This is moving to the world where the SkPipelineDataGatherer holds
a UniformManager. In that world the UniformManager will need to
incrementally allocate the uniforms it is collecting.
Bug: skia:12701
Change-Id: I6ed26948d7d0f2979130c4908d0c34c32604cd75
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527736
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Bug: skia:12698
Change-Id: I3744b281a3d22bde5adeea4f909ca65e4e4d3cfe
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527279
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
With UniformData being stored in a arena and TextureData still being
held by unique pointer, the PipelineDataCache needs to be able to handle
both.
Longer term, TextureData will also be held in an arena and this
split will go away.
Bug: skia:12701
Change-Id: Iab6c6542fbce4f886410f985e6c93ddaa125152e
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527504
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
`writeStatement` has the potential to emit a trace_line op. In practice,
right now our function's outer Block never has a position, so there is
no difference. However, Ethan's position-related work will add a
position to this Block, which would expose this issue and break tests.
Change-Id: I810edbc9b82feafd471cbcacf062aa14c8a66ba8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527841
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Ethan Nicholas <ethannicholas@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Introduce a new WGSL codegen backend. It currently only supports
fragment/vertex stage input and output variable declarations.
Bug: skia:13092
Change-Id: If9e4e432794eb95fd33719be5c0141dfc2cb5680
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/526757
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Arman Uguray <armansito@google.com>
No functional changes; just a code cleanup now that this is possible.
Change-Id: I2d153e22058c2b9be8e1dc4f2ad9ccc186bff120
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527657
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
It was already a no-op.
Bug: skia:13078
Change-Id: I7c7e1a86650ebdbd79bfcc4c1e739a3901345e76
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527836
Auto-Submit: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
We already performed a matching simplification for `0 - x`, so this
seemed like a straightforward improvement.
Performing this simplification causes the expressions in the test code
to match on both sides (e.g. `-one == -one`) which allows them to fold
away.
Change-Id: Idf87a98024dd6831b45d0384285ead2e2e039493
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527656
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
We use multiplication by 1 or -1 to branchlessly choose one of `min` and
`max` in the same function.
Change-Id: I44cf747feeae75a9c3e00f36e112e0a429871e86
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527596
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Change-Id: Id8d943b3eb752bad7f45c852c832c1f1fa0bdaca
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527776
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Hard-light is just overlay with the parameters reversed.
Change-Id: I6cf5963b1252cba3a7b71a56f4094a070188f8b2
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527503
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
These functions were functionally almost identical, except:
- Sometimes sda/dsa are flipped
- Sometimes the saturation is not updated
We now have one method (blend_hslc) which can do all four blend
operations. It takes two new parameters ("flip" and "saturate") to
handle these four variations.
This reduces our shader count on some of our most shader-heavy slides
(e.g. aaxfermodes, xfermodeimagefilter) at a pretty reasonable cost.
Change-Id: Ifa8a48399851a9badb5d50038de1e25e60d44ebd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527281
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This will allow expressions like `-x == -x` or `!y == !y` to be detected
as matching expressions (which enables various constant-folding paths).
(Also, migrated the analysis code into a separate cpp.)
Change-Id: I3e317fdaed3762f8fa19e684a5ed557fc9348c7c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527617
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
The clip isn't actually applied, but clip stack manipulation is wired
up. This may have a slight impact on how things draw because the device
clip bounds now reflect the clip actions, and that is currently used as
each draw's scissor.
Bug: skia:12698
Change-Id: I1d2e9cea49762dbfaa008c7c6af6034387b62519
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/526917
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
This avoids a situation where we didn't know the position of a binary
expression while we were in the middle of evaluating it, which was
leading to problems with upcoming error reporting changes.
Change-Id: Ie41fa82d077de1aa1041cd84c539ba29aaa2f5f0
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527500
Reviewed-by: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
This avoids a situation where we didn't know the position of a postfix
expression while we were in the middle of evaluating it, which was
leading to problems with upcoming error reporting changes.
Change-Id: Ie6d4dd8411ffb24ac7ee95d3e908483d0d92ba0f
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527499
Reviewed-by: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
This change does the following:
* Removes 2 Golo Test-TSAN_Vulkan jobs because their counterparts already exist in Skolo.
* Moves 1 Golo Test-TSAN_Vulkan job to Skolo.
Bug: skia:13129
Change-Id: I28b40320c831d109aa4f8778ddc1d03be9473592
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/526916
Reviewed-by: Kevin Lubick <kjlubick@google.com>
Commit-Queue: Ravi Mistry <rmistry@google.com>
This adds egl through swiftshader to unix builds as a separate build target, supporting cl in g3
Change-Id: Ied8c413e9dee0f78f25cbcec1bf937a5eb55dd60
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/526676
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Jorge Betancourt <jmbetancourt@google.com>
Followup CLs will improve their output.
Change-Id: I07059348f68cd6cd3154c31a41f81018b26a44e5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527616
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
Previously, we checked for division against zero, but didn't do anything
to prevent division against extraordinarily small values. Now, we only
saturate if the delta between max and min is greater than 0.00001.
