Instead of starting recording blit commands immediately we can
record them closer to where we write to the Image directly. This makes
it a bit easier to move towards hiding the command buffers. And easier
to move to using implicitly generated barriers.

Bug: angleproject:2828
Change-Id: Ida7d4081dd41c8ee19c7dc3fb234df1bb065794d
Reviewed-on: https://chromium-review.googlesource.com/1254382
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Frank Henigman <fjhenigman@chromium.org>

This reverts commit 799da6d1.

Reason for revert:Crashes have slowed down but there were two changes around the same time that may have caused it.  Relanding this to see if things improve.

Original change's description:
> Revert "D3D11: Resolve multisampled EGL surfaces at eglSwapBuffers"
> 
> An increased crash rate in rx::NativeWindow11Win32::createSwapChain has been
> seen around the same time as this patch landed.  Speculatively reverting.
> 
> BUG=887403
> 
> This reverts commit a49f60fe.
> 
> Change-Id: Id1b08efbf47196f17f4a5e509a29f7d628b07b49
> Reviewed-on: https://chromium-review.googlesource.com/1238884
> Reviewed-by: Geoff Lang <geofflang@chromium.org>
> Commit-Queue: Geoff Lang <geofflang@chromium.org>

TBR=geofflang@chromium.org

# Not skipping CQ checks because original CL landed > 1 day ago.

Bug: 887403
Change-Id: Id862fbb1c0e764470f0b7f6bdb97bf6bab95d26b
Reviewed-on: https://chromium-review.googlesource.com/1251701Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Geoff Lang <geofflang@chromium.org>

There were two code paths using nearly identical code. Fix this by
generalizing the helper method slightly.

Refactoring change only.

Bug: angleproject:2828
Change-Id: I39caccfaad4880e566e995483b8e8fb46d67d3da
Reviewed-on: https://chromium-review.googlesource.com/1249563Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Frank Henigman <fjhenigman@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>

The cache is initialized from the application's blob cache and is
occasionally written back to it for disk storage.

Bug: angleproject:2516
Change-Id: I4cba4b00a7b9641c2983ef07159bc62cd10a5519
Reviewed-on: https://chromium-review.googlesource.com/1241373Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Yuly Novikov <ynovikov@google.com>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>

We had two separate minor bugs. 1) missing a flush of the image data
and 2) missing an initial barrier to set the image to TRANSFER_DST.

Bug: angleproject:2652
Bug: angleproject:2653
Change-Id: I485aafc5c9055b5efb101cb2bc3ebab0588d2639
Reviewed-on: https://chromium-review.googlesource.com/1249564
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Frank Henigman <fjhenigman@chromium.org>

This adds support for the index layout qualifier that's used in
EXT_blend_func_extended to set whether a fragment output should be
bound to the primary or secondary blend source color.

Output locations are now validated correctly so that two outputs can
have the same location as long as they have a different index. Some
tests are fixed to allow this.

BUG=angleproject:1085
TEST=angle_unittests

Change-Id: I1de3ad1406398952287791eca367562bed59d380
Reviewed-on: https://chromium-review.googlesource.com/1245982
Commit-Queue: Olli Etuaho <oetuaho@nvidia.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

RobustBufferAccessBehaviorTest.NoBufferData/ES2_OPENGLES
RobustBufferAccessBehaviorTest.NoBufferData/ES3_OPENGLES
RobustBufferAccessBehaviorTest.NoBufferData/ES3_1_OPENGLES
ClearTest.ChangeFramebufferAttachmentFromRGBAtoRGB/ES2_OPENGLES
ClearTest.ChangeFramebufferAttachmentFromRGBAtoRGB/ES3_OPENGLES
WebGLGLSLTest.InitUninitializedGlobals/ES2_OPENGLES

Bug: angleproject:2861,angleproject:2689,angleproject:2862
Change-Id: I142594c952b6e7de24057784794b5725f3486212
Reviewed-on: https://chromium-review.googlesource.com/1252701Reviewed-by: Yuly Novikov <ynovikov@google.com>
Commit-Queue: Yuly Novikov <ynovikov@google.com>

Bug: angleproject:1388
     angleproject:1706

Change-Id: Ic6c62b0a85756bf36b6d345610a77ffd5c4665e3
Reviewed-on: https://chromium-review.googlesource.com/1248442
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>
Reviewed-by: Jiajia Qin <jiajia.qin@intel.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

Also _WIN64 implies _WIN32.

