWREQ (A008)
register (from bits 23 to 31, compared to bits 24 to 31). So updated
the gen6_set_rps function, as per this change.
Signed-off-by: Akash Goel <akash.goel@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
On SKL the frequency is specified in units of 16.66 MHZ, barring the
RP_STATE_CAP(0x5998) register, which still reports frequency in units
of 50 MHZ. So an extra conversion is required in gen6_init_rps_frequencies
function for SKL, to store the frequency values as per the actual hardware unit.
v2: Corrected the conversion from 50 to 16.66 MHZ (Ville)
Signed-off-by: Akash Goel <akash.goel@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
On SKL, frequency is specified in units of 16.66 MHZ.
Updated the intel_gpu_freq() and intel_freq_opecode() functions
to do the conversion appropriately.
Signed-off-by: Akash Goel <akash.goel@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Use the FW_WM() macro from the VLV wm code to polish up the wm
code for older gmch platforms.
Cc: Daniel Vetter <daniel@ffwll.ch>
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>
Wrap the FW register value shift+mask operations into a macro to hide
the ugliness a bit. Also might avoid bugs due to typos.
Also rename all the primary/sprite plane low order bit masks to have the
_VLV suffix, so that we can use the FW_WM_VLV() macro instead of the
FW_WM() macro for them in a consistent manner. Cursor and all the high
order bits are left to use the FW_WM() macro as there's no real
confusion with them.
Cc: Daniel Vetter <daniel@ffwll.ch>
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>
Current ILK-style watermark code assumes the primary plane and cursor
plane are always enabled. This assumption, along with the combination
of two independent commits that got merged at the same time, results in
a NULL dereference. The offending commits are:
commit fd2d61341bf39d1054256c07d6eddd624ebc4241
Author: Matt Roper <matthew.d.roper@intel.com>
Date: Fri Feb 27 10:12:01 2015 -0800
drm/i915: Use plane->state->fb in watermark code (v2)
and
commit 0fda65680e
Author: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Date: Fri Feb 27 15:12:35 2015 +0000
drm/i915/skl: Update watermarks for Y tiling
The first commit causes us to use the FB from plane->state->fb rather
than the legacy plane->fb, which is updated a bit later in the process.
The second commit includes a change that now triggers watermark
reprogramming on primary plane enable/disable where we didn't have one
before (which wasn't really correct, but we had been getting lucky
because we always calculated as if the primary plane was on).
Together, these two commits cause the watermark calculation to
(properly) see plane->state->fb = NULL when we're in the process of
disabling the primary plane. However the existing watermark code
assumes there's always a primary fb and tries to dereference it to find
out pixel format / bpp information.
The fix is to make ILK-style watermark calculation actually check the
true status of primary & cursor planes and adjust our watermark logic
accordingly.
v2: Update unchecked uses of state->fb for other platforms (pnv, skl,
etc.). Note that this is just a temporary fix. Ultimately the
useful information is going to be computed at check time and stored
right in the state structures so that we don't have to figure this
all out while we're supposed to be programming the watermarks.
(caught by Tvrtko)
v3: Fix a couple copy/paste mistakes in SKL code. (Tvrtko)
v4: Only add FB checks for ILK/SKL codepaths. Older platforms still use
intel_crtc_active() and will shortcircuit out of watermark
calculations before ever trying to dereference the primary plane's
framebuffer.
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reported-by: Michael Leuchtenburg <michael@slashhome.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=89388
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
DDR DVFS introduces massive memory latencies which can't be handled by
the PND deadline stuff. Instead the watermarks will need to be
programmed to compensate for the latency and the deadlines will need to
be programmed to tight fixed values. That means DDR DVFS can only be
enabled if the display FIFOs are large enough, and that pretty much
means we have to manually repartition them to suit the needs of the
moment.
That's a lot of change, so in the meantime let's just disable DDR DVFS
to get the display(s) to be stable.
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>
CHV has a new knob in Punit to select between some memory power savings
modes PM2 and PM5. We can allow the deeper PM5 when maxfifo mode is
enabled, so let's do so in the hopes for moar power savings.
v2: Put the thing into a separate function to avoid churn later
v3: Don't break VLV
Reviewed-by: Vijay Purushothaman <vijay.a.purushothaman@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Arun R Murthy <arun.r.murthy@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Assuming the PND deadline mechanism works reasonably we should do
memory requests as early as possible so that PND has schedule the
requests more intelligently. Currently we're still calculating
the watermarks as if VLV/CHV are identical to g4x, which isn't
the case.
The current code also seems to calculate insufficient watermarks
and hence we're seeing some underruns, especially on high resolution
displays.
To fix it just rip out the current code and replace is with something
that tries to utilize PND as efficiently as possible.
We now calculate the WM watermark to trigger when the FIFO still has
256us worth of data. 256us is the maximum deadline value supoorted by
PND, so issuing memory requests earlier would mean we probably couldn't
utilize the full FIFO as PND would attempt to return the data at
least in at least 256us. We also clamp the watermark to at least 8
cachelines as that's the magic watermark that enabling trickle feed
would also impose. I'm assuming it matches some burst size.
In theory we could just enable trickle feed and ignore the WM values,
except trickle feed doesn't work with max fifo mode anyway, so we'd
still need to calculate the SR watermarks. It seems cleaner to just
disable trickle feed and calculate all watermarks the same way. Also
trickle feed wouldn't account for the 256us max deadline value, thoguh
that may be a moot point in non-max fifo mode sicne the FIFOs are fairly
small.
On VLV max fifo mode can be used with either primary or sprite planes.
So the code now also checks all the planes (apart from the cursor)
when calculating the SR plane watermark.
