Doing this has been long overdue anyway, but now we really need it in
preparation for per connector backlight handling.
Signed-off-by: NJani Nikula <jani.nikula@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

On gmch platforms the normal pipe source CRC registers don't work for
DP and TV encoders. And on newer platforms the single pipe CRC has
been replaced by a set of CRC at different stages in the platform.

Now most of our userspace tests don't care one bit about the exact
CRC, they simply want something that reflects any changes on the
screen. Hence add a new auto target for platform agnostic tests to
use.

v2: Pass back the adjusted source so that it can be shown in debugfs.

v3: I seem to be unable to get a stable CRC for DP ports. So let's
just disable them for now when using the auto mode. Note that
testcases need to be restructured so that they can dynamically skip
connectors. They also first need to set up the desired mode
configuration, since otherwise the auto mode won't do the right thing.

v4: Don't leak the modeset mutex on error paths.

v5: Spelling fix for the i9xx auto_source function.

Cc: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Request by Ville in his review of the CRC stuff. This converts
everything but ilk_display_irq_handler since that needs a bit more
than a simple search&replace to look nice.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

The bbstate contains useful bits of debugging information such as
whether the batch is being read from GTT or PPGTT, or whether it is
allowed to execute privileged instructions.

v2: Only record BB_STATE for gen4+
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel 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: NPaulo Zanoni <paulo.zanoni@intel.com>
Signed-off-by: NImre Deak <imre.deak@intel.com>
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: NDaniel 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 6efdf354
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: NImre Deak <imre.deak@intel.com>
Reviewed-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: NDaniel 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: NImre Deak <imre.deak@intel.com>
Reviewed-by: NPaulo Zanoni <paulo.zanoni@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Daniel pointed out that it was hard to get anything lockless to work
correctly, so don't even try for this non critical piece of code and
just use a spin lock.

v2: Make intel_pipe_crc->opened a bool
v3: Use assert_spin_locked() instead of a comment (Daniel Vetter)
v4: Use spin_lock_irq() in the debugfs functions (they can only be
    called from process context),
    Use spin_lock() in the pipe_crc_update() function that can only be
    called from an interrupt handler,
    Use wait_event_interruptible_lock_irq() when waiting for data in the
    cicular buffer to ensure proper locking around the condition we are
    waiting for. (Daniel Vetter)
Suggested-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Adding stuff to the bottom of struct drm_i915_driver_private is
nowadays considered uncool.

Cc: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel 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: NImre Deak <imre.deak@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel 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: NImre Deak <imre.deak@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Upcoming patches will add tracking for a set of power domains via a
bitmask; to make things simple there remove the current gap in the
enum values.
Signed-off-by: NImre Deak <imre.deak@intel.com>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

On pre-gen5 and vlv we can't use the pipe source when TV-out or a DP
port is connected to the pipe. Hence we need to expose new CRC
sources.

Also simplify the existing pipe source platform code a bit by
rejecting all unhandled sources by default.
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: NDaniel 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: NBen Widawsky <ben@bwidawsk.net>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Once the machine gets to a certain point in the suspend process, we
expect the GPU to be idle. If it is not, we might corrupt memory.
Empirically (with an early version of this patch) we have seen this is
not the case. We cannot currently explain why the latent GPU writes
occur.

In the technical sense, this patch is a workaround in that we have an
issue we can't explain, and the patch indirectly solves the issue.
However, it's really better than a workaround because we understand why
it works, and it really should be a safe thing to do in all cases.

The noticeable effect other than the debug messages would be an increase
in the suspend time. I have not measure how expensive it actually is.

I think it would be good to spend further time to root cause why we're
seeing these latent writes, but it shouldn't preclude preventing the
fallout.

NOTE: It should be safe (and makes some sense IMO) to also keep the
VALID bit unset on resume when we clear_range(). I've opted not to do
this as properly clearing those bits at some later point would be extra
work.

v2: Fix bugzilla link

Bugzilla: http://bugs.freedesktop.org/show_bug.cgi?id=65496
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=59321Tested-by: NTakashi Iwai <tiwai@suse.de>
Tested-by: NPaulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Tested-By: NTodd Previte <tprevite@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We need this to work around a corruption when the boot kernel image
loads the hibernated kernel image from swap on Haswell systems -
somehow not everything is properly shut off.