Bug: angleproject:2858
Change-Id: I63e2ffd2e9e304171ea6adb99836733981cc1813
Reviewed-on: https://chromium-review.googlesource.com/1248441
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

It seems conceptually easier to understand that a vk::Framebuffer is
the resource used in graph. Rather than making the GraphResource be
integrated into the FramebufferVk class itself. This means that the
only objects that are graph resources are Vulkan objects: Images,
Buffers, and Framebuffers.

Refactoring change only.

Bug: angleproject:2828
Change-Id: I59d60643182287d4b1fcf9730a3c3a0da5b65973
Reviewed-on: https://chromium-review.googlesource.com/1249561
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Frank Henigman <fjhenigman@chromium.org>

Don't make TextureVk/RenderbufferVk/SurfaceVk/BufferVk own the
manipulation of the command graph. Instead put the operations close to
the buffers and images used to render.

This will lead towards implementing implicit barriers on the command
graph resources.

Bug: angleproject:2828
Change-Id: I07b742b6792c60285b280d6454f90e963d667e0e
Reviewed-on: https://chromium-review.googlesource.com/1246983Reviewed-by: Frank Henigman <fjhenigman@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>

Bug: angleproject:2828
Change-Id: Ie6bcdd10e2de415615db2bfb0b6fa17c392455b0
Reviewed-on: https://chromium-review.googlesource.com/1235655Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Frank Henigman <fjhenigman@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>

When inserting uniform layout info into the default uniform block, the array
sizes need to match but insertion was skipped when the uniform was a sampler.

This caused heap corruption when uniform data was written to the wrong offset
in the buffer.

BUG=angleproject:2859

Change-Id: Ia58a68870e2a4805391266dfe73fe8c8d238931d
Reviewed-on: https://chromium-review.googlesource.com/1249562
Commit-Queue: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>

Bug: angleproject:1944
Change-Id: If05cdf39bd4f88995ed9ea97a13bca661f826884
Reviewed-on: https://chromium-review.googlesource.com/1211782Reviewed-by: Tobin Ehlis <tobine@google.com>

This more easily allows us to compute the format alignment for use with
the vertex input stage.

Bug: angleproject:2797
Change-Id: If15281ce18fbed743b6a0c843cece4626bc4ce72
Reviewed-on: https://chromium-review.googlesource.com/1245841Reviewed-by: Frank Henigman <fjhenigman@chromium.org>
Reviewed-by: Jie A Chen <jie.a.chen@intel.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>

Bug: angleproject:2854
Change-Id: Ib9f229aa579699e90759bacd88d13953c0a7d311
Reviewed-on: https://chromium-review.googlesource.com/1245846
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

This extension is used by Chrome.

BUG=angleproject:2853

Change-Id: Ie8d4ba07857c581a4ec3f90faabdf747b6b52445
Reviewed-on: https://chromium-review.googlesource.com/1246263
Commit-Queue: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

Will be removed upstream.

BUG=chromium:688465
Change-Id: If50ac3624a5f692755797eaabcc0c7d6e7408655
Reviewed-on: https://chromium-review.googlesource.com/1244079Reviewed-by: Yuly Novikov <ynovikov@google.com>
Commit-Queue: Yuly Novikov <ynovikov@google.com>

Bug: angleproject:1951
Test: angle_end2end_tests

Change-Id: I0d8a4f8cf00fc7fd2d85315138e2b7457fd0b90c
Reviewed-on: https://chromium-review.googlesource.com/1242846
Commit-Queue: Jiajia Qin <jiajia.qin@intel.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

The dxgi_format_map_autogen.cpp was updated, but the generating script
was missed to submit in https://crrev.com/c/1142299.

Bug: angleproject:2729

Change-Id: I8d726854581d27881a737f12457482ac955e6312
Reviewed-on: https://chromium-review.googlesource.com/1242853
Commit-Queue: Jie A Chen <jie.a.chen@intel.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

This patch implements the translation from GLSL texture function
textureGatherOffset into HLSL by using the optional "offset"
parameter of Texture2D.Gather[Red|Green|Blue|Alpha].

This patch also defines the implementation-dependent limit
MIN_PROGRAM_TEXTURE_GATHER_OFFSET and MAX_PROGRAM_TEXTURE_GATHER_OFFSET
on D3D11. According to MSDN, the valid range of "offset" used in
Gather should be [-32, 31].
https://docs.microsoft.com/en-us/windows/desktop/direct3dhlsl/gather4-po--sm5---asm-

This patch also refactors OutputTextureGatherFunctionBody() so that
some redundant code can be removed.

