Make GrContext::MakeGL take interface as sk_sp.
Make GrContext::MakeVulkan take GrVkBackendContext as sk_sp.
Change-Id: I13c22a57bd281c51738f503d9ed3418d35a466df
Reviewed-on: https://skia-review.googlesource.com/81842
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
These APIs existed for Chrome. Chrome is no longer using them.
Change-Id: I15a5e2f88c7e8d1356188748fc68d4658f6f1849
Reviewed-on: https://skia-review.googlesource.com/81021
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Cary Clark <caryclark@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Bug: skia:
Change-Id: I054560b66c6cde346d939015326d8547879d2c4b
Reviewed-on: https://skia-review.googlesource.com/81160
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Reed <reed@google.com>
This reverts commit c382384848.
Reason for revert: need to add legacy flag to chrome first
Original change's description:
> keep SVG arcs axis aligned
>
> Computing the arc width introduces rounding errors that cause the
> arc to exceed 1/4 circle and cause integer anchored arcs to start
> outside their marks. A round rect may lose convexity as a result.
>
> Check if arcTo() inputs are integers and arc is 90 degrees;
> if so, output conics which are axis-aligned on integers as well.
>
> This is triggered when using SVG to represent a round rect.
>
> Possible future enhancements are recorded in bug.skia.org/7383
>
> R=reed@google.com,djsollen@google.com
> Change-Id: I6609456fcefabcda6c9560a044533ecb5cda2d31
> Reviewed-on: https://skia-review.googlesource.com/79423
> Reviewed-by: Derek Sollenberger <djsollen@google.com>
> Commit-Queue: Derek Sollenberger <djsollen@google.com>
TBR=djsollen@google.com,caryclark@google.com,reed@google.com,caryclark@skia.org
Change-Id: Ia503ea62def15322df2b11da30d377a9543d6999
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/81281
Reviewed-by: Cary Clark <caryclark@google.com>
Commit-Queue: Cary Clark <caryclark@google.com>
I was wondering how feasible using this to make downcasts safe would be.
These tests would need to build and pass on all our bots, at least.
Change-Id: I1753ba58841bf6c17d6ac3af7374518356e1bb05
Reviewed-on: https://skia-review.googlesource.com/81180
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Computing the arc width introduces rounding errors that cause the
arc to exceed 1/4 circle and cause integer anchored arcs to start
outside their marks. A round rect may lose convexity as a result.
Check if arcTo() inputs are integers and arc is 90 degrees;
if so, output conics which are axis-aligned on integers as well.
This is triggered when using SVG to represent a round rect.
Possible future enhancements are recorded in bug.skia.org/7383
R=reed@google.com,djsollen@google.com
Change-Id: I6609456fcefabcda6c9560a044533ecb5cda2d31
Reviewed-on: https://skia-review.googlesource.com/79423
Reviewed-by: Derek Sollenberger <djsollen@google.com>
Commit-Queue: Derek Sollenberger <djsollen@google.com>
Points cost 8 bytes and indices cost 4. If a point is accessed twice,
it's a wash whether we duplicate it or index. This change eliminates
texel buffers by duplicating points across instance arrays. This
reduces our dependence on extensions as well as getting rid of our
indirect memory access pattern in vertex shaders.
As a result of this change, memory usage by GPU buffers will only be a
fraction larger at worst, and slightly better at best.
Bug: skia:
Cq-Include-Trybots: skia.primary:Test-Debian9-Clang-GCE-CPU-AVX2-x86_64-Release-All-SKNX_NO_SIMD
Change-Id: I3c7f03772edd4f850d5fdd7b55552647335c1b52
Reviewed-on: https://skia-review.googlesource.com/79185
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
This partially reverts commit
1793e7bb46.
Hide SkEncodedInfo
Bug: skia:7353
Bug: skia:6839
This contains information that is not necessary for clients to know. The
Color enum tells the number of components in the input, but this is only
interesting internally (to the SkSwizzler).
Similarly, the Alpha enum differs from SkAlphaType in that it has
kBinary instead of kPremul. This is useful information only internally
for determining whether the SkColorSpaceXform needs to premultiply.
The bitsPerComponent is potentially useful for a client; Android (in
SkAndroidCodec) uses it to determine the SkColorType. Rather than
exposing bitsPerComponent, make SkAndroidCodec a friend so it can
access the SkEncodedInfo. A future change will change SkCodec to
recommend F16 for bitsPerComponent > 8, but that will be more involved;
it was the reason for the revert of this CL.
Switch conversionSupported to use an SkColorType, which is enough info.
Replace the SkEncodedInfo::Alpha field on SkCodec::FrameInfo with an
SkAlphaType.
