Idea:
1. in its mutable recording state, keep a table of drawables on the side, and store an index in the record list.
2. In "immediate-mode" draw, just call the clients drawable directly (need access to our private list to turn the stored index into a proc)
3. when we "snap", we replace the list of drawables with a list of (sub) pictures, and then during playback of the snapped picture, we invoke a private drawable which just calls "drawPicture" on the index'd subpicture.
Review URL: https://codereview.chromium.org/727363003
Remove unused NVPR related GL tokens.
Also replace GR_GL_PATH_INITIAL_END_CAP and GR_GL_PATH_TERMINAL_END_CAP
with a single call setting GR_GL_PATH_END_CAPS. Skia does not and
probably will not have different initial and terminal caps. This came up
in the review of command buffer implementation of the extension.
If more NVPR features will be used, the respective tokens can be added
back per implemented feature.
Review URL: https://codereview.chromium.org/723453002
This CL fixes the case where a bad initial vector (i.e., nearly zero) managed to short circuit all of the convexicator's logic. The initial bad vector would become the last vector and then never get displaced.
The history of this is:
https://codereview.chromium.org/298973004/
Switched the convexicator to not advance the last vector when the cross product wasn't significant
https://codereview.chromium.org/573763002/
Fixed a bug (crbug.com/412640) wherein a zero area path was being incorrectly categorized as convex b.c. opposite but equal vectors were not signaling concavity.
BUG=433683
Review URL: https://codereview.chromium.org/727283003
Chromium creates a picture to contain their picture pile in order to use MultiPictureDraw. They currently do not create a bounding box for that picture but I still need layer information for it. This change allows Chromium to continue without a BBH but still have layer information.
In the future, the brute force BBH might be suitable for their use case.
Please see gpu_raster_worker_pool.cc in (Add flag to beginRecording to request saveLayer information - https://codereview.chromium.org/721883002/) for where this is happening in Chromium.
Review URL: https://codereview.chromium.org/733963004
- move field declarations together and pack them a little tighter
- get rid of fData
- remove dead code in debugger, including unused SkPicturePlayback subclass
There are now no more long-lived SkPictureData! (Really, there never were,
but now we don't pretend to support them.)
BUG=skia:
No API changes.
TBR=reed@google.com
Review URL: https://codereview.chromium.org/725143002
This will allow us to add nonnull-attribute to the UBSAN bot.
We are in fact hitting a case where one of the arguments is null and the other
not, which seems dicey. I think the scenario is comparing the empty pathref
with another path ref that's just been COWed, without any verbs or points yet.
BUG=skia:
Review URL: https://codereview.chromium.org/732643002
We're currently overwriting the paint LCD text flag based on the the run
font data => this cancels any LCD filtering we might have performed
higher up the stack.
BUG=423362
R=reed@google.com
Review URL: https://codereview.chromium.org/718913003
This means we can store fLgMinSize in 4 bits (TBD).
Local perf comparison calls this harmless-to-slightly-helpful. Nothing to get
excited about, but seems to certainly not harm perf.
BUG=skia:
Review URL: https://codereview.chromium.org/722293003
This define was added with "Always round text position correctly."
9447103029 . The affected clients
have been rebaselined and this is no longer defined anywhere.
Review URL: https://codereview.chromium.org/722333002
This CL updates various files in the includes directory to ensure that (1) they do
not depend on headers in /src and (2) that they minimize their dependence on external
headers.
To ensure that we don't regress this behavior a new build target has been added to
build a single cpp file that contains all* public includes and is compiled with
only those directories in the include path.
* The exception is those includes that depend on OS specific headers
BUG=skia:2941
NOTRY=true
Review URL: https://codereview.chromium.org/721903002
SkRecord performance is not sensitive to these values, so we can cut some
storage. This rearranges the space-remaining logic to use a count of bytes
left rather than a pointer to the end, cutting memory usage a little more.
An SkVarAlloc used to weigh 20 or 32 bytes which now becomes 16 or 24.
