The SkStrikeServer maintains a map of SkGlyphCacheState to track the
cache for a strike on the client. This entry is evicted only on a
failure to lock, which means that the map can potentially grow unbounded
if entries are not reused after eviction.
Make sure we check that they are deleted after some limit.
R=herb@google.com
Bug:878966
Change-Id: I4adb06c35661049328f6e0bde52fb1c774d0c29b
Reviewed-on: https://skia-review.googlesource.com/150443
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Khusal Sagar <khushalsagar@chromium.org>
This was accidently removed in cl/149048
BUG=chromium:877611
Change-Id: I9608b45848c27fc79e0f26e40bb9a1c4deb50f5b
Reviewed-on: https://skia-review.googlesource.com/150148
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Khusal Sagar <khushalsagar@chromium.org>
Subtle differences like location in an atlas can cause LSB differences
in path rendering, depending on FP precision.
Bug: skia:
Change-Id: I1e6fef7eaeef833db767d9c4b6c9d29dca2fee73
Reviewed-on: https://skia-review.googlesource.com/150212
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Change-Id: Id51062e9b6231ae9eeabdc418adda8cc3974dcc3
Reviewed-on: https://skia-review.googlesource.com/149041
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Testing using serialization PaintOpPerfTest, this had the following
improvement in serialization speed:
Before: ~515,000 ops/s
After: ~600,000 ops/s
R=herb@google.com
Change-Id: Iefd80467ea4ff7cc88e8ca1f431883502d249857
Reviewed-on: https://skia-review.googlesource.com/142972
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Khusal Sagar <khushalsagar@chromium.org>
This reverts commit e2e52e46ca.
Reason for revert: See if this is blocking the Chrome roll
Original change's description:
> Remove drawTextBlob from device use drawGlyphRunList
>
> Convert all backends to use GlyphRunList instead of
> text blobs. If the device did not originally implement
> drawTextBlob it will be simulated by drawPosText on the
> device.
>
> Other changes:
> Change to using an origin from absolulte positioning. The GPU
> code uses origin change to update blobs under translation.
>
> Change cluster to use const uint32_t instead of just
> uint32_t.
>
> Add SkPaint to runs.
>
> The draw filter is hosted up to the canavas level and applied there.
>
> Change-Id: Ib105b6bd26b67db55f1c954e37c79fbdcaa9d4a2
> Reviewed-on: https://skia-review.googlesource.com/137224
> Reviewed-by: Herb Derby <herb@google.com>
> Reviewed-by: Khusal Sagar <khushalsagar@chromium.org>
> Reviewed-by: Hal Canary <halcanary@google.com>
> Reviewed-by: Jim Van Verth <jvanverth@google.com>
> Commit-Queue: Herb Derby <herb@google.com>
TBR=jvanverth@google.com,halcanary@google.com,bungeman@google.com,herb@google.com,reed@google.com,khushalsagar@chromium.org,khushalsagar@google.com
Change-Id: I4d93a534990c89deee7d3aaa00ec40d47e0d2ece
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/138120
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Convert all backends to use GlyphRunList instead of
text blobs. If the device did not originally implement
drawTextBlob it will be simulated by drawPosText on the
device.
Other changes:
Change to using an origin from absolulte positioning. The GPU
code uses origin change to update blobs under translation.
Change cluster to use const uint32_t instead of just
uint32_t.
Add SkPaint to runs.
The draw filter is hosted up to the canavas level and applied there.
Change-Id: Ib105b6bd26b67db55f1c954e37c79fbdcaa9d4a2
Reviewed-on: https://skia-review.googlesource.com/137224
Reviewed-by: Herb Derby <herb@google.com>
Reviewed-by: Khusal Sagar <khushalsagar@chromium.org>
Reviewed-by: Hal Canary <halcanary@google.com>
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Herb Derby <herb@google.com>
The remote glyph cache tests assume that the strike cache will not
change during a test. This is not true because other test also
mutate the strike cache. This is causing flaky tests.
BUG=skia:8091
Change-Id: I397d411f9412006715f6860941dfb05ad54ae1b6
Reviewed-on: https://skia-review.googlesource.com/136741
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Commit-Queue: Herb Derby <herb@google.com>
Change Validate to ValidateGlyphCacheDataSize. Make it call a
non-global version.