We don't have to worry about the WM1 watermarks since we're using the
PND deadline scheme which means the hardware ignores WM1 values.
v2: Use plane->state->fb instead of plane->fb
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>
Existing watermark code calls intel_crtc_active() to determine whether a CRTC
is active for the purpose of watermark calculations (and bails out early if it
determines the CRTC is not active). However intel_crtc_active() only returns
true if crtc->primary->fb is non-NULL, which isn't appropriate in the modern
age of universal planes and atomic modeset since userspace can now disable the
primary plane, but leave the CRTC (and other planes) running.
Note that commit
commit 0fda65680e
Author: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Date: Fri Feb 27 15:12:35 2015 +0000
drm/i915/skl: Update watermarks for Y tiling
adds a test for primary plane enable/disable to trigger a watermark update
(previously we ignored updates to primary planes, which wasn't really correct,
but we got lucky since we always pretended the primary plane was on). Tvrtko's
patch tries to update watermarks when we re-enable the primary plane, but that
watermark computation gets aborted early because intel_crtc_active() returns
false due to the disabled primary plane.
Switch the ILK and SKL watermark code over to use crtc->state->active rather
than calling intel_crtc_active() so that we'll properly compute watermarks when
re-enabling the primary plane.
Note that this commit doesn't touch callsites in the watermark code for
older platforms since there were concerns that doing so would lead to
other types of breakage.
Also note that all of the watermark calculation at the moment takes place after
new crtc/plane states are swapped into the DRM objects. This will change in
the future, so we'll be working with in-flight state objects, but for the time
being, crtc->state is what we want to operate on.
v2: Don't drop primary->fb check from intel_crtc_active(), but rather replace
ILK/SKL callsites with direct tests of crtc->state->active. There is
concern that messing with intel_crtc_active() will lead to other breakage for
old hardware platforms. (Ville)
v3: Use intel_crtc->active for now rather than crtc->state->active since
we don't have CRTC states properly hooked up and initialized yet.
We'll defer the switch to crtc->state->active until the atomic CRTC
state work is farther along. (Ville)
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Poke at the CBR1_VLV register during init_clock_gating to make sure the
PND deadline scheme is used.
The hardware has two modes of operation wrt. watermarks:
1) PND deadline mode:
- memory request deadline is calculated from actual FIFO level * DDL
- WM1 watermark values are unused (AFAIK)
- WM watermark level defines when to start fetching data from memory
(assuming trickle feed is not used)
2) backup mode
- deadline is based on FIFO status, DDL is unused
- FIFO split into three regions with WM and WM1 watermarks, each
part specifying a different FIFO status
We want to use the PND deadline mode, so let's make sure the chicken
bit is in the correct position on init.
Also take the opportunity to refactor the shared code between VLV and
CHV to a shared function.
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
VLV/CHV have similar DSPARB registers as older platforms, just more of
them due to more planes. Add a bit of code to read out the current FIFO
split from the registers. Will be useful later when we improve the WM
calculations.
v2: Add display_mmio_offset to DSPARB
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Now that we have drm_planes for the cursor and primary we can move the
pixel_size handling into vlv_compute_drain_latency() and just pass the
appropriate plane to it.
v2: Check plane->state->fb instead of plane->fb
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org> (v1)
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
[danvet: Resolve conflict with Matt's s/plane->fb/plane->state->fb/
patch.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Introduce struct vlv_wm_values to house VLV watermark/drain latency
values. We start by using it when computing the drain latency values.
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Move the DDL precision handling into vlv_compute_drain_latency() so the
callers don't have to duplicate the same code to deal with it.
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The current drain lantency computation relies on hardcoded limits to
determine when the to use the low vs. high precision multiplier.
Rewrite the code to use a more straightforward approach.
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Kill the silly DRAIN_LATENCY_PRECISION_* defines and just use the raw
number instead.
v2: Move the sprite 32/16 -> 16/8 preision multiplier
change to another patch (Jesse)
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Apparently we must yet halve the DDL drain latency from what we're
using currently. This little nugget is not in any spec, but came
down through the grapevine.
This makes the displays a bit more stable. Not quite fully stable but at
least they don't fall over immediately on driver load.
v2: Update high_precision in valleyview_update_sprite_wm() too (Jesse)
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Apparently, this has never worked reliably and is currently disabled. Also, the
gains are not particularly impressive. Thus rather than try to keep unused code
from decaying and having to update it for other driver changes, it was decided
to simply remove it.
For: VIZ-5115
Signed-off-by: John Harrison <John.C.Harrison@Intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Implicit usage of local variables in macros isn't exactly the greatest
thing in the world, especially when that variable is the drm device and
we want to move towards a broader use of the i915 device structure.
Let's make for_each_plane() take dev_priv as its first argument then.
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Chris Wilson <chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
plane->fb is a legacy pointer that not always be up-to-date (or updated
early enough). Make sure the watermark code uses plane->state->fb so
that we're always doing our calculations based on the correct
framebuffers.
This patch was generated by Coccinelle with the following semantic
patch:
@@
struct drm_plane *P;
@@
- P->fb
+ P->state->fb
v2: Rebase
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The cursor size fields in intel_crtc just duplicate the data from
cursor->state.crtc_{w,h} so we don't need them any more. Worse, their
use in the watermark code actually introduces a subtle bug since they
don't get updated to mirror the state values until the plane commit
stage, which is *after* we've already used them to calculate new
watermark values. This happens because we had to move watermark updates
slightly earlier (outside vblank evasion) in commit
commit 32b7eeec4d
Author: Matt Roper <matthew.d.roper@intel.com>
Date: Wed Dec 24 07:59:06 2014 -0800
drm/i915: Refactor work that can sleep out of commit (v7)
Dropping the intel_crtc fields and just using the state values (which
are properly updated by the time watermark updates happen) should solve
the problem.
