This patch intends to add missing initialization of MAX_INTEGER_SAMPLES,
MAX_DEPTH_TEXTURE_SAMPLES and MAX_COLOR_TEXTURE_SAMPLES required for
multisampled textures on D3D backends.

Since MAX_*_SAMPLES cannot be queried directly from D3D APIs, these
values are initially assigned a large value in renderer11.cpp and
re-calculated in context::updateCaps().

This patch also adds tests to ensure these values are greater than 1
as required in OpenGL ES3.1 spec.

BUG=angleproject:1592
TEST=angle_end2end_tests

Change-Id: Iba586e311d40d2da4569816902f96e40bbd6935b
Reviewed-on: https://chromium-review.googlesource.com/597411
Commit-Queue: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>

In D3D cube map textures use different code paths than regular 2D
textures.  Add outside-the-framebuffer handling to those paths, same as
was added the other paths earlier.

Change-Id: I51896a07f73ae8d761cd9d7b18c68076f38d32a3
Reviewed-on: https://chromium-review.googlesource.com/603050
Commit-Queue: Frank Henigman <fjhenigman@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

Make sure that the type definitions for nested structs get added to
the HLSL header, and that the std140 padding information gets
recorded. Prior to this trying to use nested structs in interface
blocks crashed.

BUG=angleproject:2084
TEST=angle_end2end_tests

Change-Id: If57870285c6feaf0c2e462f98f50f20730dd6470
Reviewed-on: https://chromium-review.googlesource.com/608449Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Olli Etuaho <oetuaho@nvidia.com>

Cube maps go through a separate path on D3D so we need to test them.

BUG=angleproject:1815

Change-Id: Ifb7a85d7e2750f25bce382fdd7a00062d23b3573
Reviewed-on: https://chromium-review.googlesource.com/597213
Commit-Queue: Frank Henigman <fjhenigman@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

In HLSL output, structs in interface blocks are not accessed directly.
Rather they get copied from the D3D constant buffer to static structs
in the shader header. Fix generating the copy/init code in the header
to handle arrays of structs correctly.

BUG=angleproject:2084
TEST=angle_end2end_tests

Change-Id: If66bd5be3f3570ba591b8b62c5284c06fc83dd45
Reviewed-on: https://chromium-review.googlesource.com/608448Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Olli Etuaho <oetuaho@nvidia.com>

BUG=angleproject:1932

Change-Id: I6474237cbb82b59a0bd40c1b9b9e2455952d3755
Reviewed-on: https://chromium-review.googlesource.com/600510
Commit-Queue: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>

If the ANGLE_multiview extension is used in a program, the number
of geometry instances is the number of instances passed to
glDraw*Instanced times the number of views in the program. The attribute
divisor has to be multiplied by the number of views so that the correct
attributes are gathered in the input assembly stage.

BUG=angleproject:2062
TEST=angle_end2end_tests

Change-Id: I960d6313c02e3eb83f7a07e72b9bcac072c736f4
Reviewed-on: https://chromium-review.googlesource.com/593953
Commit-Queue: Martin Radev <mradev@nvidia.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

The patch refactors the tests to use the GL RAII objects from
gl_raii.h. It also specifies a correct internal format for a texture 2D
array object in one of the tests so that the texture is
color-renderable.

BUG=angleproject:2062
TEST=angle_end2end_tests

Change-Id: Ie536ffeb5b2f2594f5273b795fe786a5042f63d4
Reviewed-on: https://chromium-review.googlesource.com/603610
Commit-Queue: Martin Radev <mradev@nvidia.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Olli Etuaho <oetuaho@nvidia.com>

When a function parameter name conflicts with another, the pointer
returned to ParseContext will be null.

BUG=chromium:745242
TEST=angle_unittests

Change-Id: Ie53bb06b0c6660e382d85aeda41f3a1b7df5a917
Reviewed-on: https://chromium-review.googlesource.com/603368Reviewed-by: Corentin Wallez <cwallez@chromium.org>

The patch adds two tests to measure the performance of the
ANGLE_multiview extension implementation:
1) The first test renders onto two views by issuing thousands of state
changes and draw calls. The aim of the test is to stress the CPU.
2) The second test renders onto two views by drawing with one draw call
half a million quads with multiple attributes per vertex. The attributes
are passed to the fragment shader where they are used for computing the
color. The aim of the test is to stress the GPU's memory system.

The patch also extends the ANGLEPerfTest's functionality to only run
the benchmark if the necessary extensions are available.