This is just the prep work, the next patch will implement the actual
workaround.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
[danvet: Add a commit message suitable for -fixes and add cc: stable]
Cc: stable@vger.kernel.org
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We enable the interrupt unconditionally and only control it
through the enable bit in the CRC control register.

v2: Extract per-platform helpers to compute the register values.
Reviewed-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

This will be needed for setting the HDMI pixel clock for audio
config. No functional changes.

v2: Now with a commit message.
Signed-off-by: NJani Nikula <jani.nikula@intel.com>
Reviewed-by: NRodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

At least on linux sizeof(long) == sizeof(void*) and the thinking
is that you can grab about as many references as there's memory.

Doesn't really matter, just a bit of OCD since the fixed size data
type in a pure in-kernel datastructure look off.

v2: Ville asked for an overflow check since no one prevents userspace
from incrementing the pin count forever.

v3: s/INT/LONG/, noticed by Chris.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Assuming that all framebuffer related metadata is invariant simplifies
our userspace input data checking. And current userspace always first
updates the tiling of an object before creating a framebuffer with it.

This allows us to upconvert a check in pin_and_fence to a WARN.

In the future it should also be helpful to know which buffer objects
are potential scanout targets for e.g. frontbuffer rendering tracking
and similar things.

Note that SNA shipped for one prerelease with code which will be
broken through this patch. But users shouldn't notice since it's
purely an optimization and will transparently fall back to allocating
a new fb. i-g-t also had offending code (now fixed), but we don't
really care about breaking the test-suite.

Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Grumpily-reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We have two once very similar functions, i915_gpu_idle() and
i915_gem_idle(). The former is used as the lower level operation to
flush work on the GPU, whereas the latter is the high level interface to
flush the GEM bookkeeping in addition to flushing the GPU. As such
i915_gem_idle() also clears out the request and activity lists and
cancels the delayed work. This is what we need for unloading the driver,
unfortunately we called i915_gpu_idle() instead.

In the process, make sure that when cancelling the delayed work and
timer, which is synchronous, that we do not hold any locks to prevent a
deadlock if the work item is already waiting upon the mutex. This
requires us to push the mutex down from the caller to i915_gem_idle().

v2: s/i915_gem_idle/i915_gem_suspend/

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=70334Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Tested-by: xunx.fang@intel.com
[danvet: Only set ums.suspended for !kms as discussed earlier. Chris
noticed that this slipped through.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Also use #ifdef to keep consistent with all other such cases.

Cc: Damien Lespiau <damien.lespiau@intel.com>
Acked-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

It doesn't really make sense to have two processes dequeueing the CRC
values at the same time. Forbid that usage.
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

seq_file is not quite the right interface for these ones. We have a
circular buffer with a new entry per vblank on one side and a process
wanting to dequeue the CRC with a read().

It's quite racy to wait for vblank in user land and then try to read a
pipe_crc file, sometimes the CRC interrupt hasn't been fired and we end
up with an EOF.

So, let's have the read on the pipe_crc file block until the interrupt
gives us a new entry. At that point we can wake the reading process.
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

So we don't eat that memory when not needed.
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

There are a few good properties to a circular buffer, for instance it
has a number of entries (before we were always dumping the full buffer).
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Note the "return -ENODEV;" in pipe_crc_set_source(). The ctl file is
disabled until the end of the series to be able to do incremental
improvements.
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

There are several points in the display pipeline where CRCs can be
computed on the bits flowing there. For instance, it's usually possible
to compute the CRCs of the primary plane, the sprite plane or the CRCs
of the bits after the panel fitter (collectively called pipe CRCs).

v2: Quite a bit of rework here and there (Damien)
Signed-off-by: NShuang He <shuang.he@intel.com>
Signed-off-by: NDamien Lespiau <damien.lespiau@intel.com>
[danvet: Fix intermediate compile file reported by Wu Fengguang's
kernel builder.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

I've sent this patch several times for various reasons. It essentially
cleans up a lot of code where we need to do something per ring, and want
to query whether or not the ring exists on that hardware.

