Change-Id: I97e65fcfdbab5e8ebf823ef528be3a048274a9ef
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36809Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Antonio Maiorano <amaiorano@google.com>

Bug: b/141758871
Change-Id: I8e21cb19b1cecbd996ed78cbe9fbde46beac0adc
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36769
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Antonio Maiorano <amaiorano@google.com>

ANGLE reserves 2 uniforms and 2 varyings for GLES.
Increasing maxPerStageDescriptorUniformBuffers by 2 and
maxVertexOutputComponents by 8 (2 * 4 components) allows
the following tests to pass (17 failures fixed in total):
dEQP-GLES3.functional.implementation_limits.*
dEQP-GLES3.functional.state_query.*

Bug b/141632907

Change-Id: Ie39b735286d6682f5782fca30e0f7a079f4142ec
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36668
Presubmit-Ready: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

This extension only actually adds an enum. The interesting behavior will
come with image and buffer memory barriers describing an ownership
transition between the real queue family and
VK_EXT_QUEUE_FAMILY_FOREIGN. Depending on the platform, this may be
nothing; may require cache maintenance operations; or may require
a copy to/from an external buffer.

Bug: b/141697747
Change-Id: Ia5b6d3027f21bb7c61143972fcb4d4cd8bc06f11
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36728Tested-by: Chris Forbes <chrisforbes@google.com>
Presubmit-Ready: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

True by default, allows forcing non-optimized builds of SwiftShader in
debug builds.

Bug: b/140251624
Change-Id: I51bb93b53ead5936b47894ebabbd73b36c05cae0
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36470Tested-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Change-Id: Ie2c008b40b83fa2d04f3008a5fcb9b554513639b
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36116Tested-by: Chris Forbes <chrisforbes@google.com>
Presubmit-Ready: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Change-Id: I74e4615e07478b8d0f54432762a7e630004c3aa1
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36628
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>

* changes:
  Update Marl to 12872a0df
  Squashed 'third_party/marl/' changes from 5f31df137..12872a0df

As documented at https://developer.apple.com/documentation/quartzcore/cametallayer/1478168-framebufferonly:
"If the value is true (the default), the CAMetalLayer class allocates
its MTLTexture objects in ways that are optimized for display purposes.
However, you may not sample, read from, or write to those textures. To
support sampling and pixel read/write operations on the layer’s textures
(at a cost to performance), set this value to false."

Since we call replaceRegion in MacOSSurfaceMVK::present() to write pixels
from system memory, we need to set this flag to false explicitly.

Bug: b/137673628
Change-Id: I83e28a2940261d9eedf30d7c96427c23d85ac639
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36608Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

SwiftShader has been confirmed by Khronos as a conformant Vulkan 1.1 product:
https://www.khronos.org/conformance/adopters/conformant-products#submission_400

Bug: b/129792032
Change-Id: I6893a411fbb2b3ba94ec07674bd78ca81b319680
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36589
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>

This helps applications identify the driver.

Bug: b/129792032
Test: dEQP-VK.api.driver_properties.properties
Change-Id: I300fcc5dda6886b422fb64902b3eadf070b5eea8
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/22690
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

This reverts commit 1f89f58f.

Original cl:
https://swiftshader-review.googlesource.com/c/SwiftShader/+/36548

Reason for revert: Causes timeout flakes on Regres

Bug: b/141380274

Change-Id: I184e96c586e449b7dcd5b445078064cbafb7df72
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36592Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

12872a0df Comments: Improve & make style consistent
5bb61cf6a Scheduler: Ensure all scheduled tasks are finished when the scheduler is destructed.
b5e3e1e00 src\scheduler_test.cpp: Reduce max number of in-flight tasks.
d13286050 src\osfiber_windows.h: Assert that the Win32 fiber calls succeeded.
08da5139c .gitignore - Add Visual Studio generated files.
8007ea5de Temporarily disable DestructWithPendingFibers test.
5390897be Fix alignment of Pool items.
89425dcbc CMakeLists: Bump MSVC warnings to level 4
6df4597c3 Kokoro: Shuffle presubmit file layouts.
e3b3c7df4 Add include/marl/memory.h for aligned memory allocations.
60598ef45 Fix TSAN issue with BlockingCall test.
be3628456 CMakeLists: Add TSAN build flag