SkCodec still needs an SkEncodedInfo, so move its header (which is
already not SK_API) to include/private.
TBR=mtklein@chromium.org,reed@google.com
Change-Id: I928b1f55317602cb37d29da63b53026c8d139cee
Reviewed-on: https://skia-review.googlesource.com/80860
Reviewed-by: Leon Scroggins <scroggo@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>
Also limit small radius bail in GrCircleEffect to clip out cases.
Change-Id: I14ce736969b05203219d68f30283c36c84f78f3a
Reviewed-on: https://skia-review.googlesource.com/80621
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This reverts commit c6f7a4ffa9.
Reason for revert: Causing differences in Gold, stemming from the fact that this changes the recommended SkImageInfo for 16 bits-per-component PNG from N32 to F16.
- an F16 bitmap already png-encodes to a 16 bits-per-component PNG, but it does not encode a linear colorspace (possibly a bug?). when we decode this PNG using getInfo(), it fails because it has an F16 color type and non-linear colorspace. (In the encode-srgb-png gm, this results in blank results for F16.) We could correct this on the encoder side, but it seems possible that a 16 bits-per-component PNG could be encoded with a different color space. In that case, we'd want SkCodec to recommend F16/SRGBLinear, but I think we'd want the SkCodec to store the encoded SkColorSpace so that we can Xform between the two. Currently SkCodec only stores one color space, so that will require a refactor.
- When decoding 16-bits-per-component PNGs, we are now decoding them to F16. This shows differences in Gold. The srgb/gpu results now look more like F16. I think this is fine.
Original change's description:
> Hide SkEncodedInfo
>
> Bug: skia:7353
> Bug: skia:6839
>
> This contains information that is not necessary for clients to know. The
> Color enum tells the number of components in the input, but this is only
> interesting internally (to the SkSwizzler).
>
> Similarly, the Alpha enum differs from SkAlphaType in that it has
> kBinary instead of kPremul. This is useful information only internally
> for determining whether the SkColorSpaceXform needs to premultiply.
>
> The bitsPerComponent is potentially useful for a client; Android (in
> SkAndroidCodec) uses it to determine the SkColorType. Rather than
> exposing bitsPerComponent, use it to make the same decision that Android
> would have made - 16 bits per component means to set the info to F16. Add
> a test that computeOutputColorType behaves as expected.
>
> Switch conversionSupported to use an SkColorType, which is enough info.
>
> Replace the SkEncodedInfo::Alpha field on SkCodec::FrameInfo with an
> SkAlphaType.
>
> SkCodec still needs an SkEncodedInfo, so move its header (which is
> already not SK_API) to include/private.
>
> Change-Id: Ie2cf11339bf999ebfd4390c0f448f7edd6feabda
> Reviewed-on: https://skia-review.googlesource.com/79260
> Reviewed-by: Mike Reed <reed@google.com>
> Reviewed-by: Mike Klein <mtklein@chromium.org>
> Commit-Queue: Leon Scroggins <scroggo@google.com>
TBR=mtklein@chromium.org,scroggo@google.com,reed@google.com
Change-Id: I0c5dd1461e1b70d1e55349a8e7ee6b029c3f556e
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:7353, skia:6839
Reviewed-on: https://skia-review.googlesource.com/80660
Reviewed-by: Leon Scroggins <scroggo@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>
Bug: skia:7353
Bug: skia:6839
This contains information that is not necessary for clients to know. The
Color enum tells the number of components in the input, but this is only
interesting internally (to the SkSwizzler).
Similarly, the Alpha enum differs from SkAlphaType in that it has
kBinary instead of kPremul. This is useful information only internally
for determining whether the SkColorSpaceXform needs to premultiply.
The bitsPerComponent is potentially useful for a client; Android (in
SkAndroidCodec) uses it to determine the SkColorType. Rather than
exposing bitsPerComponent, use it to make the same decision that Android
would have made - 16 bits per component means to set the info to F16. Add
a test that computeOutputColorType behaves as expected.
Switch conversionSupported to use an SkColorType, which is enough info.
Replace the SkEncodedInfo::Alpha field on SkCodec::FrameInfo with an
SkAlphaType.
SkCodec still needs an SkEncodedInfo, so move its header (which is
already not SK_API) to include/private.
Change-Id: Ie2cf11339bf999ebfd4390c0f448f7edd6feabda
Reviewed-on: https://skia-review.googlesource.com/79260
Reviewed-by: Mike Reed <reed@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Leon Scroggins <scroggo@google.com>
shader.
This fixes a bug on some intel devices where we are failing
in the ProcessorOptimizationTest.