Aside from the actual removal of the struct fields (which are formatted
in a way that I couldn't figure out how to match in Coccinelle), the
rest of this patch was generated via the following semantic patch:
// Drop assignment
@@
struct intel_crtc *C;
struct drm_plane_state S;
@@
(
- C->cursor_width = S.crtc_w;
|
- C->cursor_height = S.crtc_h;
)
// Replace usage
@@
struct intel_crtc *C;
expression E;
@@
(
- C->cursor_width
+ C->base.cursor->state->crtc_w
|
- C->cursor_height
+ C->base.cursor->state->crtc_h
|
- to_intel_crtc(E)->cursor_width
+ E->cursor->state->crtc_w
|
- to_intel_crtc(E)->cursor_height
+ E->cursor->state->crtc_h
)
v2: Rebase
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Joe Konno <joe.konno@linux.intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=89346
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
- Y tiling support for scanout from Tvrtko&Damien
- Remove more UMS support
- some small prep patches for OLR removal from John Harrison
- first few patches for dynamic pagetable allocation from Ben Widawsky, rebased
by tons of other people
- DRRS support patches (Sonika&Vandana)
- fbc patches from Paulo
- make sure our vblank callbacks aren't called when the pipes are off
- various patches all over
* tag 'drm-intel-next-2015-02-27' of git://anongit.freedesktop.org/drm-intel: (61 commits)
drm/i915: Update DRIVER_DATE to 20150227
drm/i915: Clarify obj->map_and_fenceable
drm/i915/skl: Allow Y (and Yf) frame buffer creation
drm/i915/skl: Update watermarks for Y tiling
drm/i915/skl: Updated watermark programming
drm/i915/skl: Adjust get_plane_config() to support Yb/Yf tiling
drm/i915/skl: Teach pin_and_fence_fb_obj() about Y tiling constraints
drm/i915/skl: Adjust intel_fb_align_height() for Yb/Yf tiling
drm/i915/skl: Allow scanning out Y and Yf fbs
drm/i915/skl: Add new displayable tiling formats
drm/i915: Remove DRIVER_MODESET checks from modeset code
drm/i915: Remove regfile code&data for UMS suspend/resume
drm/i915: Remove DRIVER_MODESET checks from gem code
drm/i915: Remove DRIVER_MODESET checks in the gpu reset code
drm/i915: Remove DRIVER_MODESET checks from suspend/resume code
drm/i915: Remove DRIVER_MODESET checks in load/unload/close code
drm/i915: fix a printk format
drm/i915: Add media rc6 residency file to sysfs
drm/i915: Add missing description to parameter in alloc_pt_range
drm/i915: Removed the read of RP_STATE_CAP from sysfs/debugfs functions
...
- use the atomic helpers for plane_upate/disable hooks (Matt Roper)
- refactor the initial plane config code (Damien)
- ppgtt prep patches for dynamic pagetable alloc (Ben Widawsky, reworked and
rebased by a lot of other people)
- framebuffer modifier support from Tvrtko Ursulin, drm core code from Rob Clark
- piles of workaround patches for skl from Damien and Nick Hoath
- vGPU support for xengt on the client side (Yu Zhang)
- and the usual smaller things all over
* tag 'drm-intel-next-2015-02-14' of git://anongit.freedesktop.org/drm-intel: (88 commits)
drm/i915: Update DRIVER_DATE to 20150214
drm/i915: Remove references to previously removed UMS config option
drm/i915/skl: Use a LRI for WaDisableDgMirrorFixInHalfSliceChicken5
drm/i915/skl: Fix always true comparison in a revision id check
drm/i915/skl: Implement WaEnableLbsSlaRetryTimerDecrement
drm/i915/skl: Implement WaSetDisablePixMaskCammingAndRhwoInCommonSliceChicken
drm/i915: Add process identifier to requests
drm/i915/skl: Implement WaBarrierPerformanceFixDisable
drm/i915/skl: Implement WaCcsTlbPrefetchDisable:skl
drm/i915/skl: Implement WaDisableChickenBitTSGBarrierAckForFFSliceCS
drm/i915/skl: Implement WaDisableHDCInvalidation
drm/i915/skl: Implement WaDisableLSQCROPERFforOCL
drm/i915/skl: Implement WaDisablePartialResolveInVc
drm/i915/skl: Introduce a SKL specific init_workarounds()
drm/i915/skl: Document that we implement WaRsClearFWBitsAtReset
drm/i915/skl: Implement WaSetGAPSunitClckGateDisable
drm/i915/skl: Make the init clock gating function skylake specific
drm/i915/skl: Provide a gen9 specific init_render_ring()
drm/i915/skl: Document the WM read latency W/A with its name
drm/i915/skl: Also detect eDRAM on SKL
...
Display watermarks need different programming for different tiling
modes.
Set the relevant flag so this happens during the plane commit and
add relevant data into a structure made available to the watermark
computation code.
v2: Pass in tiling info to sprite plane updates as well.
v3: Rebased for plane handling changes.
v4: Handle fb == NULL when plane is disabled.
v5: Refactored for addfb2 interface.
v6: Refactored for fb modifier changes.
v7: Updated for atomic commit by only updating watermarks when tiling changes.
v8: BSpec watermark calculation updates.
v9: Restrict scope of y_tile_minimum variable. (Damien Lespiau)
v10: Get fb from plane state otherwise we are working on old state.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Acked-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Acked-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com> (v9)
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Recent BSpect updates have changed the watermark calculation to avoid
display flickering in some cases.
v2: Fix check against DDB allocation and tidy the code a bit. (Damien Lespiau)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
I overlooked the fact that we need to allocate a minimum 8 blocks and
that just allocating the planes depending on how much they need to fetch
from the DDB in proportion of how much memory bw is necessary for the
whole display can lead to cases where we don't respect those minima (and
thus overrun).