I think if I think about it a bit more I can trim off that Block* too,
getting us to 12 or 16 bytes.
Because we now just always grow by doubling, this CL switches from storing
fSmallest to its log base 2. This has the nice effect of never having to worry
about it overflowing, and means we can probably squeeze it down into a byte
if we want, even 6 bits.
BUG=skia:
Committed: https://skia.googlesource.com/skia/+/bc415389855888af5a1282ca4b6bee30afa3d69d
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu12-ShuttleA-GTX660-x86-Debug-Trybot
Review URL: https://codereview.chromium.org/721313002
Under the hood, add SkPixelGeometry to the CreateInfo for new devices, allowing them to see their geometry (SkDeviceProperties) up front, rather than having it changed later.
The only exception is for devices that are used on the root-layer, where we don't see the device until after the fact (at least as long as we allow clients to attach a device to a canvas externally).
We also filter the geometry when we're creating a layer, so we can disable LCD text automatically if the layer is not marked as opaque.
NOTRY=True
-- gammatext flake?
Review URL: https://codereview.chromium.org/719253002
Combines adjacent DrawPaths commands into one single call when a
conservative set of conditions are met. The end result is that whole
paragraphs can be drawn with a single call to
gliStencilThenCoverFillPathInstancedNV(), rather than one call for
each line.
BUG=skia:
Review URL: https://codereview.chromium.org/712223002
Reason for revert:
Unit test failures on 32-bit machines.
test Record_Alignment: ../../tests/RecordTest.cpp:100 is_aligned(record.alloc<uint64_t>())
Original issue's description:
> Deparameterize SkVarAlloc.
>
> SkRecord performance is not sensitive to these values, so we can cut some
> storage. This rearranges the space-remaining logic to use a count of bytes
> left rather than a pointer to the end, cutting memory usage a little more.
>
> An SkVarAlloc used to weigh 20 or 32 bytes which now becomes 16 or 24.
>
> I think if I think about it a bit more I can trim off that Block* too,
> getting us to 12 or 16 bytes.
>
> Because we now just always grow by doubling, this CL switches from storing
> fSmallest to its log base 2. This has the nice effect of never having to worry
> about it overflowing, and means we can probably squeeze it down into a byte
> if we want, even 6 bits.
>
> BUG=skia:
>
> Committed: https://skia.googlesource.com/skia/+/bc415389855888af5a1282ca4b6bee30afa3d69dTBR=reed@google.com,mtklein@chromium.org
NOTREECHECKS=true
NOTRY=true
BUG=skia:
Review URL: https://codereview.chromium.org/718203006
SkRecord performance is not sensitive to these values, so we can cut some
storage. This rearranges the space-remaining logic to use a count of bytes
left rather than a pointer to the end, cutting memory usage a little more.
An SkVarAlloc used to weigh 20 or 32 bytes which now becomes 16 or 24.
I think if I think about it a bit more I can trim off that Block* too,
getting us to 12 or 16 bytes.
Because we now just always grow by doubling, this CL switches from storing
fSmallest to its log base 2. This has the nice effect of never having to worry
about it overflowing, and means we can probably squeeze it down into a byte
if we want, even 6 bits.
BUG=skia:
Review URL: https://codereview.chromium.org/721313002
The fixes include
- detect when finding the active top loops between two possible answers
- preflight chasing winding to ensure answer is consistent
- binary search more often when quadratic intersection fails
- add more failure paths when an intersect is missed
While this fixes the chrome bug, reenabling path ops in svg should be deferred until additional fixes are landed.
TBR=
BUG=421132
Committed: https://skia.googlesource.com/skia/+/6f726addf3178b01949bb389ef83cf14a1d7b6b2
Review URL: https://codereview.chromium.org/633393002
Allow the srcCoeff to be anything as long as it does not reference the dst. Previous version
required srcCoeff to be one.
BUG=skia:
Review URL: https://codereview.chromium.org/718103003
Like SkChunkAlloc, but
- does its allocation with better sympathy for malloc granularity;
- the fast path inlines entirely;
- smaller per-block overhead;
- smaller per-SkVarAlloc overhead;
- growth parameters are a little more tunable.