I've tried other "no op" type things between the floor call and abs which
also fixed the issue, as well as adding explicit checks to see if we are
less than -1 or greater than 1 where the clamp is. Thus the clamp itself
should be a no op and shouldn't secretly be fixing the problem outside
of forcing the floor and abs lines to be separate.
Bug: skia:
Change-Id: I85bf82e0e02607b78470b7a5f8f918e9f53f0154
Reviewed-on: https://skia-review.googlesource.com/76820
Reviewed-by: Robert Phillips <robertphillips@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Bug: skia:
Change-Id: I0377e6a1dd8259e944f7902a5c68af524fa588c7
Reviewed-on: https://skia-review.googlesource.com/79382
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
If a quad or cubic reverses on itself, path ops breaks
it in two. It determines the type of curve remaining,
but needs to replace near-zero with zero first.
TBR=reed@google.com
Bug:790731
Change-Id: I3a1afa14fff064ca874b5abc768ec1ec5c2cf22f
Reviewed-on: https://skia-review.googlesource.com/79400
Commit-Queue: Cary Clark <caryclark@google.com>
Reviewed-by: Cary Clark <caryclark@skia.org>
and test it.
Change-Id: Ib0c2cf93c63d8d3c36a7d4d60bbec4ecede29bc7
Reviewed-on: https://skia-review.googlesource.com/78480
Reviewed-by: Chris Dalton <csmartdalton@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Adds ultra-deferred proxies that are instantiated by a user-supplied
callback during flush.
Bug: skia:7190
Change-Id: I75a7ac6dba953c3b0a99febc203a7f4d2f3789fc
Reviewed-on: https://skia-review.googlesource.com/76461
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
The current code for this filter premuls, then clamps. We should be
clamping, then premulling.
If the matrix makes alpha greater than 1, these two orderings can result
in different color values. Alpha will clamp to 1 either way, but the
color channels are multiplied by that >1 alpha in one case, and by =1
in the other.
The left column of the gm imagefilterscropexpand demonstrates this.
Its matrix adds 32/255 to alpha and 255/255 to green. This produces
alpha ~= 1.12. That's then multiplied by the relatively small red
and blue values in the grey checkerboard, resulting in different
in-range values than the ones we would have gotten if we clamped
alpha first. Green wasn't affected because it was already fully
saturated. 255 * 1.12 == 255 no matter when we clamp.
Change-Id: I4b30bf64c30fe62526674ad5f32e9ca19ec84714
Reviewed-on: https://skia-review.googlesource.com/77902
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
Sk3DShaderContext creates its nested shader context on a SkArenaAlloc,
which handles destruction when going out of scope.
Hence, the explicit context dtor call in ~Sk3DShaderContext() is
incorrect (likely left over from before SkArenaAlloc).
BUG=chromium:787712
Change-Id: I176222e449151dcce532a839ef9587d06f61d297
Reviewed-on: https://skia-review.googlesource.com/77203
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Herb Derby <herb@google.com>
Transforms a path's points into a local buffer up front, rather than
transforming as we parse. This hopefully gets better vector
performance as well as allowing us to skip the transformation step
for paths that are known to be in device space already.
Introduces a test for parsing empty paths and does general cleanup.
Bug: skia:7190
Change-Id: Ib86d2ffcdef6fa3ec66f6d9ad4b10c0b6d44c0dc
Reviewed-on: https://skia-review.googlesource.com/74621
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
This was uncovered by the linked fuzzer issue.
I haven't looked hard at it, but I'd guess it's fuzzed an ICC profile
into one that can't be deserialized, and we get a null in CreateProc().
We could probably restrict the null check to just CreateProc(), but
putting it in Make() and asserting in the constructor feels cozy.
BUG=chromium:787718
Change-Id: Ic4b1dad28c00ee5870f22093eedbf34686c32120
Reviewed-on: https://skia-review.googlesource.com/76080
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Brian Osman <brianosman@google.com>
Simplify the SkGaussFilter API to facilitate using
ranged-for loops.
Change-Id: Id853bd6bfe342ae95b7c6248c459fbf865f75d1e
Reviewed-on: https://skia-review.googlesource.com/73262
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
Commit-Queue: Herb Derby <herb@google.com>
This reverts commit 1e09e461d2.
Change-Id: I95d5544a7baaa078536790493ce4119816a77e94
Reviewed-on: https://skia-review.googlesource.com/72903
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This reverts commit 5627d65146.
Reason for revert: Google3
Original change's description:
> Add method to sk_gpu_test::TestContext to automatically restore the previous context.