So, instead, start by allocating 8 blocks to each active display plane
and then allocate the remaining blocks like before.
v2: Rebase on top of -nightly
Cc: Mahesh Kumar <mahesh1.kumar@intel.com>
Signed-off-by: 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 W/A is put in a gen9 specific function because it may well be
needed on other gen9 platforms.
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Let's also take the opportunity the remove the comment telling it's a
pre-prod W/A, it should be obvious from the stepping test.
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We'll gather cross-gen9 W/A in a separate function later.
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Nick Hoath <nicholas.hoath@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
With Intel GVT-g, GPU power management is controlled by
host driver, so there is no need to provide virtualized
GPU PM support. In the future it might be useful to gather
VM input for freq boost, but now let's disable it simply.
v2:
take Chris' comments:
- do not special case this to gen6+
Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com>
Signed-off-by: Jike Song <jike.song@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
When enabling new platforms, we may not have any W/A to apply,
especially that, now, a bunch of them have to be done from the ring.
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Move Wa4x4STCOptimizationDisable to gen9_init_workarounds
v2: rebase
Signed-off-by: Nick Hoath <nicholas.hoath@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Replace the valleyview_set_rps() and gen6_set_rps() calls with
intel_set_rps() which itself does the IS_VALLEYVIEW() check. The
code becomes simpler since the callers don't have to do this check
themselves.
Most of the change was performe with the following semantic patch:
@@
expression E1, E2, E3;
@@
- if (IS_VALLEYVIEW(E1)) {
- valleyview_set_rps(E2, E3);
- } else {
- gen6_set_rps(E2, E3);
- }
+ intel_set_rps(E2, E3);
Adding intel_set_rps() and making valleyview_set_rps() and gen6_set_rps()
static was done manually. Also valleyview_set_rps() had to be moved a
bit avoid a forward declaration.
v2: Use a less greedy semantic patch
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
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 efficient frequency (RPe) should stay in the range
RPn <= RPe <= RP0. The pcode clamps the returned value
internally on Broadwell but not on Haswell.
Fix for missing range check in
commit 93ee29203f
Author: Tom O'Rourke <Tom.O'Rourke@intel.com>
Date: Wed Nov 19 14:21:52 2014 -0800
drm/i915: Use efficient frequency for HSW/BDW
Reference: http://lists.freedesktop.org/archives/intel-gfx/2015-February/059802.html
Reported-by: Michael Auchter <a@phire.org>
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@vger.kernel.org # v3.19
Signed-off-by: Tom O'Rourke <Tom.O'Rourke@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Move the CHV check into vlv_set_rps_idle() to simplify the caller a bit.
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>
Replace all the vlv_gpu_freq(), vlv_freq_opcode(),
*GT_FREQUENCY_MULTIPLIER, and /GT_FREQUENCY_MULTIPLIER instances
with intel_gpu_freq() and intel_freq_opcode() calls.
Most of the change was performed with the following semantic patch:
@@
expression E;
@@
(
- E * GT_FREQUENCY_MULTIPLIER
+ intel_gpu_freq(dev_priv, E)
|
- E *= GT_FREQUENCY_MULTIPLIER
+ E = intel_gpu_freq(dev_priv, E)
|
- E /= GT_FREQUENCY_MULTIPLIER
+ E = intel_freq_opcode(dev_priv, E)
|
- do_div(E, GT_FREQUENCY_MULTIPLIER)
+ E = intel_freq_opcode(dev_priv, E)
)
@@
expression E1, E2;
@@
(
- vlv_gpu_freq(E1, E2)
+ intel_gpu_freq(E1, E2)
|
- vlv_freq_opcode(E1, E2)
+ intel_freq_opcode(E1, E2)
)
@@
expression E1, E2, E3, E4;
@@
(
- if (IS_VALLEYVIEW(E1)) {
- E2 = intel_gpu_freq(E3, E4);
- } else {
- E2 = intel_gpu_freq(E3, E4);
- }
+ E2 = intel_gpu_freq(E3, E4);
|
- if (IS_VALLEYVIEW(E1)) {
- E2 = intel_freq_opcode(E3, E4);
- } else {
- E2 = intel_freq_opcode(E3, E4);
- }
+ E2 = intel_freq_opcode(E3, E4);
)
One hunk was manually undone as intel_gpu_freq() ended up
calling itself. Supposedly it would be possible to exclude
certain functions via !=~, but I couldn't get that to work.
Also the removal of vlv_gpu_freq() and vlv_opcode_freq() compat
wrappers was done manually.
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>
Rename the vlv_gpu_freq() and vlv_freq_opecode() functions to have
an intel_ prefix, and handle non-VLV/CHV platforms in them as well.
Leave the vlv_ names around for now since they're currently used.
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 the WIZ hashing mode to 16x4 for all the other gen6+
platfotrms, so let's follow suit on VLV.
My VLV is AWOL currently so I didn't test this, but since the results
for all the other platforms agree that 16x4 is the fastest we might
assume the same holds for VLV.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Arun Siluvery <arun.siluvery@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Drop WaDisablePwrmtrEvent:chv as it's no longer needed.