Its main downside is less flexibility; it supports fewer methods than SkChunkAlloc.
These current parameters bring the first allocation down from 4K to 1K,
without affecting recording time on my desktop. skiaperf.com will tell the
whole story.
BUG=skia:
Review URL: https://codereview.chromium.org/674263002
the new virtual takes a struct which we can amend in the future w/o having to
update our subclasses in chrome.
BUG=skia:
NOTRY=True
Review URL: https://codereview.chromium.org/723743002
This will help with the ability to subclass, add to, etc. GrInvariantOutput. Also it was simply
getting a little too big to be a "supporting" subclass
BUG=skia:
Review URL: https://codereview.chromium.org/699943003
This define was added in
"SK_USE_SCALED_FONTMETRICS for correct scaling"
c17c6582ec.
Users all now define this flag, so it may now be removed.
BUG=chromium:420901
Review URL: https://codereview.chromium.org/720743003
Now that CollectLayers directly uses FillBounds we can:
skip the explicit intersection with the clipBounds after an adjustAndMap call
skip the storage and use of the clipBounds in SaveLayerInfo
Review URL: https://codereview.chromium.org/719793002
Since we now snap and create the ODS one time in the inorder draw buffer,
there is no need for us to keep a cached version of it around.
BUG=skia:
Review URL: https://codereview.chromium.org/701123003
work on tests
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu13.10-GCE-NoGPU-x86_64-Debug-ASAN-Trybot,Test-Ubuntu12-ShuttleA-GTX660-x86-Debug-Trybot,Test-Win7-ShuttleA-HD2000-x86_64-Debug-Trybot,Test-Win7-ShuttleA-HD2000-x86-Debug-Trybot
BUG=skia:
Review URL: https://codereview.chromium.org/704923003
This CL:
1) removes the EXPERIMENTAL_optimize on SkCanvas & SkDevice
2) moves the saveLayer gathering step to endRecording
3) Replaces GPUOptimize with SkRecordComputeLayers
4) Update bench_pictures & render_pictures to provide the new flag
#2 also necessitated moving the BBH computation (and record optimization) out of SkPicture's ctor (and into endRecording)
Review URL: https://codereview.chromium.org/718443002
This removes the old guarded code and enables the new api
introduced with "Update fontMgr to take list of bcp47 language tags."
c20386e393 . Blink on Android is
already using the new code.
Review URL: https://codereview.chromium.org/705843004
We previously saw crashes decoding bad ICO files. Add tests for
known bad files.
While testing, I learned that one of them still crashes. Check for
large offset and size separately to fix the crash.
BUG=skia:2878
Review URL: https://codereview.chromium.org/712123002
intended uses:
- flag a SkSurface as sRGB (only supported by Ganesh for now)
- flag images (e.g. png or jpeg) as sRGB if the codec tells us that
wins:
- faster gamma-correct text (esp. w/ distance-fields) when we can use sRGB for text
- better color fidelity when the screen really is sRGB
Review URL: https://codereview.chromium.org/676883003
This CL removes CollectLayers' reliance on having the top most picture (by removing the unused fPictureID member). This then allows making CollectLayers' API closer to that of SkRecordFillBounds in order to facilitate using them interchangeably.
Review URL: https://codereview.chromium.org/714533002
This became relevant whilst attempting to rebaseline the multipicturedraw GMs after turning on layer hoisting inside Skia.
Review URL: https://codereview.chromium.org/709943003
Rather than freeing almost all of its memory on calls to reset(), this
change updates GrTRecorder so it keeps around enough to satisfy the
storage requirements from last time, plus up to ~50% growth. This is
based on the assumption that subsequent draw calls require roughly the
same amount of memory.
BUG=skia:
Review URL: https://codereview.chromium.org/684203003
Removes the case for x-only glyph positions from nvpr text, opting to
always send 2d glyph positions instead. The 1d glyph positions saved a
bit on memory bandwidth, but ended up a net loss because they required
more updates to the view matrix. Now we can draw an entire paragraph
without touching the GL state, whereas before we would have to update
the view matrix at every new line.
