Add hw limitation checks for MDP composition once
we identify actual layers for MDP. This should be generic
way for all basic hw limitations of underlying MDP/MDSS.
Current limitations:
- a-family : MDP can not handle alpha scaling.
- 8226 & 8974 : Can not handle if multiple layers need
downscaling and blending
Change-Id: Ia944802d798f2d0e9cd4515d7e9854c7314cac78
CRs-fixed: 538803
The programYUV used for video-only cases is the same as programMDP
function. Remove the redundant function
Change-Id: I1d891404a77b2a47f90c236d4d6dcbbd02143d93
Use the low and high bandwidth values exported from driver per target
Use the low bandwidth value when the system has at least 1 video
mode panel, else use the high bandwidth value.
Newer targets could have bandwidth exceeding what an uint32 could
hold, so change the bytes read to be in gigs.
Do not use the framebuffer handle to get size, since the handle may
be null.
Change-Id: If3124ddc33334eba80d34ab7d8fc64d1b46697ce
To minimize usage of decimation when 4K x 2K video is played
on primary, split the yuv layer into 2 halves and configure
each layer to individual VG pipes. With this approach,
decimation can be avoided in some 4K x 2K video use cases,
thereby improving performance.
Change-Id: I7cddfab787c50f6ff132eaa7f5a3d9cfd636c282
GPU's perf is proportional to the number of pixels it processes, so
the load based strategy that prefers GPU, attempts to load it less
in terms of pixels, and at the same time makes sure the MDP has
enough pipes and mixer stages.
MDP's perf is real time and proportional to the instantaneous b/w
which is a function of overlaps. So the load based strategy that
prefers MDP would look at sending only as many full-screen layers
that MDP could support.
This mode is used when the GPU-preferred mode fails.
Change-Id: I3221a94bb01c43112370c248cd4ab217f2bd8ed1
Bytes claimed doesn't reflect the panel refresh rate.
So use the bandwidth claimed instead which factors this in.
This is important when 2 displays are running at different
refresh rates.
Change-Id: I3ab9573e43b8c25c31e01888fb2b181469b9e91b
Move resource checks, pipes, stages of mixer, bandwidth to a common
function and use it from all strategies.
Change-Id: I161fc957e2793d543b5943fbd23ae3607a4cfc87
When MDP cannot handle an updating layer and we marked it for GPU,
make sure framebuffer is refreshed for that frame. In the current design,
only video layers fall into these catches. But we may need
this change for future composition strategies.
Change-Id: I1d6a92da13feef4b353fa5e196d42f91148412e5
CRs-Fixed: 575340
Redraw the cache if current frame cache/drop data doesn't
match with cached frame data. Frame buffer z-order, fbcount
and mdp count are not reliable. With same fbz, fbcount and
mdpcount it is possible that layers in previous cached
frame are not part of current frame fb cache.
Change-Id: I858228dc1a3a8ab8673ab8d294e54fdbf43cc1ad
Add support for load based partial mdp comp.
This is used on geometry changes where a redraw is unavoidable.
We select a batch of layers, that has minimum pixels for FB comp,
the rest go to MDP.
Conflicts:
libhwcomposer/hwc_utils.cpp
libhwcomposer/hwc_utils.h
Change-Id: Ifc5eeb4785c75c37de97a2bb89ca81409d324691
While marking layers for cache, neglect the z-order when
possible, to improve the scope of mixed mode usage.
It is safe to neglect the z-order if a static layer
doesn't have any overlapping region with updating layers
in between static layer & contiguous static layers batch.
- if above layer doesn't have any overlapping with updating
layer in middle, push it to the batch.
- If above case fails, but layers below(existing batch) does not
have any overlapping region with the updating layer, bring the
batch to top and modify fb-zorder accordingly.
Change-Id: I4882a750d0a9a2e11272fcbb146202160673d04f
This change adds support in HWC for MDP partial update applicable
only for command mode panels. Presence of MDP FB (GRAM) in
command mode panels allows MDP clients to update only the updating region
of the frame as rest of the frame will be cached in the GRAM.
HWC calculates the updating region of a frame (ROI) by deducing
outer bounds of its updating layer destinations. Layers not lying
within the calculated ROI will be dropped from the composition
since its illegal to program MDP pipes for non participating layers.
Change-Id: I890d98ff7960fe888787981803cac62f68471201
When MDP cannot support a layer, we send the whole list of layers to
GPU. If such layers are not updating they can be cached on FB while
allowing the rest to take the MDP route.