Commands:
git subtree pull --prefix third_party/marl https://github.com/google/marl master --squash

Bug: b/141380274
Change-Id: Ic33110f917737d0d84d882f63a4e1c0a87a16c33

12872a0df Comments: Improve & make style consistent
5bb61cf6a Scheduler: Ensure all scheduled tasks are finished when the scheduler is destructed.
b5e3e1e00 src\scheduler_test.cpp: Reduce max number of in-flight tasks.
d13286050 src\osfiber_windows.h: Assert that the Win32 fiber calls succeeded.
08da5139c .gitignore - Add Visual Studio generated files.
8007ea5de Temporarily disable DestructWithPendingFibers test.
5390897be Fix alignment of Pool items.
89425dcbc CMakeLists: Bump MSVC warnings to level 4
6df4597c3 Kokoro: Shuffle presubmit file layouts.
e3b3c7df4 Add include/marl/memory.h for aligned memory allocations.
60598ef45 Fix TSAN issue with BlockingCall test.
be3628456 CMakeLists: Add TSAN build flag

git-subtree-dir: third_party/marl
git-subtree-split: 12872a0dfe717ab75a7cf8e10621d723b8a1a382

Change-Id: I98428c5f15232f9f7347642cc42f9da5c579b86c
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36508
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>

This is causing a deadlock on Windows 7. The process teardown is
calling a destructor function that ends up calling a bunch of
synchronization logic that should not be called on process exit.

Bug: b/141380274
Bug: angleproject:3909
Change-Id: I4787d495d965e8b43d15be9c952c4b1de9bdf197
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36548
Presubmit-Ready: Jamie Madill <jmadill@chromium.org>
Tested-by: Jamie Madill <jmadill@chromium.org>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

The libswiftshader_common_debug target adds debug.cpp which uses
utils/String8.h from libutils.

Bug: b/138239233
Change-Id: I7b078e47e56e02d12e44e21a42679a4cd4b6b6df
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36528Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Jason Macnak <natsu@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

* changes:
  Update Marl to 5f31df137
  Squashed 'third_party/marl/' changes from bf3e23083..5f31df137

ANGLE needs to be able to use the depth buffer in a sampled
texture to emulate some GLES2 features, even when the image
also contains a stencil. Added support for sampling the
depth+stencil format we support.

Tests: dEQP-VK.*

Bug b/141304420

Change-Id: Ia4ccce2a155aa16178f16eb92efae9fc357b6e45
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36449
Presubmit-Ready: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Fixing this issue required using the WrapOpKill SPIRV-Tools
pass. This pass wraps OpKill instructions inside a function.
See the following for the explanation:
https://github.com/KhronosGroup/SPIRV-Tools/issues/2726

dEQP-GLES2.functional.shaders.discard.function_always fails
on ANGLE on top of SwiftShader Vulkan without this fix.

Bug b/141246700

Change-Id: I48ce3c19a1b02160fd8f23b60a75e5f8f35a4d37
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36450
Presubmit-Ready: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

5f31df137 aarch64: Fix saving/restoring register r29 on fiber switch
76f093bce Kokoro:Add bazel presubmits
ea671dced Kokoro: Split build rules by build system
767915fdb Scheduler: Don't work-steal on the single threaded workers.
4fbdf5946 Bazel build support (#28)

Bug: b/140967243
Change-Id: Ia7ee9d37f83313e8524fda9757fd23b487f937be

5f31df137 aarch64: Fix saving/restoring register r29 on fiber switch
76f093bce Kokoro:Add bazel presubmits
ea671dced Kokoro: Split build rules by build system
767915fdb Scheduler: Don't work-steal on the single threaded workers.
4fbdf5946 Bazel build support (#28)

git-subtree-dir: third_party/marl
git-subtree-split: 5f31df137868d306ecfb4c908cc6c5a725aeda63

* changes:
  Update Marl to bf3e23083
  Squashed 'third_party/marl/' changes from 59068ee4c..bf3e23083

Example:
ComputeProgramKey a, b;
a.shader = 1;
b.shader = 2;

a.layout = 2;
b.layout = 1;

// This assertion will fail because a < b since a.shader < b.shader,
// but also b < a because b.layout < a.layout.
assert( (a < b) && !(b < a) );

Change-Id: I1061f517cee60a28a1f4239704655e9d320ade46
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36348
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

This allows the TSAN/ASAN builds to properly locate the library.
Makes the fuchsia workaround apply everywhere.