BUG=skia:
Review URL: https://codereview.chromium.org/700283002
Reason for revert:
Not compiling in ANGLE build
Original issue's description:
> Get gpudft support working in dm, gm, nanobench and bench_pictures
>
> Adds a new config to test distance field text.
> Clean up some flags and #defines to read "distance field text",
> not "distance field fonts" to be consistent with Chromium
>
> NOTREECHECKS=true
>
> Committed: https://skia.googlesource.com/skia/+/06ba179838ba4fe187cf290750aeeb4a02a2960bTBR=bsalomon@google.com,mtklein@google.com,reed@google.com
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.chromium.org/707723005
Adds a new config to test distance field text.
Clean up some flags and #defines to read "distance field text",
not "distance field fonts" to be consistent with Chromium
NOTREECHECKS=true
Review URL: https://codereview.chromium.org/699453005
Chrome's tracing framework appears to be intentionally racy on its
quick-reject checks, trading some data loss for better performance
when disabled. People will never notice the data loss, but TSAN does.
Let's assuage TSAN with some annotations.
The 'volatile' val in SK_ANNOTATE_UNPROTECTED_WRITE was making this
not compile, but that volatile doesn't really make sense there: the value we're
writing is not what we care about, it's the destination.
CQ_EXTRA_TRYBOTS=client.skia:Test-Ubuntu13.10-GCE-NoGPU-x86_64-Release-TSAN-Trybot
No API changes.
TBR=reed
BUG=skia:
Review URL: https://codereview.chromium.org/702883002
In cull_line we must also check if both points are the same. Otherwise we
fail the assert in the else "SkASSERT(dy && !dx)".
This is currently blocking the roll as it fails a webkit test.
BUG=skia:
Review URL: https://codereview.chromium.org/703783002
Rendering a glyph with a path wants to place it at the (sx, sy) we get
as input to the method, but we add (dx, dy) for the clipRect check.
Hence, we need to subtract that out before we render the path.
Review URL: https://codereview.chromium.org/699283003
This removes the code which forces 400 and 700 weights only,
and also overrides the font weight with the configured weight.
Review URL: https://codereview.chromium.org/694533006
For the Android framework build, we get our defines from
SkUserConfig, rather than from the makefile, so we need to
include it (via SkTypes) before we can use our defines.
Fixes Android framework build.
Review URL: https://codereview.chromium.org/700893002
The issue was that, with caching disabled, the layer cache code was removing layers outside of a purge (i.e., in unlock) but not correctly cleaning up the GrPictureInfo objects (as is done in purgePlot).
Review URL: https://codereview.chromium.org/703643002
Tested by running DM on XP. Before this patch, it fails at startup (even just out/Debug/dm --help). Now it asserts for other reasons later on in user code, which is just fine by me.
The net effect is that SkTaskGroups will always be synchronous on XP. That's not ideal, but a step up from crashing.
CQ_EXTRA_TRYBOTS=client.skia:Test-Win7-ShuttleA-HD2000-x86-Release-Trybot,Test-Win7-ShuttleA-HD2000-x86_64-Release-Trybot
BUG=skia:
Review URL: https://codereview.chromium.org/700683002
- Update spacing of LODs to get better results scaling up to 256
- Remove unnecessary "bolding" from dftext
- Add debug colors for dftext LODs
BUG=skia:2933,skia:2935
Review URL: https://codereview.chromium.org/703463002
The calls to visit() to execute the SkRecord::Draw::operator() code path
is not clear to read, so adding some comments to help other new-comers
follow this through to the SkCanvas calls.
R=mtklein@chromium.org
Review URL: https://codereview.chromium.org/695403003
This CL shrinks the bound computed for saveLayers that possess both an explicit
bound and a complex paint (e.g., one that affects transparent black). In this
case the bound of the layer should be the clipped explicit bound rather then
the clip prior/after the saveLayer/restore block.