Change-Id: I7d1df3430941c7e1a7f94e597d5449f9259612d6
Bug: skia:9320
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527498
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>
Change-Id: Ibff49d1928d7f82d04930c8cfd9d574780732c0d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527497
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
We were not propagating the position into a double-negated expression,
leading to an assertion failure in PrefixExpression.
Change-Id: I1970ff1a06d9631582626c68e151f12f6b3ef278
Bug: oss-fuzz:46381
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527507
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: Ethan Nicholas <ethannicholas@google.com>
GrProcessors should use BlendExpression instead. In practice, this is
currently a wash for GrXferProcessor generation, because anything that
BlendExpression can optimize right now is already be handled by the
PorterDuffXferProcessor. But it does allow us to remove AppendMode.
Also, any future improvements to BlendExpression will be accessible by
CustomXfermode.
Change-Id: Iaad82e4fe62ba5326b69020f254be9d2468813a5
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527276
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
This gm is using ToolUtils::emoji_typeface() to draw, but hardcoding the
result of ToolUtils::emoji_sample_text() to remove the space. Remove
the hardcoded string and use the emoji_sample_text instead. Adjust the
sample text by using only the first two non-space codepoints.
Change-Id: Ifae79b3f73951a2b6216d838b1fe0a63608bd678
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527067
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
Auto-Submit: Ben Wagner <bungeman@google.com>
The goal here is to replace GrGLSLBlend::AppendMode with a version that
supports blend_porter_duff (i.e., uniform handling must be supported).
AppendMode is removed in the followup CL.
Change-Id: I166232b1238cef57af53d00246e7fee19d1ea156
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/526516
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
This also adds a make target to build a release version of
the debugger, which is handy when we want to push a fix
quickly. The make targets will also create the local_build
folder in the infra repo if it does not exist.
Change-Id: I073615a35068c425608454311a51c0c4e0a8152a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527181
Reviewed-by: Brian Osman <brianosman@google.com>
This FreeType union is 80 bytes in size and sometimes two of them are
being passed by value. Pass them by const reference instead.
Change-Id: I69f2a326e586c1ddf80d79db96f3df7ed2d00e31
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527285
Commit-Queue: Ben Wagner <bungeman@google.com>
Auto-Submit: Ben Wagner <bungeman@google.com>
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Herb Derby <herb@google.com>
Once the UniformManager is moved into the SkPipelineDataGatherer, it
will need to be reset-able and write{Uniforms} will need to be able
to be called repeatedly. This CL starts setting up for that future.
Bug: skia:12701
Change-Id: Ia652c05b4eb508dce12c66b2298930d799d1e445
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527280
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
Bug: skia:13036
Change-Id: I31ccc3d1cfb271a2c2000476b91a695390cef4a6
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/526518
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
The internal ClipGeometry enum is moved to SimplifyResult in header.
The internal get_clip_geometry is renamed to Simplify and declared
header. This was so that the internal types (SaveRecord, Element, etc)
and TransformedShape could easily interoperate.
TransformedShape is a view that bundles references together instead
of requiring that all the internal types have some magical set of
functions that get_clip_geometry used.
The contains and intersect logic has all been consolidated into
TransformedShape instead of spread across many different contains()
functions. Intersect() is also more detailed because it can take
coordinate systems into account the coordinate systems, which will be
particularly useful when draw checking is added in a follow-up patch.
Bug: skia:12698
Change-Id: Ifac41a40715c7b0369fe31c273f12edc1efa956a
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/526736
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Michael Ludwig <michaelludwig@google.com>
The SkPipelineDataGatherer is going to subsume all the UniformManagers
scattered across the code so it will need the Layout in its ctor and
be plumbed where ever there is a UniformManager (in order to replace it).
Bug: skia:12701
Change-Id: I4f42bf50023e9e66c90f9a14833b976e214e1cc1
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/526275
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
The unused variable was breaking compilation for some versions of clang
Bug: skia:12701
Change-Id: I8ded5bf3d18136b5679878090186746b88292c3d
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527236
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: John Stiles <johnstiles@google.com>
Commit-Queue: John Stiles <johnstiles@google.com>
We didn't have any existing tests that exercised this path; it is
separate from most operators since it has no C++ equivalent.
Change-Id: I95b538dad01f8c8b122954fb5f66337371a398a8
Bug: oss-fuzz:46289
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527196
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
This is currently unused and can be re-obtained from the paint params
key if needed. Currently, it is getting in the way of updating how we
collect uniform data.
Bug: skia:12701
Change-Id: I5057cdfa42252b03b0f9756290735bf4b20dcb6b
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/526272
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
This fixes an issue with "go embed" not liking symlinks.
Bug: skia:13126
Change-Id: I994751e63ec467e37555f5bfc192243498b0d6cd
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/527057
Reviewed-by: Ravi Mistry <rmistry@google.com>
Commit-Queue: Eric Boren <borenet@google.com>
The `owner` field was previously restricted to GrFragmentProcessor,
which excludes other GrProcessor classes like GrXferProcessors from
using the uniform handlers.
Change-Id: I465190334f0fb1e27393108ab7b18d2d3089e69c
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/526461
Commit-Queue: John Stiles <johnstiles@google.com>
Auto-Submit: John Stiles <johnstiles@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