Bug: chromium:1000668
Bug: angleproject:3876
Change-Id: Ide2e8527c89c614a648204e953def1f24b702a62
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36468
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Jamie Madill <jmadill@chromium.org>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

dEQP doesn't try to render this format with blending enabled, so we
missed the attachment read path.

Change-Id: Ib1decc0dbdf1e9895b5ac00b478a65fd717e534c
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36488Tested-by: Chris Forbes <chrisforbes@google.com>
Presubmit-Ready: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Several weeks ago I did some build system rationalization which subtly
broke cuttlefish. Unfortunately nothing screamed very loudly so we
didn't notice.

- Place the libraries in the correct directory
- Use libarect (static lib for this purpose) rather than libui (which
is inaccessible to sphal even though the build system will allow it)

Bug: b/141209513
Change-Id: I4e9010bfb83686a876968daab8e99df736d4b1e8
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36451Tested-by: Chris Forbes <chrisforbes@google.com>
Presubmit-Ready: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>

Change-Id: Ifc8cf06bcc18d197fd45bb08779342ddc7f20265
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36448
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

Bug b/140294254

Test: dEQP-VK.binding_model.*
Test: dEQP-VK.glsl.atomic_operations.*
Test: dEQP-VK.robustness.vertex_access.*
Test: dEQP-VK.synchronization.op.single_queue.*

Change-Id: I5f3cd246d93b6550a2ad982f1c838ad9c02d3858
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35789Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Robust buffer access for vertex inputs/attributes
was not implemented. It's fixed in this cl by adding
robustBufferAccess to the Context structure, adding a
check in Renderer::advanceInstanceAttributes() to prevent
advancing a buffer past its end and adding out of bounds
checks in VertexRoutine::readStream()

This affects dEQP-VK.robustness.vertex_access.* tests,
which will be enabled when the vertexPipelineStoresAndAtomics
feature is turned on.

For now, this change should be noop in test results.

Bug b/140294254 b/131224163

Change-Id: Ib4d4cdb73f48495a556b6bd05b1d0e3ce800ad15
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36268Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Extra vertices are added at the end of the vertex list when
processing a batch in order to avoid SIMD width overrun.
A single comparison can prevent us from processing these
vertices by disabling them in the stores and atomics mask.

This would fix the following tests:
dEQP-VK.synchronization.op.single_queue.*
when enabling the vertexPipelineStoresAndAtomics feature.

Note that these tests are affected because they wrongly
assume vertices won't be processed more than once. These
tests should still get fixed.

Bug b/140294254

Change-Id: I04185b899a9770537c3d10bcfd87e00e314582de
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36368
Presubmit-Ready: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>

Instead of packing each point in a triangle primitive,
tightly pack all point vertices and write them out 3
at a time in the output primitive. This should be
roughly twice as fast. Explanation:

Currently:
Vertices: 0 0 0 1 1 1 2 2 2 ...
Processing:
1) 0 0 0 1
2) 0 (cache hit)
3) 0 (cache hit)
2) 1 (cache hit)
3) 1 (cache hit)
4) 1 (cache hit)
5) 2 2 2 3
...
-> We processed 8 vertices to get points 0 1 2 3

New way:
1) 0 1 2 3 -> We processed 4 vertices to get points 0 1 2 3
2) 4 5 6 7

Will affect these tests once vertexPipelineStoresAndAtomics
is enabled:
dEQP-VK.glsl.atomic_operations.*

Note that these tests are affected because they wrongly assume
vertices won't be processed more than once. These tests should
still get fixed.

Bug b/140294254

Change-Id: Idb21085838317db7b7a6630a18de4d7284534429
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36349
Presubmit-Ready: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>

Change-Id: I65ed5dc54a648a148c1cf95546902e64df629bca
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36128
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>

While blitting, it's possible to generate a float that is outside the
representable values of an integer. Since converting such a float
results in undefined behavior, we instead throw an invalid operation
error and return without continuing the blit.