Also remove the WaSetMaskForGfxBusyness:chv note, but we still
leave the GEN6_RP_MEDIA_IS_GFX bit enabled as that's still the
recommended setting.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Arun Siluvery <arun.siluvery@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Enable coarse power gating for Gen9. This feature allows render and
media engine to enter RC6 independently. Policies are configured
together with RC6. This feature will only be enabled when RC6 is
enabled.
v2: Rebase after Chris'/Mika's forcewake change (Damien)
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Zhe Wang <zhe1.wang@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
v2: Use the new function, gen6_init_rps_frequencies() (Damien)
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> (v1)
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Per latest PM programming guide.
v2: the wrong flavour of the function updating the ring frequency was
called, leading to dead locks (Tvrtko)
v3: Add GEN6_RP_MEDIA_IS_GFX to RP_CONTROL (Imre, done by Damien)
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
[danvet: Fixup conflicts with Mika's forcewake refactor.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We have multiple forcewake domains now on recent gens. Change the
function naming to reflect this.
v2: More verbose names (Chris)
v3: Rebase
v4: Rebase
v5: Add documentation for forcewake_get/put
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Deepak S <deepak.s@linux.intel.com> (v2)
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
To match the semantics of drm_crtc->state, which this will eventually
become. The allocation of the memory for config will be fixed in a
followup patch. By adding the extra _config field to intel_crtc it was
possible to generate this entire patch with the cocci script below.
@@ @@
struct intel_crtc {
...
-struct intel_crtc_state config;
+struct intel_crtc_state _config;
+struct intel_crtc_state *config;
...
}
@@ struct intel_crtc *crtc; @@
-memset(&crtc->config, 0, sizeof(crtc->config));
+memset(crtc->config, 0, sizeof(*crtc->config));
@@ @@
__intel_set_mode(...) {
<...
-to_intel_crtc(crtc)->config = *pipe_config;
+(*(to_intel_crtc(crtc)->config)) = *pipe_config;
...>
}
@@ @@
intel_crtc_init(...) {
...
WARN_ON(drm_crtc_index(&intel_crtc->base) != intel_crtc->pipe);
+intel_crtc->config = &intel_crtc->_config;
return;
...
}
@@ struct intel_crtc *crtc; @@
-&crtc->config
+crtc->config
@@ struct intel_crtc *crtc; identifier member; @@
-crtc->config.member
+crtc->config->member
@@ expression E; @@
-&(to_intel_crtc(E)->config)
+to_intel_crtc(E)->config
@@ expression E; identifier member; @@
-to_intel_crtc(E)->config.member
+to_intel_crtc(E)->config->member
v2: Clarify manual changes by splitting them into another patch. (Matt)
Improve cocci script to generate even more of the changes. (Ander)
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The objective is to make this structure usable with the atomic helpers,
so let's start with the rename. Patch generated with coccinelle:
@@ @@
-struct intel_crtc_config {
+struct intel_crtc_state {
...
}
@@ @@
-struct intel_crtc_config
+struct intel_crtc_state
v2: Completely generate the patch with cocci. (Ander)
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The performance regression from the CHV RC6 EI->TO change is now fixed
so re-enable TO mode for better RC6 resicency.
This reverts commit e85a5c7989.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Deepak S<deepak.s@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
CherryViewA0_iGfx_BIOS_DRIVER_PUNIT_spec_y14w28d5 tells us not to enable
the RP down timeout interrupt, and says that the timeout value is hence
not used. We do enable that interrupt currently though, so leaving the
timeout as 0 results in very poor performance as the GPU frequency keeps
dropping constantly. So just program the register with the recommended
value.
Leaving the interrupt enabled doesn't seem to do any harm so far. So
I've decided to leave it on for now, just to avoid making CHV a
special case.
This fixes the performance regression from:
commit 5a0afd4b78
Author: Deepak S <deepak.s@linux.intel.com>
Date: Sat Dec 13 11:43:27 2014 +0530
drm/i915/chv: Use timeout mode for RC6 on chv
Cc: Deepak S <deepak.s@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Deepak S<deepak.s@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We use decimal for all the other RP magic values, so change
GEN6_RP_DOWN_TIMEOUT to decimal as well. Also change the order
of the register writes to match the BIOS spec for easier verification.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Deepak S<deepak.s@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Follow the sequence in the BIOS spec and clear the RC_CONTROL register
before changing any of the other RC6/RP registers.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Deepak S<deepak.s@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Use new Sideband offset to read max/min/gaur freq based on the SKU it
is running on. Based on the Number of EU, we read different bits to
identify the max frequencies at which system can run.
v2: reuse mask definitions & INTEL_INFO() to get device info (Ville)
v3: add break in switch conditions (Ville)
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala at linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
- refactor i915/snd-hda interaction to use the component framework (Imre)
- psr cleanups and small fixes (Rodrigo)
- a few perf w/a from Ken Graunke
- switch to atomic plane helpers (Matt Roper)
- wc mmap support (Chris Wilson & Akash Goel)
- smaller things all over
* tag 'drm-intel-next-2015-01-17' of git://anongit.freedesktop.org/drm-intel: (40 commits)
drm/i915: Update DRIVER_DATE to 20150117
i915: reuse %ph to dump small buffers
drm/i915: Ensure the HiZ RAW Stall Optimization is on for Cherryview.
drm/i915: Enable the HiZ RAW Stall Optimization on Broadwell.
drm/i915: PSR link standby at debugfs
drm/i915: group link_standby setup and let this info visible everywhere.
drm/i915: Add missing vbt check.
drm/i915: PSR HSW/BDW: Fix inverted logic at sink main_link_active bit.
drm/i915: PSR VLV/CHV: Remove condition checks that only applies to Haswell.
drm/i915: VLV/CHV PSR needs to exit PSR on every flush.
drm/i915: Fix kerneldoc for i915 atomic plane code
drm/i915: Don't pretend SDVO hotplug works on 915
drm/i915: Don't register HDMI connectors for eDP ports on VLV/CHV
drm/i915: Remove I915_HAS_HOTPLUG() check from i915_hpd_irq_setup()
drm/i915: Make hpd arrays big enough to avoid out of bounds access
Revert "drm/i915/chv: Use timeout mode for RC6 on chv"
drm/i915: Improve HiZ throughput on Cherryview.
drm/i915: Reset CSB read pointer in ring init
drm/i915: Drop unused position fields (v2)
drm/i915: Move to atomic plane helpers (v9)
...