BUG=angleproject:2062
TEST=angle_end2end_tests

Change-Id: Ic63d54a69fdddb72439eeeb1951a500fb1247e95
Reviewed-on: https://chromium-review.googlesource.com/597630Reviewed-by: Olli Etuaho <oetuaho@nvidia.com>

BUG=angleproject:2123

Change-Id: Ic3c70c7148297662d4fd9e08edcd208f5a1cc885
Reviewed-on: https://chromium-review.googlesource.com/601370Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

Add API entry and validation checks(GLES 3.1 section 7.3).
Add the first 2 interfaces(PROGRAM_INPUT and PROGRAM_OUTPUT) implementation.

BUG=angleproject:1920
TEST=angle_end2end_tests:ProgramInterfaceTestES31.*

Change-Id: I50057f7b99f4dc7c23ca87fa9b797ca424f66e3d
Reviewed-on: https://chromium-review.googlesource.com/475075Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

The patch enables the ANGLE_multiview extension in the D3D11 renderer
and determines the maximum number of available views.

BUG=angleproject:2062
TEST=angle_end2end_tests

Change-Id: I56e434a14c61df3d6ab66c2863ce783fcdcc5a93
Reviewed-on: https://chromium-review.googlesource.com/595507
Commit-Queue: Martin Radev <mradev@nvidia.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Olli Etuaho <oetuaho@nvidia.com>

The __has_attribute macro was one of improvements in GCC 5 release.
See https://gcc.gnu.org/gcc-5/changes.html.
This checks to make it work with older GCC.

BUG=angleproject:2119

Change-Id: I751249b9d028d1322829bebc1ca8d423edf27057
Reviewed-on: https://chromium-review.googlesource.com/599152
Commit-Queue: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

Was skipped on Windows and Mac before, now seen flaky also on Linux.

BUG=angleproject:1541

Change-Id: Ic993eeadb104989416607271ce100b02aea7cecc
Reviewed-on: https://chromium-review.googlesource.com/603087Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Commit-Queue: Yuly Novikov <ynovikov@chromium.org>

TextureRectangleTest.TexStorage2D and TextureRectangleTest.TexImage2D
failed on NVIDIA Quadro P400, previously disabled on Linux.

BUG=angleproject:2122
TBR=cwallez@chromium.org

Change-Id: I471dc040ed51a0e4588cfaab7d1c4c46e69932d5
Reviewed-on: https://chromium-review.googlesource.com/602779Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Commit-Queue: Yuly Novikov <ynovikov@chromium.org>

TextureRectangleTest.TexStorage2D and TextureRectangleTest.TexImage2D
failed on NVIDIA Quadro P400.

BUG=angleproject:2122
TBR=cwallez@chromium.org

Change-Id: I72d1755881cb06ef4363266b368ada5ca954f6dd
Reviewed-on: https://chromium-review.googlesource.com/601348Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Commit-Queue: Yuly Novikov <ynovikov@chromium.org>

This is needed to support binding IOSurfaces to textures on OSX. This
commit adds support in the API and tests, but didn't need to implement
compiler changes as it already supported ARB_texture_rectangle.

Implementation of CHROMIUM_opy_texture for rectangle texture and the
spec are left for follow-up commits.

Change-Id: I45c66be763a9d3f6f619640f9f95f39b05c70867
Reviewed-on: https://chromium-review.googlesource.com/559106
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

BUG=angleproject:1932

Change-Id: I4d19892852540797c765916131498c96511b4f55
Reviewed-on: https://chromium-review.googlesource.com/600509Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Geoff Lang <geofflang@chromium.org>

Bug: 495204
Change-Id: I6d460a9e6d110be4697e3cc1685e16395276dd42
Reviewed-on: https://chromium-review.googlesource.com/600258Reviewed-by: Jamie Madill <jmadill@chromium.org>

By marking the images as dirty after binding a surface, the surface would
be cleared when it was first read or written to.

BUG=750813
BUG=angleproject:1635

Change-Id: Ic0d1c985151d55a0f1a1af67bb1edc4b0e8f2063
Reviewed-on: https://chromium-review.googlesource.com/598731
Commit-Queue: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

BUG=angleproject:2121

Change-Id: Ie7d09aca453753577dd26dcd35b1c743dbfcfce2
Reviewed-on: https://chromium-review.googlesource.com/598495
Commit-Queue: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

Also includes some tracking of if a cache hit comes from a binary
inserted after a Link call, or sourced from the external disk cache.