Example # 1: Volume bar on video in 90 rotated mode. Since volume bar
is not h/w accelerated, there is no prerotation and we invoke the GPU
for redraws each round. With this change we won't because of caching.
Example # 2: If prerotation is applied to layers on primary, the
secondary layers would show a transform and the whole list is redrawn
with GPU. Now we could make use of caching.
This change also imposes a 1 layer limit on secondary, but now allows
full and partial MDP comp, thus benefiting secondary-only content.
Change-Id: I7eb43ba817f1081f317fefa99fa0c507344060e8
In video handling this patch makes sure:
1) Sufficient VG pipes are available, not just any pipes
2) If VG pipes are insufficient in multi-video scenarios
preference is given to secure videos
Change-Id: I170592463a1c28348108a1b12d60908cf3063d7d
A display could be split if its dimensions exceed 2048 or in case
of primary if we make it so via device tree property.
Add an API to find if a display is split. Use this API instead of
going through a set of conditions in various use cases.
Rename the HighRes, LowRes objects to a more appropriate Split and
NonSplit respectively
Change-Id: Id847c2bf36ea9250e1cdbc0259fdec64bd124537
Add support for checking whether the current frame's required
bandwidth can be handled by MDP.
The max limit is configurable via a property and the value is
derived from simulation runs and profiling
Conflicts:
libhwcomposer/hwc_mdpcomp.cpp
Change-Id: I46b46c346973fcb6487a8a656268171470ce69b9
Remove the unnecessary blank lock, mdp comp lock, secure lock.
Rename the ext lock to a more appropriate draw lock.
The mdp comp lock is at an incorrect place and causes unwanted
objects to show up in dumpsys, since configDone hasnt cleaned
them up yet.
dump(), blank(), draw() should all acquire a common lock.
draw() includes prepare() and set().
Change-Id: I595547dd5a393a8af6cd8c9297d50793b715e658
For split displays, earlier we allowed pipes to switch mixers in
subsequent rounds. This change prevents that and makes sure there
is one composition round where a pipe being transferred to another
mixer of the same display is UNSET
Change-Id: I3c679cc4256363eeb70c5cf8bcaf5047b8a064c2
If framebuffer needs to be configured, configure it before all the
other layers in mdp composition. This helps in cases where we are
out of SMPs and framebuffer configuration fails owing to that.
Framebuffer being the fallback path, should always get highest
priority when reserving SMP blocks
Change-Id: Ie7a6903d1b9fb98b308689c81522571449bf2e8e
Dumpsys runs in different thread not in composition thread.
Accessing mdp composition frame info in dumpsys with out
synchronizing could cause crash since frame info can get reset
in composition thread.
CRs-fixed: 503268
Change-Id: If3f7d8cc9fce8b14622e666fa52583ff0cfdedd2
Add support for unevenly split primary displays.
The driver provides info about the split via msm_fb_split sysfs node
For external we assume even split. If driver doesn't specify any split
for primary, we default to even split.
Change-Id: I4d541f41de2d7a5d2b62653fa33cab079a6d5d30
This change adds adb property support to disable mixed mode in
MDP Comp. Will be helpful in debugging issues to narrow down the
root cause.
Change-Id: I49b737bb8755dc3a35db2e1ea8ccb2043135d2b0
- Check for MAX_NUM_APP_LAYERS before updating yuv indices array.
- Fall back to GPU composition when number of app layers exceeds
MAX_NUM_APP_LAYERS to avoid heap corruption.
Change-Id: Ieb91b705a0a5f50ce2f8829d1f1ee048d44b7d2e
Reset idle fall back at the end of prepare, so that irrespective of
whether primary or external is configured first, both will fall back.
Change-Id: I6206f7cf0cccf71d1b33153044546515029f4e08
There is h/w limitation when MDP_OP_SCALE_RGB_ENHANCED bit and
MDP4_OP_SCALEX_EN or MDP_OP_SCALEY_EN bit is set.
The minimum block size is 2x2. MDP will hang if the source RGB image
height or width is equal to 1. Hence fallback to GPU in such cases.
Change-Id: I7333b097fd26bfb888865de9c39d1664bc87acca
Fix reseting on handles, they never were in most cases,
owing to wrong size.
Fix redrawing, which needs to be done also when the fb's zorder changes.