It has various uses coming up, but for now it shouldn't be too
offensive.

v2: Big conflict resolution on Damien's DEV_INFO_FOR_EACH stuff

v3: Resolved vebox addition

v4: Rebased after months of disuse. Also made failed ringbuffer init
cleaner.

v5: Remove the init cleaner from v4. There is a better way to do it.
(Chris)
Reviewed-by: NChris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Signed-off-by: NDaniel 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: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NPaulo 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: NDaniel Vetter <daniel.vetter@ffwll.ch>

Boots Just Fine (tm)!

The only glitch seems to be that at least on Fedora the boot splash
gets confused and doesn't display much at all.

And since there's no ugly console flickering anymore in between, the
flicker while switching between X servers (VT support is still enabled)
is even more jarring.

Also, I'm unsure whether we don't need to somehow kick out vgacon, now
that nothing else gets in the way. But stuff seems to work, so I
don't care. Also everything still works as well with VGA_CONSOLE=n

Also the #ifdef mess needs a bit of a cleanup, follow-up patches will
do just that.

To keep the Kconfig tidy, extract all the i915 options into its own
file.

v2:
- Rebase on top of the preliminary hw support option and the
  intel_drv.h cleanup.
- Shut up warnings in i915_debugfs.c

v3: Use the right CONFIG variable, spotted by Chon Ming.

Cc: Lee, Chon Ming <chon.ming.lee@intel.com>
Cc: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: NChon Ming Lee <chon.ming.lee@intel.com>
Signed-off-by: NDaniel 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: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

Now that MMIO has been split up into gen specific functions it is
obvious when HAS_FPGA_DBG_UNCLAIMED, HAS_FORCE_WAKE are needed. As such,
we can remove this extraneous condition.

As a result of this, as well as previously existing function pointers
for forcewake, we no longer need the has_force_wake member in the device
specific data structure.
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Reviewed-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

In preparation for having per GEN MMIO functions, create, and start
using MMIO functions in our uncore data structure. This simply makes the
transition easier by allowing us to just plug in the per GEN stuff
later.

For simplicity, I moved the intel_uncore_init() function down since
those rely on static functions defined lower in the file. This is most
of the churn in this patch.

I made one unrelated change here by using off_t datatype for the offset
of the register to write. I like the clarity that this brings to the
code. If I did it as a separate patch, I am pretty certain it would get
bikeshedded to oblivion.
Requested-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: NBen Widawsky <ben@bwidawsk.net>
Reviewed-by: NDamien Lespiau <damien.lespiau@intel.com>
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

We can get the PCI vendor and device IDs via dev->pdev. So we can drop
the duplicated information.
Signed-off-by: NVille Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: NDaniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

All drivers embed gem-objects into their own buffer objects. There is no
reason to keep drm_gem_object_alloc(), gem->driver_private and
->gem_init_object() anymore.

New drivers are highly encouraged to do the same. There is no benefit in
allocating gem-objects separately.

Cc: Dave Airlie <airlied@gmail.com>
Cc: Alex Deucher <alexdeucher@gmail.com>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Rob Clark <robdclark@gmail.com>
Cc: Inki Dae <inki.dae@samsung.com>
Cc: Ben Skeggs <skeggsb@gmail.com>
Cc: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
Signed-off-by: NDavid Herrmann <dh.herrmann@gmail.com>
Acked-by: NAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: NDave Airlie <airlied@redhat.com>

for igt test case.

v2: remove trailing spaces and fix conflicts
Signed-off-by: NRodrigo Vivi <rodrigo.vivi@gmail.com>
[danvet:
- make it comipile
- s/IS_HASWELL/HAS_PSR/]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>

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: NChris 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: NJesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: NDaniel 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: NPaul Neumann <paul104x@yahoo.de>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68716Signed-off-by: NChris 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: NJesse Barnes <jbarnes@virtuousgeek.org>
[danvet: No extern for function prototypes in headers.]
Signed-off-by: NDaniel Vetter <daniel.vetter@ffwll.ch>