Chromium-side CL: https://chromium-review.googlesource.com/c/592151/

BUG=angleproject:2118

Change-Id: I80eefd203d8ce31d1ac03dd1a36459d3581128f5
Reviewed-on: https://chromium-review.googlesource.com/590689
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

This is a follow-up CL for
https://chromium-review.googlesource.com/c/544634/.
In this commit, we add drawElements perf test for d3d9, d3d11 and
OpenGL. Specially, we test DrawElements with unchanged index buffer
every frame and changed index buffer every frame. This is also a
preparation for further remove IndexRange in validation.
Paste a set of data (repeated 30) for reference on Intel skylake
Win10 desktop.
DrawElementsPerfBenchmark.Run/d3d11:
mean: 13186.1666667, stddev: 200.479605502,
truncated mean: 13192.5416667, stddev: 156.284190704
DrawElementsPerfBenchmark.Run/d3d11_index_buffer_changed:
mean: 43.6, stddev: 4.02160830183,
truncated mean: 43.5833333333, stddev: 2.97092391166

BUG=angleproject:1393
TEST=angle_perftests

Change-Id: I94bd181aa2b4515df1affc79f591f5db399feac3
Reviewed-on: https://chromium-review.googlesource.com/585981
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

Because the ANGLE_multiview extension uses instancing to multiply
geometry for each view, Draw* calls with an active multiview program
have to be handled in the follwing way:
1) Convert non-instanced Draw calls to their instanced versions.
2) Multiply the number of instances in an instanced Draw call by the
number of views.

BUG=angleproject:2062
TEST=angle_end2end_tests

Change-Id: I19119e8178f70336e5dbb1e5eed0658b5b9f43d7
Reviewed-on: https://chromium-review.googlesource.com/593657Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Martin Radev <mradev@nvidia.com>

This patch intends to split all vector<Varying> into two vectors
to store input and output varyings separately in the compiler.

This patch is a base of implementing the built-ins, inputs and
outputs of a geometry shader to ANGLE GLSL compiler. Unlike the
vertex shaders (their outputs are varyings) and fragment shaders
(their inputs are varyings), the inputs and outputs of geometry
shaders are all varyings, so we need two vector<Varying> to store
them correctly.

BUG=angleproject:1941

Change-Id: I9e8cc16045d5e29e9a80a09dc31b33a7ae39b345
Reviewed-on: https://chromium-review.googlesource.com/593347
Commit-Queue: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

Clear* commands for side-by-side framebuffers require special handling
because only the scissor rectangle of the first viewport is used in the
scissor test as defined in the OpenGL 4.1+ specs.

To enable clearing of each view of a side-by-side framebuffer all views
are iterated over, the corresponding scissor rectangle is set as first,
and a Clear* call is made to the driver. Afterwards the scissor state is
restored.

BUG=angleproject:2062
TEST=angle_end2end_tests

Change-Id: I138663ea61b4f0c9302872108e7dfbadf451b3ec
Reviewed-on: https://chromium-review.googlesource.com/590233
Commit-Queue: Martin Radev <mradev@nvidia.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>

Thanks to Markus from NVIDIA. This inlines some methods and optimizes
the check if a uniform is a bool type.

BUG=angleproject:2119

Change-Id: I6f2c2d22c577458b39de600d3c56ec8e1a456a7a
Reviewed-on: https://chromium-review.googlesource.com/591699
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Kenneth Russell <kbr@chromium.org>

WebGL can request contexts with a stencil buffer but no depth buffer.

BUG=angleproject:2075

Change-Id: I46b272089e44d5f1725fa6252b68d34c9e9f37f6
Reviewed-on: https://chromium-review.googlesource.com/565052Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Geoff Lang <geofflang@chromium.org>

Regressed: https://chromium-review.googlesource.com/531798

This was because the std::stringstream constructor was being called
every draw call, for the InfoLog of the dynamically recompiled
shaders in the D3D back-ends. The constructor was allocating memory
and freeing it on destruction. Instead use a lazy init for the
std::stringstream in InfoLog like we do for Error.

BUG=chromium:750685

Change-Id: I9cd429ae4e1439ae504943a9cad31cbbed17ef32
Reviewed-on: https://chromium-review.googlesource.com/594629
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

Side-by-side framebuffers have viewport offsets as part of their state
which have to be applied to the viewport and scissor rectangles to
generate the final viewport and scissor rectangles of each view.
Whenever there is a transition to or from a side-by-side framebuffer,
viewport and scissor state has to be synced. Also, because rendering is
done on the same 2D texture the scissor test has to be always enabled to
guarantee that no fragments leak to a neighboring view.

The patch addresses this by extending the viewport and scissor state
in StateManagerGL to be a vector of rectangles instead of a single
rectangle. Two new dirty bits are added to cover changes in the viewport
offsets and whether the framebuffer has a side-by-side layout.

