This enhances to use the native parallel compile extensions if
available.

Bug: 873724

Change-Id: I0aaed314accd75e1bfa596b322225b56d729d3a6
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1475234
Commit-Queue: Jie A Chen <jie.a.chen@intel.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

ssbo.a[uniformBlock.index] should be a valid usage. So in this change,
we forward EOpIndexDirectInterfaceBlock to OutputHLSL to process when
the left node is not a SSBO.

Bug: angleproject:3280
Change-Id: I2696d5b1cf420db133a8ba0728116261dd9e9f1e
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1531980Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Jiajia Qin <jiajia.qin@intel.com>

This both speeds up test execution and cuts down on the number of new
windows and displays created for a test config. This feature is only
currently enabled for Windows NVIDIA and Intel. On every other config
there were blocking issues that would need investigation. Several tests
were manually flagged as needed new displays on each iteration to
prevent test flakiness.

This feature might fix the issues with Intel test flakiness that have
been prominent on the ANGLE CQ.

WGL configurations have also been removed from ANGLE tests. So this
removes more of the code from ANGLETest.cpp.

Bug: angleproject:3261
Change-Id: Ic2864d4806ad38e0eeaa3c0afcd54ae1c548090f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1520995
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Yuly Novikov <ynovikov@chromium.org>

Bug: angleproject:3227
Change-Id: I3bbdb6956cf97298eff9bd2a766d048f1de4a70f
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1531538
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

This perf test measures the GPU time spent clearing framebuffers.

Bug: angleproject:2361
Change-Id: I5dc2e12c08543330f7e2565596b73e9c2f4e53ef
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1529864
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

This allows us to more easily compare sets of parameters used in our
tests. The config parameters are stuff like the red / gree / blue bits
used in an EGL config. Or particular sets of extensions or other
EGL options.

This will more easily allow us to determine when we need to use a new
EGL display instead of reusing a prior.

Bug: angleproject:3261
Change-Id: Ia1f0ede988e0b4084fbb4d55097e94fd89ee4899
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1531535
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Yuly Novikov <ynovikov@chromium.org>

Fixes the suprious newlines messages. Also prefer a Release directory
when writing shaders for performance. Also adds glslang's git HEAD to
the list of generated hashes. This should ensure we regenerate all
shaders when there's a roll of glslang.

Glslang was rolled prior to this CL. Also update the SPRI-V libs.

Bug: angleproject:3227
Change-Id: I1fd212a08000c08ef41b60c4201edd3ac771f91d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1516512Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>

Otherwise we could potentially be destroying resources that are
currently in use.

Was caught by changing some of the test tear down ordering. Also might
affect some of the dEQP EGL multithread tests.

Requires fixing the serial tracking of the "null" buffer. Otherwise we
could be destroying the null buffer in the incorrect order.

Bug: angleproject:3261
Bug: angleproject:3059
Change-Id: I57d1059ce9b1da19ced8ac753fab80c625901066
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1531291Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>

We were using a static SystemInfo in two places. Consolidate the
SystemInfo collection to a single location. Also renames Nvidia to
NVIDIA to be consistent with the company naming. And adds a few helpers
to SystemInfo for GPU detection.

Will lead to test changes to reduce flakiness on Intel Windows.

Bug: angleproject:3261
Change-Id: I4e0addf19d6fe26b4d31a1289efce72092a0a1dd
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1531533Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
Reviewed-by: Yuly Novikov <ynovikov@chromium.org>

Calling eglMakeCurrent between different D3D9 contexts could lead to
state sync bugs. This fixes the problem by ensuring all internal
state is dirtied between makeCurrent calls.

This bug turned up when refactoring the tests to run multiple Contexts
with the same EGLDisplay.

Bug: angleproject:3261
Change-Id: Ic91e1af8b2615e7e7e0d0faabeb533d075ea2e91
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1531534
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>

From angle_vulkan build target,
so that everyone including Vulkan Headers
would get them for the right platform.

Also remove obsolete defines.