BUG=skia:8091
Change-Id: Iec31a06569a0ab1ec318e693e699a808eb9ad247
Reviewed-on: https://skia-review.googlesource.com/136638
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
This reverts commit 3e36ce6e66.
Reason for revert: Too much verification
Original change's description:
> Make SkRemoteGlyphCache tests use private glyph cache
>
> Change-Id: If6aa189f3badc7558ab8ecf71ee3d704b275b20f
> Reviewed-on: https://skia-review.googlesource.com/136225
> Commit-Queue: Herb Derby <herb@google.com>
> Reviewed-by: Mike Klein <mtklein@google.com>
TBR=mtklein@google.com,herb@google.com
Change-Id: Ia0c096abd0ab651bc7c907f0595af4f07c88fb5e
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/136478
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Herb Derby <herb@google.com>
Change-Id: If6aa189f3badc7558ab8ecf71ee3d704b275b20f
Reviewed-on: https://skia-review.googlesource.com/136225
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
If the client finds that the server re-initializes a cached path/image,
we consider this an error and invalid data. But since we might
initialize a glyph using a fallback on the client, and receive the
correct version from the server later, this is not longer true.
Bug: 829622
Change-Id: I34ab17b54139d89a15179265d4aed4a1fe36fd47
Reviewed-on: https://skia-review.googlesource.com/133566
Commit-Queue: Khusal Sagar <khushalsagar@chromium.org>
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
When allocating the mask for a fallback glyph, we allocate it on the
arena on the SkScalerContext while the image belongs to a glyph on a
different cache. This can lead to use-after-free bugs if accessing the
image after the context owning that memory is destroyed. Fix this by
allocating on the arena from the owning cache.
R=herb@google.com, mtklein@google.com
Bug: 829622
Change-Id: Ife53e24f5bc868f36c43f2adcd7a2629ab5577fe
Reviewed-on: https://skia-review.googlesource.com/134182
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Change-Id: I0ae768c5517c3ee3f6822fea0926b3f27214a0e4
Reviewed-on: https://skia-review.googlesource.com/132260
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
Use GrContext::FallbackTextHelper in SkTextBlobCacheDiffCanvas to
replicate glyph generation logic for fallback text during analysis. This
ensures that we correctly handle these fallback cases when using
distance field or paths for text rendering.
R=herb@google.com, jvanverth@google.com
Bug: skia:7913
Change-Id: I3067c4f1bd09231a564ac7c4cd89efcb876d2abd
Reviewed-on: https://skia-review.googlesource.com/132285
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Khusal Sagar <khushalsagar@chromium.org>
Most of this is (obviously) not necessary to do, but once
I started, I figured I'd just get it all. Tools (nanobench,
DM, skiaserve), all GMs, benches, and unit tests, plus support
code (command line parsing and config stuff).
This is almost entirely mechanical.
Bug: skia:
Change-Id: I209500f8df8c5bd43f8298ff26440d1c4d7425fb
Reviewed-on: https://skia-review.googlesource.com/131153
Reviewed-by: Mike Klein <mtklein@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This reverts commit 4f078f7cfa.
Reason for revert: Seems to have caused these breakages:
* https://chromium-swarm.appspot.com/task?id=3dc6788dc7b77010&refresh=10
* https://chromium-swarm.appspot.com/task?id=3dc67452db797c10&refresh=10
Original change's description:
> fonts: Use correct SurfaceProps in analysis canvas for remoting.
>
> SaveLayer may not preserve the behaviour for lcd text, in which case
> the surfaceProps used for desc generation are inconsistent.
>
> R=herb@google.com
>
> Bug: 829622
> Change-Id: I3adfc6780f26e4eb333a8aff76eaa4b5a9f0a0a7
> Reviewed-on: https://skia-review.googlesource.com/129557
> Reviewed-by: Herb Derby <herb@google.com>
> Commit-Queue: Khusal Sagar <khushalsagar@chromium.org>
TBR=herb@google.com,khushalsagar@chromium.org
Change-Id: I0372a3d834f8c4c929feb71c2b6b13739443eaaa
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: 829622
Reviewed-on: https://skia-review.googlesource.com/130680
Reviewed-by: Ravi Mistry <rmistry@google.com>
Commit-Queue: Ravi Mistry <rmistry@google.com>
SaveLayer may not preserve the behaviour for lcd text, in which case
the surfaceProps used for desc generation are inconsistent.