Bug chromium:1001874

Change-Id: Ic6938adf75176e34021d0ca1404176e4979a3ca6
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36408Reviewed-by: Alexis Hétu <sugoi@google.com>
Tested-by: Sean Risser <srisser@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

bf3e23083 pool.h: Don't over-qualify Loan type parameters
dc76e9653 CMake: Fix configuration file installation path
fca4771b4 Scheduler: Increase max number of threads from 64 to 256.
1b3fd77b2 debug: Add MARL_WARN for printing warnings
3928f44d6 [Grammar] Fix passive tense.
a2918916e CMake: Add install target (#21)
1ec7c36cf examples/primes.cpp: Add missing header including
ed21a07a1 Minor cleanup

Commands:
git subtree pull --prefix third_party/marl https://github.com/google/marl bf3e23083979c3bd3de1c77346b655eec423b3bc --squash

Bug: b/141003126
Change-Id: I5e9afad957d52919c48a60d8db49c04e43102870

bf3e23083 pool.h: Don't over-qualify Loan type parameters
dc76e9653 CMake: Fix configuration file installation path
fca4771b4 Scheduler: Increase max number of threads from 64 to 256.
1b3fd77b2 debug: Add MARL_WARN for printing warnings
3928f44d6 [Grammar] Fix passive tense.
a2918916e CMake: Add install target (#21)
1ec7c36cf examples/primes.cpp: Add missing header including
ed21a07a1 Minor cleanup

git-subtree-dir: third_party/marl
git-subtree-split: bf3e23083979c3bd3de1c77346b655eec423b3bc

The Renderer needs to be aligned when allocated, otherwise
it leads to all sorts of hard to track issues in the 32 bit
version of the library.

In this instance, Angle's end to end tests on SwiftShader
Vulkan had spurious failures due to having unaligned members
in PixelProcessor, which wasn't aligned because it's used as
a base class of the Renderer class, which wasn't aligned when
allocated. Using an allocator to align the memory fixes the
issue.

I also did a small cleanup of PixelProcessor::setBlendConstant,
as it was a bit hard to debug and changing it this way made it
easier to track what was going on.

Bug b/140888415

Change-Id: Ie5559642456ee4dcb73f9d07c777349f558879ba
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36248
Presubmit-Ready: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

A previous CL broke the Android build due to the fact that
sw::trace() is only defined by System/Debug.hpp on non-Android
platforms (while DebugAndroid.hpp defines trace() instead).

Use the TRACE() macro instead of calling the function directly.

Bug: b/73656151
Change-Id: I85e9f592a43df8bc4dc719b995934bb9c3bfa97f
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36329Tested-by: David Turner <digit@google.com>
Kokoro-Presubmit: David Turner <digit@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

src/System/Timer.cpp which was duplicated for Vulkan uses x86 intrinsic __rdtsc
to tick count. This intrinsic is not available on ARM and needs to be fixed.
This issue broke Windows ARM64 build bot and the same issue was fixed for
src/Common/Timer.cpp before (see below links).

https://ci.chromium.org/p/chromium/builders/ci/win32-arm64-rel/7763
https://swiftshader-review.googlesource.com/c/SwiftShader/+/23508

Bug: chromium:893460
Change-Id: I48e3673034650a89fead71447d75f4e486ee68b5
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36308
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>

This creates a new helper class to allocate memfd-backed shared
memory regions with the Linux kernel. The feature has been
supported since at least Linux 3.17, which is sufficient for
a large number of desktop Linux and Android devices.

However, only recent version of GLibc or the Android C library
expose memfd_create(), so keep implementing it as a direct syscall.

NOTE: There is a similar function named memfd_create() inside
      of src/Reactor/ExecutableMemory.cpp, but it is easier to
      keep them separate for now to avoid introducing subtle
      dependency issues in the CMakeLists.txt file.

This will used in future CLs that implement Vulkan external
semaphores and external memory on Linux and Android for SwiftShader.

Bug: b/140419396

Change-Id: I378c34760930d8fef7d0659128ededb5227c323f
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35128Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: David Turner <digit@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>