Backmerge Linus tree after rc5 + drm-fixes went in.
There were a few amdkfd conflicts I wanted to avoid,
and Ben requested this for nouveau also.
Conflicts:
drivers/gpu/drm/amd/amdkfd/Makefile
drivers/gpu/drm/amd/amdkfd/kfd_chardev.c
drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c
drivers/gpu/drm/amd/amdkfd/kfd_priv.h
drivers/gpu/drm/amd/include/kgd_kfd_interface.h
drivers/gpu/drm/i915/intel_runtime_pm.c
drivers/gpu/drm/radeon/radeon_kfd.c
This reverts commit 5a0afd4b78.
Although timeout mode allows higher residency it impact badly on performance.
I believe while we don't have a way to balance between performance and
power savings at runtime I believe we have to revert and prioritize
performance that was impacted a lot.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=88103
Cc: Deepak S <deepak.s@linux.intel.com>
Cc: Wendy Wang <wendy.wang@intel.com>
Cc: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Conflicts:
drivers/gpu/drm/i915/intel_runtime_pm.c
Separate branch so that Takashi can also pull just this refactoring
into sound-next.
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
We apply the RPS interrupt workaround on VLV everywhere except when
writing the mask directly during idling the GPU. For consistency do this
also there.
While at it also extend the code comment about affected platforms.
I couldn't reproduce the issue on VLV fixed by this workaround, by
removing the workaround from everywhere, while it's 100% reproducible on
SNB using igt/gem_reset_stats/ban-ctx-render. So also add a note that
it hasn't been verified if the workaround really applies to VLV/CHV.
Suggested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
In
commit dbea3cea69
Author: Imre Deak <imre.deak@intel.com>
Date: Mon Dec 15 18:59:28 2014 +0200
drm/i915: sanitize RPS resetting during GPU reset
we disable RPS interrupts during GPU resetting, but don't apply the
necessary GEN6 HW workaround. This leads to a HW lockup during a
subsequent "looping batchbuffer" workload. This is triggered by the
testcase that submits exactly this kind of workload after a simulated
GPU reset. I'm not sure how likely the bug would have triggered
otherwise, since we would have applied the workaround anyway shortly
after the GPU reset, when enabling GT powersaving from the deferred
work.
This may also fix unrelated issues, since during driver loading /
suspending we also disable RPS interrupts and so we also had a short
window during the rest of the loading / resuming where a similar
workload could run without the workaround applied.
v2:
- separate the fix to route RPS interrupts to the CPU on GEN9 too
to a separate patch (Daniel)
Bisected-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Testcase: igt/gem_reset_stats/ban-ctx-render
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=87429
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
GEN8+ HW has the option to route PM interrupts to either the CPU or to
GT. For GEN8 this was already set correctly to routing to CPU, but not
for GEN9, so fix this. Note that when disabling RPS interrupts this was
set already correctly, though in that case it didn't matter much except
for the possibility of spurious interrupts.
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
- plane handling refactoring from Matt Roper and Gustavo Padovan in prep for
atomic updates
- fixes and more patches for the seqno to request transformation from John
- docbook for fbc from Rodrigo
- prep work for dual-link dsi from Gaurav Signh
- crc fixes from Ville
- special ggtt views infrastructure from Tvrtko Ursulin
- shadow patch copying for the cmd parser from Brad Volkin
- execlist and full ppgtt by default on gen8, for testing for now
* tag 'drm-intel-next-2014-12-19' of git://anongit.freedesktop.org/drm-intel: (131 commits)
drm/i915: Update DRIVER_DATE to 20141219
drm/i915: Hold runtime PM during plane commit
drm/i915: Organize bind_vma funcs
drm/i915: Organize INSTDONE report for future.
drm/i915: Organize PDP regs report for future.
drm/i915: Organize PPGTT init
drm/i915: Organize Fence registers for future enablement.
drm/i915: tame the chattermouth (v2)
drm/i915: Warn about missing context state workarounds only once
drm/i915: Use true PPGTT in Gen8+ when execlists are enabled
drm/i915: Skip gunit save/restore for cherryview
drm/i915/chv: Use timeout mode for RC6 on chv
drm/i915: Add GPGPU_THREADS_DISPATCHED to the register whitelist
drm/i915: Tidy up execbuffer command parsing code
drm/i915: Mark shadow batch buffers as purgeable
drm/i915: Use batch length instead of object size in command parser
drm/i915: Use batch pools with the command parser
drm/i915: Implement a framework for batch buffer pools
drm/i915: fix use after free during eDP encoder destroying
drm/i915/skl: Skylake also supports DP MST
...
Haswell significantly improved the performance of sampler_c messages,
but the optimization appears to be off by default. Later platforms
remove this bit, and apparently always enable the optimization.
Improves performance in "Counter Strike: Global Offensive" by 18%
at default settings on Iris Pro.
This may break sampling of paletted formats (P8/A8P8/P8A8). It's
unclear whether it affects sampling of paletted formats in general,
or just the sample_c message (which is never used).