Fix mdpcount update which should happen before we assess the pipes
available versus the pipes needed.
Handle video only cases, where other UI layers could be present, but we
cannot use MDP comp for them.
Restructure MDP comp decision model and some redesign.
Change-Id: I86a2efb77edccbbb9e9d7701d06c5b2578a9d2aa
With this change we are moving MDP composition from all-or-nothing
implementation to mixed mode implementation where layers can
partially be composited through MDP and while rest of the layers are
composed / cached in framebuffer.
- Mixed mode design is based on layer caching
- Mixed mode path is configured only when non-bypassing
layers are cached.
- Never allow mixed mode when FB needs to be udpated.
- If we cannot bypass all MDP comp marked layers, bail
out completely.
Change-Id: Ie08f39db07e032b537f042d0d2bfe772ebfed049
This change adds support to MDP composition to use DMA pipes when
layer scaling is not needed. Also makes sure not to invoke
MDP rotator when DMA pipes are configured since MDP rotator will
expect both the DMA pipes to available for performing rotation.
In such cases, we let the first frame of rotated video frame to
fall back to FB.
Change-Id: I3c3697dd2713ce9b86ba1e23192ec5841abd25ec
-Add support for Video via overlay on 4k2k external panel.
-Add support for rotating videos on 4k2k panels. We use
pre-rotation in hwc to rotate a video into a single buffer,
irrespective of panel size. Then this buffer is fed to MDP.
Rotator objects are managed by the new RotMgr.
-Cleaup mdpcomp and overlay.
Change-Id: Ifb08534747e8e18b6c58dd8a3e1a9947409100f1
This patch:
-Implements MDP composition for higher resolution
panels of width more than 2k.
-Adds DMA pipe support for high res. MDP composition
Change-Id: Ib7b1f822e3600eca81521294fedfe05e948b65c3
Although 6 pipes are available in total but only up to 4 pipes are
available per display (single mixer) for MDP Composition.
CRs-Fixed: 450985
Change-Id: I99df91ab955a1a7110dfbea2aaf790ba26058a78
This is a software workaround for MDP hardware limitation where
DSI mode panels wont give more than 30 FPS if buffers in RGB
pipes are of width less than 5
MDP can't handle layers of width less than 3. Fail MDP
Composition for those frames.
Change-Id: If45efab19817ea9e724fd2857fcc1c5d98ce0eab
Restructure MDPComp class for 4kx2k displays.
Cleanup MDPComp class to remove unused stuff and formatting.
Add a MDPCompLowRes class for current functionality.
Add a hollow MDPCompHighRes class for future.
Change-Id: I343540e712d31ca907257a73b49567578f49d5f7
Does a basic implementation of the dump function
called by SurfaceFlinger in dumpsys.
Further logs in different areas can be added later on
using the dumpsys_log utility.
Change-Id: I2194b4ed7f10947272b7b824ef802ce2ee3c8a48
This change extends MDP composition to support
upto 4 layers by making border fill as base pipe and
using all MDP h/w pipes in overlay mode.
Conflicts:
libhwcomposer/hwc.cpp
libhwcomposer/hwc_utils.cpp
libhwcomposer/hwc_utils.h
Bug: 7626586
Change-Id: I01a0e53ddfbcf3ed46734f6f3bb0ef7d912ceac6
Signed-off-by: Jeykumar Sankaran <jsanka@codeaurora.org>
Signed-off-by: Iliyan Malchev <malchev@google.com>
Bug: 7124159
--External disabled
--MDP comp coded but disabled.
--Videos via overlay, so that secure videos can work.
Change-Id: Ie48b264143d5e4237ab9724e28930e3f68ba49ee
Signed-off-by: Iliyan Malchev <malchev@google.com>
SurfaceFlinger re-creates HWC worklist only for new frame and on layer geometry
change. Invalidate triggered by idle timeout refreshes the existing worklist
retaining its previous composition type. So its MDP comp's responsibility to
clear layer composition types to HWC_FRAMEBUFFER on timeout.
Change-Id: I2407c44f6aa3b11365586364e5865ae629d5326a
This change
1) implements MDP Composition upto 3 layers
2) adds invalidator support to fall back to FB
composition during idle screen condition.
Change-Id: I55f27321fd0df096b353d66aaad1cc720cd4b84b
Acked-by: Jeykumar Sankaran <jsanka@codeaurora.org>