In the following the first bound is the currently computed bound
while the second is the new/desired one:
For a 100x100 picture
saveLayer (no bound, no paint) [ 0 0 100 100 ] [ 50 50 100 100 ]
setMatrix (translate 50, 50) [ 0 0 100 100 ] [ 50 50 100 100 ]
saveLayer (bound of 0, 0, 50, 50 - complex paint) [ 0 0 100 100 ] [ 50 50 100 100 ]
restore [ 0 0 100 100 ] [ 50 50 100 100 ]
restore [ 0 0 100 100 ] [ 50 50 100 100 ]
Review URL: https://codereview.chromium.org/696763002
- Add clipRect check
- Remove creation of scalerContext to check for color fonts in canDraw()
(no longer needed)
BUG=skia:2933
Review URL: https://codereview.chromium.org/696503004
When an error occurs reading a flattenable object, it should be unrefed instead of deleted because, in the case of SkXferMode, for example, we'd actually be deleting a cached object kept in a static array.
BUG=428789
Review URL: https://codereview.chromium.org/695633003
This change is here since previously color bitmap text was rendered using a
geometry processor in the coverage stage. The problem with this is that we
cannot correctly do xfer modes with this method. So I now make color bitmap text
draw using a color stage in the same was as a draw bitmap call.
One issue that arrises from this fix is that we end up adding this final color
processor after any previous color processors. Thus if we have a custom blend
implemented as a color processor it will be before this text one and we won't
blend correctly. This issue will get fixed once an xfer processor is fully
implemented. I have hacked a test locally to show that if we can add the text
color processor to the begining of the color stages we do blend correctly in all
cases (so the xfer processor will be a fix).
BUG=skia:
Review URL: https://codereview.chromium.org/689923004
This will be a bit hairy to review.
The FillBounds and CollectLayers code has diverged significantly resulting in the rendering path seeing different bounds than the hoisting path. This CL merges the FillBounds changes into CollectLayers. A follow on CL will, hopefully, find a way to layer CollectLayers on top of FillBounds.
The only code in CollectLayers that is different from FillBounds is bracketed by "LAYER HOISTING" comments.
NOTREECHECKS=true
Review URL: https://codereview.chromium.org/685263004
Got a few crashes running the fuzzer locally, all related to handling NULL members/parameters in an inconsistent way.
BUG=skia:
Review URL: https://codereview.chromium.org/675013003
This assists debugging layer hoisting errors (e.g., when a layer is hoisted but not actually used in subsequent rendering).
Review URL: https://codereview.chromium.org/694533004
A picture may possess many layers that get placed in one plot of the atlas. In this case we can only remove the plot from the plotUsage tracking structure when all the layers belonging to the picture in that plot have been removed.
Review URL: https://codereview.chromium.org/654463004
Adds miplevel as part of dfpath key, and scale factor so we know
how much to adjust to fit desired scale.
BUG=skia:2935
Review URL: https://codereview.chromium.org/687283002
This should produce tighter conservative bounding boxes for text than the
approximation code it replaces.
Recording performance is neutral on my desktop. Playback performance
improves by up to 15% on text heavy pages, e.g.
desk_pokemonwiki.skp_1 3.24ms -> 2.83ms 0.87x
desk_baidu.skp_1 1.91ms -> 1.58ms 0.83x
Committed: https://skia.googlesource.com/skia/+/bf8dc343df4fbdcb8af546eb68b640e011a33489
CQ_EXTRA_TRYBOTS=client.skia:Test-Win7-ShuttleA-HD2000-x86-Debug-Trybot
Review URL: https://codereview.chromium.org/680363003
The SkRectShaderImageFilter had the same bug as previously fixed for
SkBitmapSource and SkPictureImageFilter. Rather than copy-and-paste
the implementation, this change makes all filters with 0 inputs return
their source bounds, instead of returning false.
BUG=427251
Review URL: https://codereview.chromium.org/681643003
This should produce tighter conservative bounding boxes for text than the
approximation code it replaces.