Bug: angleproject:3279
Change-Id: I4bbbeb72acfbf0dd22a1d6a13bd9df7ddaec84dd
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1530029Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Tobin Ehlis <tobine@google.com>
Commit-Queue: Yuly Novikov <ynovikov@chromium.org>

Bug: chromium:943604
Test: angle_end2end_tests

Change-Id: Id4bd293bea0587687d1656998bc6d986fd9fb8e9
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1530027
Commit-Queue: Michael Spang <spang@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

Bug: chromium:941770
Test: angle_unittests

Change-Id: I7dbee848b251daaaf1b4f85c5e7e76bb1fa1fbbf
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1530026
Commit-Queue: Michael Spang <spang@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

The model name, model version, and model manufacturer are easy to expose for
Android via <sys/system_properties.h>. This should be exposed and added to the
angle test framework for easy management of test expectations on different
devices.

Bug: angleproject:3274
Change-Id: I8ee6b8fa66ff7f4d6ee4688b335f2e6ef03baed6
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1529207
Commit-Queue: Jonah Ryan-Davis <jonahr@google.com>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>

Bug: angleproject:2280

Change-Id: Iad82d5838a7efdb6f6287aafb9ab980e9e86468d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1496017Reviewed-by: Jamie Madill (use @chromium please) <jmadill@google.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Commit-Queue: Jiajia Qin <jiajia.qin@intel.com>

Bug: angleproject:3125
Change-Id: Ifa45d489cb082355a23f33f05f805e5f6162ac0c
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1529232Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>

Useful for binary size attribution (and bug filing)

Change-Id: Ie77a331deff7094f8a386082141fbc49d902dd2a
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1528926Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Commit-Queue: Yuly Novikov <ynovikov@chromium.org>

GetLoadFunctionsMap() is expected to cover all possible combinations of
texture internalformat and upload data format.  Several of these
combinations were missing, especially w.r.t to compressed types.  This
change adds the missing combinations that are hit on any of the
configurations being tested.  An assert is added to catch missing
conversion functions in the future.

Bug: angleproject:2670
Change-Id: I793ac2c57f65eeb1cc8da8f1b223f6a9a44c7708
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1524462Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>

On D3D the vertex ID is always indexed from 0, unlike GL. The D3D
backend had assumed the wrong behavior. This forwards the true offset to
D3D by using the ConstantsBuffer.

Bug: angleproject:3090
Change-Id: Ia19e3490503c97541af14979b75af0c94c67ab6b
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1520988
Commit-Queue: Jonah Ryan-Davis <jonahr@google.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

Usage: trigger.py <GN path> <test> <os_dim> <gpu_dim> <-s shards>

This script can circumvent the trybots when attempting a job on a
specific configuration. It must be used within a Chromium checkout
currently. In the future we can likely use it in standalone.

Optional arguments can specify the number of swarming shards, the pool,
and extra command line test parameters like --gtest_filter.

Bug: angleproject:3272
Change-Id: Ia79f4422786c463378fc7e0f7f5ae4d07bd097a0
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1524700
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Yuly Novikov <ynovikov@chromium.org>

Bug: angleproject:1944
Change-Id: I6c47487f4dd43b3d3caecbc9256654865518d775
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1524467Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>

ANGLE's ASSERT macro was often not printing the error message before
crashing. The printing happens in the destructor of a temporary object,
but the temporary object is not guaranteed to be destroyed before ASSERT
calls __builtin_trap(). This change fixes it by introducing a new FATAL
log severity that moves the __builtin_trap() call into the destructor
after the message is printed.

Bug: 780507, angleproject:2215
Change-Id: Ica91e00d67cc95d7b59c3d59e6925ac3a93c19c5
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1521839
Commit-Queue: James Darpinian <jdarpinian@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

Additionally, only one result is output from the tests.  If running
every scene, only the final result is output.

Bug: angleproject:3125
Change-Id: I9592a6e07757663bfdd6653f541fece95c4b295b
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1524695Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>

The sampler/texture type validation validation added in
https://chromium-review.googlesource.com/c/1377611 did not correctly
consider sampler objects. This is a minimal change to disable the
validation temporarily, which I intend to merge back to 74. A fix
for the validation is in progress at:
https://chromium-review.googlesource.com/c/angle/angle/+/1522364

Bug: 940080, 809237
Change-Id: If8a1ae46af7daf23cd92ccb8a985329dfd404dc9
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1524475
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

Many TearDown calls were missing. Also a few tests had other small
issues. Discovered while changing the test harness.

Bug: angleproject:3261
Change-Id: I8915088a6ca82e6cc4d6c922d03dc447e6051518
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1524699Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>

Previously we would ignore unsetting calls to MakeCurrent. But this
would prevent ANGLE from unsetting driver state on the current thread.