R=herb@google.com
Bug: 829622
Change-Id: I3adfc6780f26e4eb333a8aff76eaa4b5a9f0a0a7
Reviewed-on: https://skia-review.googlesource.com/129557
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Khusal Sagar <khushalsagar@chromium.org>
Change-Id: I309b39425afc9b45095241eeb299096bc426afed
Reviewed-on: https://skia-review.googlesource.com/130029
Reviewed-by: Herb Derby <herb@google.com>
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Herb Derby <herb@google.com>
This reverts commit 35e0a1a690.
Reason for revert: looks like we're leaking paths?
https://chromium-swarm.appspot.com/task?id=3da25e2f0cadb210&refresh=10
Original change's description:
> fonts: Add support for distance field text to font remoting.
>
> R=jvanverth@google.com, herb@google.com
>
> Bug: skia:7913
> Change-Id: Id3f5b3e75005be9a7234df774268359b406c99a8
> Reviewed-on: https://skia-review.googlesource.com/128970
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Reviewed-by: Jim Van Verth <jvanverth@google.com>
> Commit-Queue: Khusal Sagar <khushalsagar@chromium.org>
TBR=jvanverth@google.com,bsalomon@google.com,bungeman@google.com,herb@google.com,khushalsagar@chromium.org
Change-Id: I37c54c8748db9b20e1f48016d8298808a1999fdb
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:7913
Reviewed-on: https://skia-review.googlesource.com/129681
Commit-Queue: Mike Klein <mtklein@google.com>
Reviewed-by: Mike Klein <mtklein@google.com>
This reverts commit 2b8a0d1844.
Reason for revert: reverting a CL this builds on
Original change's description:
> fonts: Cleanup cache miss logging for font remoting.
>
> Add hooks to notify the embedder if there is a cache miss during draw.
> Also remove the reference to SkStrikeClient from SkTypefaceProxy and
> SkScalerContextProxy, since the proxies can outlive the client.
>
> R=herb@google.com
>
> Bug: 829622
> Change-Id: Ib2fd1b91ebd057856c1d4e717cf50b49f08c903b
> Reviewed-on: https://skia-review.googlesource.com/129402
> Commit-Queue: Khusal Sagar <khushalsagar@chromium.org>
> Reviewed-by: Herb Derby <herb@google.com>
TBR=herb@google.com,khushalsagar@chromium.org
Change-Id: I8a331545988885c620685008f4b60240d80f3712
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: 829622
Reviewed-on: https://skia-review.googlesource.com/129682
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Add hooks to notify the embedder if there is a cache miss during draw.
Also remove the reference to SkStrikeClient from SkTypefaceProxy and
SkScalerContextProxy, since the proxies can outlive the client.
R=herb@google.com
Bug: 829622
Change-Id: Ib2fd1b91ebd057856c1d4e717cf50b49f08c903b
Reviewed-on: https://skia-review.googlesource.com/129402
Commit-Queue: Khusal Sagar <khushalsagar@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>
SkRemoteGlyphCache only sends images for glyphs, even for cases where
the gpu falls back to drawing text as paths. This includes cases in
SkDraw::ShouldDrawTextAsPaths and when the glyph exceeds the max bounds
that can fit on the atlas. Fix this by identifying these cases in the
renderer and sending paths instead.
Note: We still don't handle distance field text correctly.
R=herb@google.com, bsalomon@google.com
Bug: skia:7913
Change-Id: I17d4eccbeaa2e995ae67b61c76cebd27f8280329
Reviewed-on: https://skia-review.googlesource.com/128203
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Khusal Sagar <khushalsagar@chromium.org>
When deserializing glyphs in the SkRemoteGlyphCache, we allocate from
the arena for the SkGlyphCache but don't account for it in the total
memory used by the cache. Fix that and avoid exposing the SkArenaAlloc
from SkGlyphCache, since that can result in such brittle use.
R=herb@google.com
Bug: 829622
Change-Id: Iecff9ce6e0ed2c641957535363edec3e3fad178d
Reviewed-on: https://skia-review.googlesource.com/128112
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Khusal Sagar <khushalsagar@chromium.org>
This reverts commit 101d56359a.