>
> The motivation for this is to allow a GM to create a GL context, do some some work in it, and then return to the context that was set when it was invoked.
>
> Change-Id: Ie8496072a10f8f3ff36a08889e593a6ca961b61a
> Reviewed-on: https://skia-review.googlesource.com/70720
> Commit-Queue: Brian Salomon <bsalomon@google.com>
> Reviewed-by: Brian Osman <brianosman@google.com>
TBR=bsalomon@google.com,brianosman@google.com
Change-Id: Ifb79638c9d4500ca3be9a5be39a5ad78b20247c1
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://skia-review.googlesource.com/72981
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This is a reland of e421800227
Original change's description:
> Fix precision caps and rrect/ellipse effect precisions
>
> Replaces all the complex precision caps with a single flag that says
> whether "float" == fp32. Updates the ellipse and rrect effects to
> use float coords, and use the scale workaround when float != fp32.
>
> Bug: skia:7190
> Change-Id: Ieccff9f38acd05e5cec78fe90d01a5da901a9307
> Reviewed-on: https://skia-review.googlesource.com/70961
> Commit-Queue: Chris Dalton <csmartdalton@google.com>
> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
> Reviewed-by: Brian Salomon <bsalomon@google.com>
TBR=bsalomon@google.com
Bug: skia:7190
Change-Id: I7ced37a64164b83d86f6a957c35e10ce9085aba0
Reviewed-on: https://skia-review.googlesource.com/72760
Reviewed-by: Chris Dalton <csmartdalton@google.com>
Commit-Queue: Chris Dalton <csmartdalton@google.com>
The motivation for this is to allow a GM to create a GL context, do some some work in it, and then return to the context that was set when it was invoked.
Change-Id: Ie8496072a10f8f3ff36a08889e593a6ca961b61a
Reviewed-on: https://skia-review.googlesource.com/70720
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
This reverts commit e421800227.
Reason for revert: Also may be responsible for layout test failures? Playing it safe.
Original change's description:
> Fix precision caps and rrect/ellipse effect precisions
>
> Replaces all the complex precision caps with a single flag that says
> whether "float" == fp32. Updates the ellipse and rrect effects to
> use float coords, and use the scale workaround when float != fp32.
>
> Bug: skia:7190
> Change-Id: Ieccff9f38acd05e5cec78fe90d01a5da901a9307
> Reviewed-on: https://skia-review.googlesource.com/70961
> Commit-Queue: Chris Dalton <csmartdalton@google.com>
> Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
> Reviewed-by: Brian Salomon <bsalomon@google.com>
TBR=egdaniel@google.com,bsalomon@google.com,csmartdalton@google.com,ethannicholas@google.com
Change-Id: Idca2f0390e7a0eb85010255183f2f27332b8d26d
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: skia:7190
Reviewed-on: https://skia-review.googlesource.com/72540
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Replaces all the complex precision caps with a single flag that says
whether "float" == fp32. Updates the ellipse and rrect effects to
use float coords, and use the scale workaround when float != fp32.
Bug: skia:7190
Change-Id: Ieccff9f38acd05e5cec78fe90d01a5da901a9307
Reviewed-on: https://skia-review.googlesource.com/70961
Commit-Queue: Chris Dalton <csmartdalton@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This is to alleviate problems due to the command buffer getting bent out of shape when the current
OpenGL context is switched out from under it (because we ran a test with a native GL context). This,
however is not a full solution. More changes will be required to ensure that after running each
command buffer or native test we bind the null context. This does allow us to take a step in that
direction without breaking anything too badly. Moreover, there is no real benefit to reusing a
GrContextFactory.
Modifies DEF_GPUTEST to take GrContextOptions rather than a factory to use. Tests were already using
their own factories anyway.
In tests that use GrContextFactory the factory instance is moved to the inner loop.
Modifies gpucts and skia_test to not use persistent GrContextFactories.
Change-Id: Ie7a36793545c775f2f30653ead6fec93a3d22717
Reviewed-on: https://skia-review.googlesource.com/71861
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This patch fixes an assert that triggers when removing the last
extension in the list, since the index operator goes out of bounds.
Added a test that fails without the code changes and passes with the
changes.
R=bsalomon@chromium.org
Change-Id: I0d561b150899c178f638dde088af773fddf112d9
Reviewed-on: https://skia-review.googlesource.com/72000
Commit-Queue: Vladimir Levin <vmpstr@chromium.org>
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
This isn't used and has become a maintenance burden.
Change-Id: I5f3af8f91e5c4f073fe4ea30e0a7f1f61efeea47
Reviewed-on: https://skia-review.googlesource.com/70640
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>