Recording performance is neutral on my desktop. Playback performance
improves by up to 15% on text heavy pages, e.g.
desk_pokemonwiki.skp_1 3.24ms -> 2.83ms 0.87x
desk_baidu.skp_1 1.91ms -> 1.58ms 0.83x
Review URL: https://codereview.chromium.org/680363003
This is intended to prevent ghosting on tiled architectures.
This CL also defers creation of the atlas (and its texture) until it is actually needed.
Review URL: https://codereview.chromium.org/678403002
Adds the following:
- Use cached geometry processor rather than recreating all the time.
- Use context's quad index buffer.
Review URL: https://codereview.chromium.org/683923002
Reason for revert:
Compile errors on bots
Original issue's description:
> These tests stress pathops by describing the union of circle-like paths that have tiny line segments embedded and double back to create near-coincident conditions.
>
> The fixes include
> - detect when finding the active top loops between two possible answers
> - preflight chasing winding to ensure answer is consistent
> - binary search more often when quadratic intersection fails
> - add more failure paths when an intersect is missed
>
> While this fixes the chrome bug, reenabling path ops in svg should be deferred until additional fixes are landed.
>
> TBR=
> BUG=421132
>
> Committed: https://skia.googlesource.com/skia/+/6f726addf3178b01949bb389ef83cf14a1d7b6b2TBR=caryclark@google.com
NOTREECHECKS=true
NOTRY=true
BUG=421132
Review URL: https://codereview.chromium.org/686843002
The fixes include
- detect when finding the active top loops between two possible answers
- preflight chasing winding to ensure answer is consistent
- binary search more often when quadratic intersection fails
- add more failure paths when an intersect is missed
While this fixes the chrome bug, reenabling path ops in svg should be deferred until additional fixes are landed.
TBR=
BUG=421132
Review URL: https://codereview.chromium.org/633393002
This CL alters layer hoisting to defer creation of the free floating layers until they are actually needed (rather than creating _all_ the hoisted layers at the start).
It also fixes a pre vs. post Concat bug with how matrices were being accumulated.
BUG=skia:2315
Review URL: https://codereview.chromium.org/657383004
Reason for revert:
try again
Original issue's description:
> Fix bounds computation of all 0-input filters.
>
> The SkRectShaderImageFilter had the same bug as previously fixed for
> SkBitmapSource and SkPictureImageFilter. Rather than copy-and-paste
> the implementation, this change makes all filters with 0 inputs return
> their source bounds, instead of returning false.
>
> BUG=427251
>
> Committed: https://skia.googlesource.com/skia/+/ba036cc82b5a543a13cafd11a19ba0e3087fca38TBR=bsalomon@google.com,senorblanco@chromium.org
NOTREECHECKS=true
NOTRY=true
BUG=427251
Review URL: https://codereview.chromium.org/678273002
The SkRectShaderImageFilter had the same bug as previously fixed for
SkBitmapSource and SkPictureImageFilter. Rather than copy-and-paste
the implementation, this change makes all filters with 0 inputs return
their source bounds, instead of returning false.
BUG=427251
Review URL: https://codereview.chromium.org/681643003
- The expected case is now a single bulk-load insert() call instead of N;
- reserve() and flushDeferredInserts() can fold into insert() now;
- SkBBH subclasses may take ownership of the bounds
This appears to be a performance no-op on both my Mac and N5. I guess
even the simplest indirect branch predictor ("same as last time") can predict
the repeated virtual calls to SkBBH::insert() perfectly.
BUG=skia:
Review URL: https://codereview.chromium.org/670213002
This only affects the PNG image decoder, where we have already created
the color table without premultiplication. Since the RowProcChooser is
just providing a proc that chooses indices into the color table, it can
just return the same RowProc.
Update test_row_proc_choice. It was testing to ensure that we hadn't
changed the behavior from the original version of setPrefConfigTable.
In this case, we deliberately changed the behavior, so we need to
change the test.
BUG=b/12024301
Review URL: https://codereview.chromium.org/657863005