Fix this by checking for 0/null explicitly in DisplayWGL::makeCurrent.

This turned up when running test configs without recreating the
Display.

Bug: angleproject:3261
Change-Id: I7f747dc7649a06f019f1b89655cb71dd3b151aa0
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1520992
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Geoff Lang <geofflang@chromium.org>

Zero-sized loops or loops that might never be executed may be unrolled into
constant array accesses that are invalid, generating HLSL compiler errors.

This issue showed up with the passthrough command decoder because it only does
a single compilation step of WebGL shader -> HLSL instead of the validating
decoder which does WebGL shader -> ESSL -> HLSL and eliminates these zero-
sized loops in the process.

BUG=angleproject:3269

Change-Id: Iaa14149c542f23a6da58c90cf6de51f4df7e27d6
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1524696Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Jonah Ryan-Davis <jonahr@google.com>
Commit-Queue: Geoff Lang <geofflang@chromium.org>

Bug: angleproject:3172
Change-Id: I4ce55a71db204acdd73a4f9db0c5480b31e0b3ad
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1521317
Commit-Queue: Ian Elliott <ianelliott@google.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

A previous refactor made queries no longer a graph resource.  This in
turn meant that the query nodes no longer had an ID, which tripped up
the ID assignment when dumping the graph.

This affected fences as well.

Bug: angleproject:2853
Change-Id: I359f8145dca068edaa5fcd5bcb0a231ee050ab76
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1520990
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

The barrier dependencies were set in two loops:

- From Previous to (Previous, Current]
- From (Previous, Current) to Current

The "Previous to Current" dependency was set in the first loop.  This
loop is conditioned to a previous barrier existing at all.

The bug is that for the first barrier node, the first loop does not
execute and the second loop does not create the dependency from the
first node (Previous index being 0 when no previous barrier exists) to
the barrier node.

Bug: angleproject:2361
Change-Id: I61a78dbbeba6cf97788bf8ea32908c7865ad48c2
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1520994
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

When copying to a non-cube-complete level of a texture we would fail
because we would try to get the base level's current format. Instead
we can get the format of the level that is being copied to.

Bug: angleproject:2911
Change-Id: Ib6a5c5cef67ad9c880c1b76ea71ed9317eb20c97
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1483951
Commit-Queue: Shahbaz Youssefi <syoussefi@chromium.org>
Reviewed-by: Yuly Novikov <ynovikov@chromium.org>

The EXT_color_buffer_half_float spec states:

    If OES_texture_half_float is supported, textures created with:

        <internalformat> = RGBA
        <format> = RGBA
        <type> = HALF_FLOAT_OES

    are renderable.

But WebGL content relies on being able to render to both RGB and RGBA unsized
HALF_FLOAT_OES textures.

Framebuffers may still return unsupported for this format.

BUG=941671

Change-Id: I42d909d26508b9cdf7cda6d1dc025b0e2ed87732
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1523530Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Commit-Queue: Geoff Lang <geofflang@chromium.org>

The changes are to allow build files to be generated using Python 3 on Windows.

The scripts still work with Python 2.
There are no intended behaviour changes.

Bug: 941669
Change-Id: Ia9d8d61373ee37b4470fd5479e9db01e17994ffa
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1520649
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

Opengl32.lib/opengl32.dll are not available on Windows ARM64. Disable building
glmard2_wgl which depends them.

Bug: chromium:941621

Change-Id: Iddae231764c6135ef1a74c294da1393f0fe51eeb
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1522353Reviewed-by: Jamie Madill <jmadill@chromium.org>

This reverts commit 13a8c4d8.

Reason for revert: Seems to have led to a flakier situation:

https://ci.chromium.org/p/chromium/builders/ci/Win10%20FYI%20Debug%20%28NVIDIA%29/3006
https://ci.chromium.org/p/chromium/builders/try/win-angle-rel/1231