BUG=angleproject:2826
TEST=dEQP-GLES31.functional.texture.gather.offset.min_required_offset.2d.rgba8.*
     dEQP-GLES31.functional.texture.gather.offset.min_required_offset.2d_array.rgba8.*
     dEQP-GLES31.functional.texture.gather.offset.implementation_offset.2d.rgba8.*
     dEQP-GLES31.functional.texture.gather.offset.implementation_offset.2d_array.rgba8.*
     (without texture_swizzle)

Change-Id: Id0f411257b64c8a97428f16b1a86950ec6d36e2f
Reviewed-on: https://chromium-review.googlesource.com/1237303Reviewed-by: Jiajia Qin <jiajia.qin@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Jiawei Shao <jiawei.shao@intel.com>

Bug: chromium:889303
Change-Id: Ie06ea7ff27e64d8b149a2200efd159f3a09780b7
Reviewed-on: https://chromium-review.googlesource.com/1244001Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>

CopyTextureChromium sometimes failed because of incorrect image dimensions
being used in the image copy command.  Simply use 0, the images are all
tightly packed.

BUG=angleproject:2723

Change-Id: I2332561bdc9ee70ba07101bcf2db7ea9f2355844
Reviewed-on: https://chromium-review.googlesource.com/1194907
Commit-Queue: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

This adds build_overrides/spirv_tools.gni to tell it where to find its
dependencies.

BUG=chromium:870747
Change-Id: Id1d9fb836e74597860a4411df0a4fd0651840f57
Reviewed-on: https://chromium-review.googlesource.com/1162240Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>

BUG=angleproject:2508

Change-Id: I28782e488495a31da885f35e9c37b0dd22d951a3
Reviewed-on: https://chromium-review.googlesource.com/1236754Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Geoff Lang <geofflang@chromium.org>

Bug: angleproject:2463
Bug: angleproject:2847
Bug: angleproject:2848
Change-Id: I4036363c517dbe293566bb9d51a2470685d41ac2
Reviewed-on: https://chromium-review.googlesource.com/1240108Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>

Line rasterization rules are implemented using a shader patch. The
patch does a small test and discards pixels that are outside of the
OpenGL line region.

The feature is disabled on Android until we can determine the root
cause of the test failures.

Bug: angleproject:2598
Change-Id: Ic76c5e40fa3ceff7643e735e66f5a9050240c80b
Reviewed-on: https://chromium-review.googlesource.com/1120153
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>

It seems that vkCmdCopyBuffer on Windows with Intel GPU neglects
the last region when given more than one.  Work around that in
LineLoopHelper::getIndexBufferForElementArrayBuffer() by adding an
unused region.  Enable corresponding test.

BUG=angleproject:2838

Change-Id: I8847c7b2cfdb94526d4d28ba5bf1f162da3a1ed4
Reviewed-on: https://chromium-review.googlesource.com/1238887Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Frank Henigman <fjhenigman@chromium.org>

In some state change scenarios the driver uniforms would be updated but
not reapplied to the current state. Fix this by setting the descriptor
sets dirty when we dirty the driver uniforms.

Includes a test using gl_DepthRange. Also fixes a bug with the upcoming
line segment rasterization emulation.

Bug: angleproject:2598
Bug: angleproject:2845
Change-Id: Ia66f3c86f9770ceb145069eec879fd8725111a76
Reviewed-on: https://chromium-review.googlesource.com/1240413Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>

Deleting an object that was acting as a framebuffer color attachment
in the current framebuffer didn't previously update all of the
framebuffer state correctly. Fix this by going through the usual
resetAttachment path when any framebuffer attachment is deleted
instead of having custom code that only updated part of the state.

Also early out from clearbuffer calls in case of a missing color
buffer - even now that the draw buffer mask is being updated correctly,
some backend code doesn't take it into account. One example is
querying attachment format when the SRGB clear for linear framebuffer
attachments workaround is active.

BUG=angleproject:2831
TEST=angle_end2end_tests

Change-Id: I1071a60dc0251946fed00e88e43a244fe59f4863
Reviewed-on: https://chromium-review.googlesource.com/1235656Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Olli Etuaho <oetuaho@nvidia.com>

This fixes a crash appearing in robust access mode when a buffer has
no data by stopping the draw call.

A better fix would be to ensure that fetches from an empty buffer
result in zero, but this fix requires a lot more effort. This is not
necessarily warranted since the client is either way doing something
incorrect in this case.

BUG=angleproject:2840
TEST=angle_end2end_tests

Change-Id: Ie4ef5b1ca7c0091d7d26f91678d6e4768cc0c650
Reviewed-on: https://chromium-review.googlesource.com/1238622Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Olli Etuaho <oetuaho@nvidia.com>

When a line loop is drawn with byte indices from a buffer, translate
to short.  Enable test.