While libva does have support for using paletted formats (primarily
for OSDs), that support appears to have been broken for at least a
year, so I couldn't observe a regression from this:
I tried to get libva-intel to use paletted formats, and observe a
regression...but the only thing I found that used it was mplayer's OSD
(on screen display). Even without my patch, the colors were totally
wrong with that, and it's according to a few distro wikis, that's been
the case for over a year.
If libva's code for paletted formats /is/ broken, they could always
add code to disable this bit using the command validator when fixing
it.
Further investigation from Haihao shows that libva mplayer OSD seems
to work at least on his setup (still unclear what's wron with Ken's),
and that it's not affected by this patch. Quoting the discussion
between Haihao and Ken:
> > > If you use "-vo gl" or "-vo xv", the OSD is solid white text with a black
> > > border around it. I presume that it's supposed to be white with vaapi as
> > > well, but I guess I'm not entirely sure.
> > >
> > > It's possible that the optimization doesn't affect the palette as long as
> > > you never use sample_c with the paletted textures.
> >
> > I verified the palette takes effect in the following way:
> >
> > 1. Only support P8A8 format in the driver
> >
> > 2. ran the above command and I saw white OSD text
> >
> > 3. Only support P4A4 format in the driver and don't use
> > 3DSTATE_SAMPLER_PALETTE_LOAD0 to load the value to the texture palette,
> > so the palette keeps unchanged.
> >
> > 4. ran the above command and I saw black OSD text.
> >
> > 5. Load the right value to the texture palette and ran the above command
> > again, I saw white OSD text.
> >
> > Hence I think sample_c with the paletted textures is used in the driver.
>
> That sounds like the palette is actually working, then. Great :)
>
> I doubt that libva would use sample_c - sampling with a shadow comparison?
> It looks like it just uses sample and sample+killpix.
You are right, libva driver doesn't use sample_c message.
> I'm pretty sure the sample_c optimization just uses the palette memory as
> storage for some stuff, so it's quite possible it just works if you're
> only using sample and sample+killpix.
Thanks for the explanation, it makes sense to me.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
[danvet: Add wa name from Ville's review to the comment and copypaste
the explanation why we don't care about libva (already broken) from
Ken. Also add conclusion from libva devs that&why this is all fine.]
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: "Xiang, Haihao" <haihao.xiang@intel.com>
Cc: libva@lists.freedesktop.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Higher RC6 residency is observed using timeout mode
instead of EI mode. It's Recommended to use TO Method for RC6.
v2: Add comment about timeout threshold. (Tom)
Signed-off-by: Deepak S <deepak.s@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Atm, we don't disable RPS interrupts and related work items before
resetting the GPU. This may interfere with the following GPU
initialization and cause RPS interrupts to show up in PM_IIR too early
before calling gen6_enable_rps_interrupts() (triggering a WARN there).
Solve this by disabling RPS interrupts and flushing any related work
items before resetting the GPU.
v2:
- split out the common parts of the gt suspend and the new gt reset
functions (Paulo)
v3:
- remove the check for UMS, it's a NOP nowadays (Daniel)
Reported-by: He, Shuang <shuang.he@intel.com>
Testcase: igt/gem_reset_stats/ban-render
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=86644
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
We were incorreectly bypassing the flush everytime which led to fifo
underrun when more than one plane is enabled.
Signed-off-by: Sonika Jindal <sonika.jindal@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Satheeshakrishna M<satheeshakrishna.m@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The pipe wm parameters is not correctly updated with sprite parameters
because it copies them for each plane from plane_list to the sprite
offset in pipe wm parameters. Since plane_list also contains primary and
cursor planes, we end up updating wrong params for sprites.
Signed-off-by: Sonika Jindal <sonika.jindal@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
No functional changes. This is just the begin of a FBC rework.
v2 (Paulo):
- Revert intel_fbc_init() changed parameter.
- Revert set_no_fbc_reason() rename.
- Rebase.
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
I was playing with clang and oh surprise! a warning trigerred by
-Wshift-overflow (gcc doesn't have this one):
WA_SET_BIT_MASKED(GEN7_GT_MODE,
GEN6_WIZ_HASHING_MASK | GEN6_WIZ_HASHING_16x4);
drivers/gpu/drm/i915/intel_ringbuffer.c:786:2: warning: signed shift result
(0x28002000000) requires 43 bits to represent, but 'int' only has 32 bits
[-Wshift-overflow]
WA_SET_BIT_MASKED(GEN7_GT_MODE,
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
drivers/gpu/drm/i915/intel_ringbuffer.c:737:15: note: expanded from macro
'WA_SET_BIT_MASKED'
WA_REG(addr, _MASKED_BIT_ENABLE(mask), (mask) & 0xffff)
Turned out GEN6_WIZ_HASHING_MASK was already shifted by 16, and we were
trying to shift it a bit more.
The other thing is that it's not the usual case of setting WA bits here, we
need to have separate mask and value.
To fix this, I've introduced a new _MASKED_FIELD() macro that takes both the
(unshifted) mask and the desired value and the rest of the patch ripples
through from it.
This bug was introduced when reworking the WA emission in:
Commit 7225342ab5
Author: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Date: Tue Oct 7 17:21:26 2014 +0300
drm/i915: Build workaround list in ring initialization
v2: Invert the order of the mask and value arguments (Daniel Vetter)
Rewrite _MASKED_BIT_ENABLE() and _MASKED_BIT_DISABLE() with
_MASKED_FIELD() (Jani Nikula)
Make sure we only evaluate 'a' once in _MASKED_BIT_ENABLE() (Dave Gordon)
Add check to ensure the value is within the mask boundaries (Chris Wilson)
v3: Ensure the the value and mask are 16 bits (Dave Gordon)
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Cc: Arun Siluvery <arun.siluvery@linux.intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
The conversion table can be replaced with simple enough function.
text data bss dec hex filename
839688 10987 24 850699 cfb0b drivers/gpu/drm/i915/i915.ko
839224 10987 24 850235 cf93b drivers/gpu/drm/i915/i915.ko
Result is 494 saved bytes (.05525%).
v2: - no run on sentences from subject (Chris, Jani)
- be verbose about the savings (Chris, Daniel)
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com> (v1)
Signed-off-by: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This function can be called now with i915 interrupts enabled, so the
corresponding WARN is incorrect, remove it. I think this was spotted by
Paulo during his review, but since I already removed the same WARN
from intel_suspend_gt_powersave() I missed then his point.