Original change's description:
> Fix several WGL test failures.
> 
> SimpleOperationTest.ClearAndSwap/ES2_WGL failed when run in isolation,
> since getGLWindow()->hasError() would report a previous error,
> instead of result of swapBuffers().
> When running after an OPENGL test, swapBuffers() would clear
> the previous error, but that doesn't happen in isolation.
> 
> The previous error is from loading WGL functions, some of which are
> expected not to be present. Clear the error in GetProcAddressWithFallback,
> but verify that there is no error entering it.
> 
> This uncovers more errors in angle_perftests:
> DrawCallPerfBenchmark.Run/wgl
> DrawCallPerfBenchmark.Run/wgl_tex_change
> DrawCallPerfBenchmark.Run/wgl_vbo_change
> DrawElementsPerfBenchmark.Run/wgl_ushort
> They come from redundant calls when destroying a window. Fix this as well.
> 
> Several more errors where uncovered by debug prints, fix those, too.
> 
> Bug: angleproject:3153
> Change-Id: I559c098be9dcdfd3add83f045f745d190250b986
> Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1515602
> Commit-Queue: Yuly Novikov <ynovikov@chromium.org>
> Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>

TBR=ynovikov@chromium.org,geofflang@chromium.org,syoussefi@chromium.org,jmadill@chromium.org

Change-Id: I095fadc0dd3a2c998c1dc86f3760184ae6fd7309
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: angleproject:3153
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1523527Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>

The spec says that EGL_COLOR_BUFFER_TYPE "must be either
EGL_RGB_BUFFER for an RGB color buffer, or EGL_LUMINANCE_BUFFER for a
luminance color buffer."  However, Config::Config() was setting it to
EGL_NONE, which causes EGLChooseConfigTest to fail on the Ozone
platform (which doesn't override the value).  This sets the proper
value for all platforms (which can still override this default).

Bug: angleproject:3260
Change-Id: Iaf59bd76b11c8d80359c760c10baad3b79c79f9d
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1521310Reviewed-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Commit-Queue: Ian Elliott <ianelliott@google.com>

Bug: chromium:930993, chromium:938602
Change-Id: Ie2e07369eca663b1507c7811a1a06ce5efc892ee
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1516007
Commit-Queue: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

The test had an additional check to make sure we have coverage on every
platform, but this is unnecessary and requires too much overhead to
maintain a list of good locales for all platforms.

Bug: 941249

Change-Id: I3e6da2ad61453fc385edf5f18101c8cb13ae0503
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1520989
Commit-Queue: Jonah Ryan-Davis <jonahr@google.com>
Reviewed-by: Geoff Lang <geofflang@chromium.org>
Reviewed-by: Jamie Madill <jmadill@chromium.org>

SimpleOperationTest.ClearAndSwap/ES2_WGL failed when run in isolation,
since getGLWindow()->hasError() would report a previous error,
instead of result of swapBuffers().
When running after an OPENGL test, swapBuffers() would clear
the previous error, but that doesn't happen in isolation.

The previous error is from loading WGL functions, some of which are
expected not to be present. Clear the error in GetProcAddressWithFallback,
but verify that there is no error entering it.

This uncovers more errors in angle_perftests:
DrawCallPerfBenchmark.Run/wgl
DrawCallPerfBenchmark.Run/wgl_tex_change
DrawCallPerfBenchmark.Run/wgl_vbo_change
DrawElementsPerfBenchmark.Run/wgl_ushort
They come from redundant calls when destroying a window. Fix this as well.

Several more errors where uncovered by debug prints, fix those, too.

Bug: angleproject:3153
Change-Id: I559c098be9dcdfd3add83f045f745d190250b986
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1515602
Commit-Queue: Yuly Novikov <ynovikov@chromium.org>
Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>

The implementation of eglChooseConfig does not match the
specification for attributes not provided by the calling function.  It
is supposed to use the default value and match semantics, per the
specification.  This is fine for many attributes (where the value in
the EGLConfig doesn't matter).  Currently, this affects the following
attributes:

- EGL_COLOR_BUFFER_TYPE
- EGL_LEVEL
- EGL_RENDERABLE_TYPE
- EGL_SURFACE_TYPE
- EGL_TRANSPARENT_TYPE
- EGL_OPTIMAL_SURFACE_ORIENTATION_ANGLE
- EGL_COLOR_COMPONENT_TYPE_EXT

This change causes 55 of 65 of the dEQP-EGL.functional.choose_config.* tests
to start passing.

Bug: angleproject:3172
Change-Id: I287af5ba7d296694d9a78ded5d1e3bc4e7043d03
Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/1506696
Commit-Queue: Ian Elliott <ianelliott@google.com>
Reviewed-by: Yuly Novikov <ynovikov@chromium.org>