The idea behind keeping the saturation mask local to a context backfired
spectacularly. The premise with the local mask was that we would be more
proactive in attempting to use semaphores after each time the context
idled, and that all new contexts would attempt to use semaphores
ignoring the current state of the system. This turns out to be horribly
optimistic. If the system state is still oversaturated and the existing
workloads have all stopped using semaphores, the new workloads would
attempt to use semaphores and be deprioritised behind real work. The
new contexts would not switch off using semaphores until their initial
batch of low priority work had completed. Given sufficient backload load
of equal user priority, this would completely starve the new work of any
GPU time.
To compensate, remove the local tracking in favour of keeping it as
global state on the engine -- once the system is saturated and
semaphores are disabled, everyone stops attempting to use semaphores
until the system is idle again. One of the reason for preferring local
context tracking was that it worked with virtual engines, so for
switching to global state we could either do a complete check of all the
virtual siblings or simply disable semaphores for those requests. This
takes the simpler approach of disabling semaphores on virtual engines.
The downside is that the decision that the engine is saturated is a
local measure -- we are only checking whether or not this context was
scheduled in a timely fashion, it may be legitimately delayed due to user
priorities. We still have the same dilemma though, that we do not want
to employ the semaphore poll unless it will be used.
v2: Explain why we need to assume the worst wrt virtual engines.
Fixes: ca6e56f654 ("drm/i915: Disable semaphore busywaits on saturated systems")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Dmitry Rogozhkin <dmitry.v.rogozhkin@intel.com>
Cc: Dmitry Ermilov <dmitry.ermilov@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190618074153.16055-8-chris@chris-wilson.co.uk
Though we pin the context first before taking the pm wakeref, during
retire we need to unpin before dropping the pm wakeref (breaking the
"natural" onion). During the unpin, we may need to attach a cleanup
operation on to the engine wakeref, ergo we want to keep the engine
awake until after the unpin.
v2: Push the engine wakeref into the barrier so we keep the onion unwind
ordering in the request itself
Fixes: ce476c80b8 ("drm/i915: Keep contexts pinned until after the next kernel context switch")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190618074153.16055-1-chris@chris-wilson.co.uk
Currently, we perform a locked update of the shadow entry when
allocating a page directory entry such that if two clients are
concurrently allocating neighbouring ranges we only insert one new entry
for the pair of them. However, we also need to serialise both clients
wrt to the actual entry in the HW table, or else we may allow one client
or even a third client to proceed ahead of the HW write. My handwave
before was that under the _pathological_ condition we would see the
scratch entry instead of the expected entry, causing a temporary
glitch. That starvation condition will eventually show up in practice, so
fix it.
The reason for the previous cheat was to avoid having to free the extra
allocation while under the spinlock. Now, we keep the extra entry
allocated until the end instead.
v2: Fix error paths for gen6
Fixes: 1d1b5490b9 ("drm/i915/gtt: Replace struct_mutex serialisation for allocation")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190617140426.7203-1-chris@chris-wilson.co.uk
Move fw_info out of struct intel_package_header to allow it to grow more
easily in future. To make a cleaner move, let's also extract a function to
search the header for the dmc_offset.
While reviewing this code I wondered why we continued the search even
after finding a suitable firmware. Add a comment to explain we will
continue to try to find a more specific firmware version, even if this
is not required by the spec.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Anusha Srivatsa <anusha.srivatsa@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190607091230.1489-3-lucas.demarchi@intel.com
ICL introduces a new gamma correction mode in display engine, called
multi-segmented-gamma mode. This mode allows users to program the
darker region of the gamma curve with sueprfine precision. An
example use case for this is HDR curves (like PQ ST-2084).
If we plot a gamma correction curve from value range between 0.0 to 1.0,
ICL's multi-segment has 3 different sections:
- superfine segment: 9 values, ranges between 0 - 1/(128 * 256)
- fine segment: 257 values, ranges between 0 - 1/(128)
- corase segment: 257 values, ranges between 0 - 1
This patch:
- Changes gamma LUTs size for ICL/GEN11 to 262144 entries (8 * 128 * 256),
so that userspace can program with highest precision supported.
- Changes default gamma mode (non-legacy) to multi-segmented-gamma mode.
- Adds functions to program/detect multi-segment gamma.
V2: Addressed review comments from Ville
- separate function for superfine and fine segments.
- remove enum for segments.
- reuse last entry of the LUT as gc_max value.
- replace if() ....cond with switch...case in icl_load_luts.
- add an entry variable, instead of 'word'
V3: Addressed review comments from Ville
- extra newline
- s/entry/color/
- remove LUT size checks
- program ilk_lut_12p4_ldw value before ilk_lut_12p4_udw
- Change the comments in description of fine and coarse segments,
and try to make more sense.
- use 8 * 128 instead of 1024
- add 1 entry in LUT for GCMAX
V4: Addressed review comments from Ville
- Remove unused macro
- missing shift entry in blue
- pick correct entry for GCMAX
- Added Ville's R-B
Note: Tested and confirmed the programming sequence of odd/even
registers in the HW. The correct sequence should be:
ilk_lut_12p4_udw
ilk_lut_12p4_ldw
v5: Addressed Ville's review comments and renamed odd/even register
helpers to be more consistent with the values.
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Suggested-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Shashank Sharma <shashank.sharma@intel.com>
Signed-off-by: Uma Shankar <uma.shankar@intel.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1560321900-18318-5-git-send-email-uma.shankar@intel.com
To continue the onslaught of removing the assumption of a global
execution ordering, another casualty is the engine->timeline. Without an
actual timeline to track, it is overkill and we can replace it with a
much less grand plain list. We still need a list of requests inflight,
for the simple purpose of finding inflight requests (for retiring,
resetting, preemption etc).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190614164606.15633-3-chris@chris-wilson.co.uk
We need to keep the context image pinned in memory until after the GPU
has finished writing into it. Since it continues to write as we signal
the final breadcrumb, we need to keep it pinned until the request after
it is complete. Currently we know the order in which requests execute on
each engine, and so to remove that presumption we need to identify a
request/context-switch we know must occur after our completion. Any
request queued after the signal must imply a context switch, for
simplicity we use a fresh request from the kernel context.
The sequence of operations for keeping the context pinned until saved is:
- On context activation, we preallocate a node for each physical engine
the context may operate on. This is to avoid allocations during
unpinning, which may be from inside FS_RECLAIM context (aka the
shrinker)
- On context deactivation on retirement of the last active request (which
is before we know the context has been saved), we add the
preallocated node onto a barrier list on each engine
- On engine idling, we emit a switch to kernel context. When this
switch completes, we know that all previous contexts must have been
saved, and so on retiring this request we can finally unpin all the
contexts that were marked as deactivated prior to the switch.
We can enhance this in future by flushing all the idle contexts on a
regular heartbeat pulse of a switch to kernel context, which will also
be used to check for hung engines.
v2: intel_context_active_acquire/_release
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190614164606.15633-1-chris@chris-wilson.co.uk