Spotted-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Imre Deak <imre.deak@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
I saw punit timeouts in vlv_set_rps_idle() while running various
subtests of pm_rpm. Increasing the timeout to 100ms got rid of the
issue.
Testcase: igt/pm_rpm
Reference: https://bugs.freedesktop.org/show_bug.cgi?id=82939
Signed-off-by: Imre Deak <imre.deak@intel.com>
Tested-by: Guo Jinxian <jinxianx.guo@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
In __gen6_update_ring_freq, use the full range of
possible gpu frequencies from max_freq to min_freq.
The actual gpu frequency could be outside the range
from max_freq_softlimit to min_freq_softlimit due
to power/thermal constraints.
Signed-off-by: Tom O'Rourke <Tom.O'Rourke@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
In gen8_enable_rps, change the initial rps setting
to the min_freq_softlimit (same as gen6_enable_rps).
Signed-off-by: Tom O'Rourke <Tom.O'Rourke@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Set the min_freq_softlimit to max(RPe, 450MHz).
Setting a floor can ensure a minimum experience
level. The 450MHz value came from a power and
performance study of various types of workloads
(3D, Media, GPGPU, idle, etc).
v2: rebased
Signed-off-by: Tom O'Rourke <Tom.O'Rourke@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Added gen6_init_rps_frequencies() to initialize
the rps frequency values. This function replaces
parse_rp_state_cap(). In addition to reading RPn,
RP0, and RP1 from RP_STATE_CAP register, the new
function reads efficient frequency (aka RPe) from
pcode for Haswell and Broadwell and sets the turbo
softlimits. The turbo minimum frequency softlimit
is set to RPe for Haswell and Broadwell and to RPn
otherwise.
For RPe, the efficiency is based on the frequency/power
ratio (MHz/W); this is considering GT power and not
package power. The efficent frequency is the highest
frequency for which the frequency/power ratio is within
some threshold of the highest frequency/power ratio.
A fixed decrease in frequency results in smaller
decrease in power at frequencies less than RPe than
at frequencies above RPe.
v2: Following suggestions from Chris Wilson and
Daniel Vetter to extend and rename parse_rp_state_cap
and to open-code a poorly named function.
Signed-off-by: Tom O'Rourke <Tom.O'Rourke@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
[danvet: Remove unused variables.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
So with all the code movement and extraction in intel_pm.c in -next
git is hopelessly confused with
commit 2208d655a9
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date: Fri Nov 14 09:25:29 2014 +0100
drm/i915: drop WaSetupGtModeTdRowDispatch:snb
from -fixes. Worse even small changes in -next move around the
conflict context so rerere is equally useless. Let's just backmerge
and be done with it.
Conflicts:
drivers/gpu/drm/i915/i915_drv.c
drivers/gpu/drm/i915/intel_pm.c
Except for git getting lost no tricky conflicts really.
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
After the previous patch RPS disabling doesn't depend any more on the
first level interrupts being disabled, so we can move it everywhere
earlier. Doing so let's us think about the uninitialization steps
afterwards independently of any asynchronous RPS events that can happen
atm. It also makes the system/runtime suspend time RPS disabling more
uniform. Finally this gets rid of the WARN in
intel_suspend_gt_powersave(), which we can hit if a final RPS work runs
after we disabled the first level interrupts.
Testcase: igt/pm_rpm
Reference: https://bugs.freedesktop.org/show_bug.cgi?id=82939
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>
When disabling the RPS interrupts there is a tricky dependency between
the thread disabling the interrupts, the RPS interrupt handler and the
corresponding RPS work. The RPS work can reenable the interrupts, so
there is no straightforward order in the disabling thread to (1) make
sure that any RPS work is flushed and to (2) disable all RPS
interrupts. Currently this is solved by masking the interrupts using two
separate mask registers (first level display IMR and PM IMR) and doing
the disabling when all first level interrupts are disabled.
This works, but the requirement to run with all first level interrupts
disabled is unnecessary making the suspend / unload time ordering of RPS
disabling wrt. other unitialization steps difficult and error prone.
Removing this restriction allows us to disable RPS early during suspend
/ unload and forget about it for the rest of the sequence. By adding a
more explicit method for avoiding the above race, it also becomes easier
to prove its correctness. Finally currently we can hit the WARN in
snb_update_pm_irq(), when a final RPS work runs with the first level
interrupts already disabled. This won't lead to any problem (due to the
separate interrupt masks), but with the change in this and the next
patch we can get rid of the WARN, while leaving it in place for other
scenarios.
To address the above points, add a new RPS interrupts_enabled flag and
use this during RPS disabling to avoid requeuing the RPS work and
reenabling of the RPS interrupts. Since the interrupt disabling happens
now in intel_suspend_gt_powersave(), we will disable RPS interrupts
explicitly during suspend (and not just through the first level mask),
but there is no problem doing so, it's also more consistent and allows
us to unify more of the RPS disabling during suspend and unload time in
the next patch.
v2/v3:
- rebase on patch "drm/i915: move rps irq disable one level up" in the
patchset
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>