Reason for revert: 5 of 5
Original change's description:
> fonts: Set up remote glyph caching to push fonts.
>
> Currently the SkStrikeClient is designed to pull fonts from the server
> on demand, and to pre-fetch a batched request by analyzing the ops using
> a SkTextBlobCacheDiffCanvas. This change modifies the design to support
> a push based model, where the server pushes fonts required by the client
> and sets up the requisite SkGlyphCaches on the client prior to
> rasterizing the ops.
>
> This model still relies on the SkTextBlobCacheDiffCanvas for analyzing
> the glyphs required for rasterizing an op. The glyph caches required for
> raster are locked and missing glyphs to be sent to the client are tracked
> by the SkStrikeServer. The embedder can serialize this font data at any
> point, but must ensure that this data is deserialized by the
> SkStrikeClient at the remote end, before rasterizing any ops analyzed
> prior to serialization. Any refs on the caches are released once the
> font data is serialized by the server.
>
> The locking of glyph caches relies on the embedder providing discardable
> handles. These handles can be created on the server and serialized to be
> sent to the client, and map to an instance of SkGlyphCache. This allows
> the server to control the lifetime of the caches on the client.
>
> Bug: skia:7515
> Change-Id: Id39f346b47b60899778404bbd0429ee811d0e53b
> Reviewed-on: https://skia-review.googlesource.com/120283
> Commit-Queue: Khusal Sagar <khushalsagar@chromium.org>
> Reviewed-by: Herb Derby <herb@google.com>
TBR=mtklein@google.com,herb@google.com,khushalsagar@chromium.org
Change-Id: If72caf968ddcbf70b8b9d71782a2339a118ed202
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:7515
Reviewed-on: https://skia-review.googlesource.com/125264
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
This reverts commit ef4142a9bc.
Reason for revert: 2 of 5
Original change's description:
> fonts: Bandaid fix for desc mismatch in SkRemoteGlyphCache.
>
> Since the typeface proxies on the client don't perform the same
> filtering done on the server during SkDescriptor generation, it causes
> the desc mismatches during raster. Disable this filtering on the server
> until this is resolved.
>
> Bug: 831354
> Change-Id: I5683372fb497a4874dede5aec9c734cd1392872c
> Reviewed-on: https://skia-review.googlesource.com/125140
> Commit-Queue: Khusal Sagar <khushalsagar@chromium.org>
> Reviewed-by: Herb Derby <herb@google.com>
TBR=herb@google.com,khushalsagar@chromium.org
Change-Id: I8e732f57aa49323c186e3c4ea6120ff1caf8e25b
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: 831354
Reviewed-on: https://skia-review.googlesource.com/125261
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Mike Klein <mtklein@google.com>
Since the typeface proxies on the client don't perform the same
filtering done on the server during SkDescriptor generation, it causes
the desc mismatches during raster. Disable this filtering on the server
until this is resolved.
Bug: 831354
Change-Id: I5683372fb497a4874dede5aec9c734cd1392872c
Reviewed-on: https://skia-review.googlesource.com/125140
Commit-Queue: Khusal Sagar <khushalsagar@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>
Currently the SkStrikeClient is designed to pull fonts from the server
on demand, and to pre-fetch a batched request by analyzing the ops using
a SkTextBlobCacheDiffCanvas. This change modifies the design to support
a push based model, where the server pushes fonts required by the client
and sets up the requisite SkGlyphCaches on the client prior to
rasterizing the ops.
This model still relies on the SkTextBlobCacheDiffCanvas for analyzing
the glyphs required for rasterizing an op. The glyph caches required for
raster are locked and missing glyphs to be sent to the client are tracked
by the SkStrikeServer. The embedder can serialize this font data at any
point, but must ensure that this data is deserialized by the
SkStrikeClient at the remote end, before rasterizing any ops analyzed
prior to serialization. Any refs on the caches are released once the
font data is serialized by the server.
The locking of glyph caches relies on the embedder providing discardable
handles. These handles can be created on the server and serialized to be
sent to the client, and map to an instance of SkGlyphCache. This allows
the server to control the lifetime of the caches on the client.
Bug: skia:7515
Change-Id: Id39f346b47b60899778404bbd0429ee811d0e53b
Reviewed-on: https://skia-review.googlesource.com/120283
Commit-Queue: Khusal Sagar <khushalsagar@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>