BUG=angleproject:2802

Change-Id: Ie178ec866387be85b91291c28e1978db0a09c20a
Reviewed-on: https://chromium-review.googlesource.com/1237293Reviewed-by: Frank Henigman <fjhenigman@chromium.org>
Commit-Queue: Frank Henigman <fjhenigman@chromium.org>

For DynamicBuffer avoid calling vkFlushMappedMemoryRanges() or
vkInvalidateMappedMemoryRanges() if the underlying memory is host-
coherent. These calls are not required for host-coherent memory and
may negatively impact performance.

When currently allocating HOST_VISIBLE memory it's possible to also get
HOST_COHERENT memory even though it's not explicitly requested. Because
of this, the change updates the memory allocaiton interfaces through
the stack to pass back the actual allocated memory properties so that
it can be checked to see if it's host-coherent.

Bug: angleproject:2804
Change-Id: Ife95c4b98115b16d16c087dd72dba7d9661fdb46
Reviewed-on: https://chromium-review.googlesource.com/1234276
Commit-Queue: Tobin Ehlis <tobine@google.com>
Reviewed-by: Yuly Novikov <ynovikov@chromium.org>

The new API is named 'recordCommands'.

These two APIs were basically doing the same thing. We don't need to
have an understanding of creating a new graph node to know that we
want to record some Vulkan commands to a command buffer. The prior
design was actually masking a bug where we would allow appending
commands to a command graph node that had already started a render
pass. Fix this by adding a render pass check to recordCommands.

Also removes 'hasStartedWriteResource' since this method wasn't used
anywhere.

Also renames 'onResourceChanged' to 'finishCurrentCommands'.

Bug: angleproject:2828
Change-Id: I00bd5b893fcfc37172b6c1706cb2f5fc57e79f54
Reviewed-on: https://chromium-review.googlesource.com/1235654
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Frank Henigman <fjhenigman@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

Bug: angleproject:2843
Change-Id: If86ed33f7fbc70e4f10580d2b52ff04454e95f52
Reviewed-on: https://chromium-review.googlesource.com/1240416Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>

We sometimes generate local dirty bits in TextureGL. To make sure the local
dirty bits don't get skipped we use a Subject/Observer pattern between the
TextureGL and gl::Texture. This allows us to skip syncState in the hot path.

Also inlines a couple of other texture functions. And fixes a stray header
in EGLBlobCacheTest.

Bug: angleproject:2763
Change-Id: Ie1d8a5865deaf2a563a358c31ae28bef6b2458b1
Reviewed-on: https://chromium-review.googlesource.com/1228374
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Yuly Novikov <ynovikov@chromium.org>

In single-threaded cases we don't need to use a TLS lookup for the
current Context.

Assume a global context has 3 states: unset, set, and multi-threaded:

* global context unset -> set the global context
* global context set, matches TLS context -> set the global context
* global context set, doesn't match TLS -> set multi-threaded
* global context multi-threaded -> leave it

This also inlines the global context/TLS check into global_state.h.
Increases performance of the binding perf overhead test by up to 25%
and the draw call perf test by up to 3%.

Bug: angleproject:2824
Change-Id: Ib95953ca918f2408fe72ca83061f4e2d1bd1adad
Reviewed-on: https://chromium-review.googlesource.com/1231993
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Yuly Novikov <ynovikov@chromium.org>

These were already being signaled in the front end.

Bug: angleproject:2763
Change-Id: Id78d3d764e80e687c29c08395b59048d314bfbe2
Reviewed-on: https://chromium-review.googlesource.com/1204490Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>

This patch is the third one to support atomic functions on D3D11.

In this patch we enable support for atomic function returns outside of
assignments (e.g. part of arithmetic operations or to index into arrays) and
when used directly initialize a variable.

Note that we are still missing the functionality to be tag loops with
[allow_uav_condition] as required by InterlockedCompareExchange.

BUG=angleproject:2682
TEST=angle_end2end_tests

Change-Id: Ia409ebb10621fd5c514cf6c76f366a320a9d9fc1
Reviewed-on: https://chromium-review.googlesource.com/1208317Reviewed-by: Bryan Bernhart <bryan.bernhart@intel.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>

An increased crash rate in rx::NativeWindow11Win32::createSwapChain has been
seen around the same time as this patch landed.  Speculatively reverting.

BUG=887403

This reverts commit a49f60fe.

Change-Id: Id1b08efbf47196f17f4a5e509a29f7d628b07b49
Reviewed-on: https://chromium-review.googlesource.com/1238884Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Geoff Lang <geofflang@chromium.org>