BUG=angleproject:2062
TEST=angle_end2end_tests

Change-Id: I8107d7ba97d06b20cf24358f19963fa494844592
Reviewed-on: https://chromium-review.googlesource.com/585012
Commit-Queue: Martin Radev <mradev@nvidia.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

According to the ANGLE_multiview spec Draw* commands should generate
an INVALID_OPERATION error if there is an active query object for
target TIME_ELAPSED_EXT and the number of views in the active draw
framebuffer is greater than 1.

BUG=angleproject:2062
TEST=angle_end2end_tests

Change-Id: I8a4434784ecec753a39c5ef82fa3ee46255a0851
Reviewed-on: https://chromium-review.googlesource.com/593315Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Martin Radev <mradev@nvidia.com>

According to the ANGLE_multiview spec, glReadPixels must generate an
INVALID_FRAMEBUFFER_OPERATION error if the active read framebuffer has
a multi-view layout.

BUG=angleproject:2062
TEST=angle_end2end_tests

Change-Id: Ia5311ff7a62c5ff732491eb80befd32de57b9d44
Reviewed-on: https://chromium-review.googlesource.com/591368
Commit-Queue: Martin Radev <mradev@nvidia.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

BUG=angleproject:1987

Change-Id: Iebf9aeba61bfaa300de3ed50d7d57c61996f67dc
Reviewed-on: https://chromium-review.googlesource.com/592139Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>

This simply translates the offset qualifier and synchronizes
the buffer binding state for GL translator and renderer.

BUG=angleproject:1729
TEST=angle_end2end_tests:AtomicCounterBufferTest31

Change-Id: Ib8424918478ae4c47049d5856ea20ef022f12913
Reviewed-on: https://chromium-review.googlesource.com/575913
Commit-Queue: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

According to the ANGLE_multiview spec, glBlitFramebuffer must
generate an INVALID_FRAMEBUFFER_OPERATION error if either the active read
framebuffer, or active draw framebuffer has a multi-view layout.

BUG=angleproject:2062
TEST=angle_end2end_tests

Change-Id: I885bdc970c9606cfad882f31759f5780c65d15e5
Reviewed-on: https://chromium-review.googlesource.com/590237
Commit-Queue: Martin Radev <mradev@nvidia.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

According to the ANGLE_multiview spec, glCopyTex* functions must
generate an INVALID_FRAMEBUFFER_OPERATION error if the active read
framebuffer has a multi-view layout.

BUG=angleproject:2062
TEST=angle_end2end_tests

Change-Id: Icadc4ac79843986076594da25a90ba807e511d1e
Reviewed-on: https://chromium-review.googlesource.com/589447Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Martin Radev <mradev@nvidia.com>

According to the ANGLE_multiview spec Draw* commands should generate
an INVALID_OPERATION error if there is an active transform feedback
object and the number of views in the active draw framebuffer is greater
than 1. The patch addresses this by extending the base draw call
validation.

BUG=angleproject:2062
TEST=angle_end2end_tests

Change-Id: I67221cb2cfee6febae8d97697b234aeffff313de
Reviewed-on: https://chromium-review.googlesource.com/589268
Commit-Queue: Martin Radev <mradev@nvidia.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

According to the ANGLE_multiview spec Draw* commands should generate
an INVALID_OPERATION error if the program uses the multiview extension
and the number of views in the active draw framebuffer and active
program differs. The patch addresses this by extending the base draw
call validation.

BUG=angleproject:2062
TEST=angle_end2end_tests

Change-Id: I369070beb5ccb17211dbe61ebec40bfcbcf5bc4e
Reviewed-on: https://chromium-review.googlesource.com/586605
Commit-Queue: Martin Radev <mradev@nvidia.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>

It is legal to blit between GL_RGBA8 and GL_BGRA8 sources and
destinations when resolving multisampled renderbuffers. Expand
BlitFramebuffer's validation to handle this case.

Query GL_MAX_VERTEX_OUTPUT_COMPONENTS on the Core Profile to work
around an error generated on macOS when querying
GL_MAX_VARYING_COMPONENTS.

Expand the BlitFramebuffer tests to cover these cases and start
running them on the OpenGL backend.

Fix detectition of multisampled D3D11 rendertargets when the sample count
is 1.

BUG=angleproject:891

Change-Id: Ief5531756651caa66f612e647d3d5c05c8c51ff5
Reviewed-on: https://chromium-review.googlesource.com/587459Reviewed-by: Kenneth Russell <kbr@chromium.org>