Nothing is actually scheduled, but is a first step towards using yarn.

As yarn is now actually used by SwiftShader, drop the BUILD_YARN build
flags, as failing to build yarn would result in linker errors.

Bug: b/139142453
Change-Id: Ibd9a69f72a248f58a62bb41eeb196c4647876e82
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35153Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

They're big and take a while to initialize. Only construct them if we need them.
This also pushes the construction onto the queue, speeding up the main thread.

Bug: b/139010488
Change-Id: I2d1d25f24162339de62cea4d946532b1a2e346ca
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/34330Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>

TIL: For POD type T, `new T` and `new T()` are not the same.
new T() will zero initialize, which for large arrays can be very costly.

Bug: b/139142453
Change-Id: I9a1a3ef1bb0a167d9955310d1cf331e917019397
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35569
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

There is no reason to plumb this through Context.

Bug: b/132280877
Change-Id: I92819af587505b75b279a6aed70aa8067455df8d
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35549Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>

We only support these as far as is required for Vulkan 1.1 core
functionality. If we expose the extension bits, we have to support
more.

Change-Id: I2a8b1648e88fb55b4c87075efc696792540144fc
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35548Tested-by: Chris Forbes <chrisforbes@google.com>
Presubmit-Ready: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

Bug: b/119621736
Change-Id: I8a25e6a5dba7e05649c35dc65160f69300bf5791
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35508Tested-by: Chris Forbes <chrisforbes@google.com>
Presubmit-Ready: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Bug: b/119621736
Change-Id: I9be8dc680f68622e795fccf4ff8213dccd5debb2
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35488Tested-by: Chris Forbes <chrisforbes@google.com>
Presubmit-Ready: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

This function works with VkFormat, not spv::ImageFormat (due to input
attachments)

Change-Id: I36a36a94127533a009df8f806b55ff2845162a76
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35528Tested-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Change-Id: Iba8f2c6f469c7c34dfaa25811c6bcc8f784b600d
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35448Reviewed-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

A single RenderPass object may be used to record multiple command
buffers simultaneously. Storing the subpass index there produces
crosstalk between the command buffer recordings, and incorrect
rendering.

Bug: b/139824232
Change-Id: Ib8689a47733a991c61785d2d6332584f7a96f05d
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35468
Presubmit-Ready: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Change-Id: I8189bb729fb5843979bc719638d25474c2d69308
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35428
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>

Section 15.2.1. RGB to Shared Exponent Conversion of the Vulkan 1.1 spec
states "NaN, if supported, is handled as in IEEE 754-2008 minNum() and
maxNum(). That is the result is a NaN is mapped to zero."

This currently does not affect anything because E5B9G9R9 is not a
renderable format. The blitter merely has to support writing this format
for computing the corner texel of cube map borders, which are averaged
from other E5B9G9R9 format texels and thus no NaN can be encountered.

Bug: b/138944025
Change-Id: Ib69aca3af5afee240d9b697c47b2f1a0204fe697
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35008
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Sean Risser <srisser@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

The Vulkan 1.1 spec states that:
"For VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, VK_DESCRIPTOR_TYPE_STORAGE_IMAGE,
or VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT, only the imageView and
imageLayout members of each element of VkWriteDescriptorSet::pImageInfo
are accessed."

Also remove the zeroing of sampler descriptor data in
SampledImageDescriptor when no sampler handle is provided. This was
intended to detect bugs but we didn't do it consistently for all
descriptor types (VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER also uses
SampledImageDescriptor and has no sampler data). We now pass a null
pointer to the sampling routine when no sampler descriptor is available
(specifically for OpImageFetch) so it is less likely to still access it
unintentionally.

Bug: b/139401791
Change-Id: Id1394493e88b7465fe3628c4c7d410cd3eaa3ccf
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35388
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>

OpImageFetch takes an Image operand, not a SampledImage, and thus we
don't explicitly have a sampler descriptor. However, our
vk::SampledImageDescriptor structure includes both image view descriptor
data and sampler data, and we were relying on the sampler descriptor
data to be present for setting the addressing mode. Other state would
also inadvertently be affected.

This change ensures that the sampler descriptor for fetch operations is
null and we set the addressing mode to repeat but leave other unused
state at their defaults.

Bug: b/139401791
Change-Id: I9f1af35940c2fa9c7d30e771ebdb244f72f5bdbe
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35348
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 is a mandatory feature for 1.1, but Vulkan CTS 1.1.3 doesn't
enforce that -- it only tests the feature if it is present.

Highlights:
- Multiple views implemented by running each draw multiple times. We
  could do something more efficient, but this is a fringe feature so
  far.
- Render targets and input attachments are adjusted to use the layer
  corresponding to the current view.
- Explicit attachment clears and end-of-subpass resolves are broadcast
  to the layers corresponding to the current subpass's view mask.
- Renderpass attachment load ops are executed for the layers corresponding
  to the union of the subpass view masks for all subpasses which use the
  attachment. The actual load ops are still performed together at the
  beginning of the renderpass.
- ViewIndex builtin variable is exposed to the shaders. In a
  non-multiview draw call, ViewIndex is still available, but is always
  zero.

Bug: b/139862810
Test: dEQP-VK.*multiview*
Change-Id: Iaf40cfdb2f5afa61253cc756f97c0db30fb4d813
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35408
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>

Change-Id: I8090e55dae72f1b48b46147bad0439d22b513b74
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35409Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>

Chrome has started requesting RGB formats for IOSurfaces to support the alpha=false
flag for WebGL context creation.

BUG=chromium:995142

Change-Id: I0c9d994f2d8208e87a0d167ebc0787fc15add08f
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35389Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Geoff Lang <geofflang@chromium.org>

This is Vulkan 1.1's "preferred method of line rasterization, and must
be used when the implementation advertises the strictLines limit in
VkPhysicalDeviceLimits as VK_TRUE."

The 'Diamond test convention' code path, which implemented coverage
corresponding to the Bresenham algorithm (except for endpoints, see
b/80135519), is now inactive, but might become of use again for
implementing VK_LINE_RASTERIZATION_MODE_BRESENHAM_EXT so it is kept,
for now.

Bug: b/139755234
Tests: dEQP-VK.*line*
Change-Id: I0113c04e35c1aca8031194223e589962942e18fa
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35308Reviewed-by: Chris Forbes <chrisforbes@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

This ports the Vulkan side change to OpenGL ES and silences the GCC 8
class-memaccess warning/error, while ensuring proper key initialization.

Defines a Memset<T> class to be used as the first base class of cache
key types, which makes it explicit that their underlying memory will be
fully initialized before any member constructors are run.

In particular this fixes Blitter::Options state having uninitialized
bits after the bitfield, and Blitter::State having uninitialized padding
bytes after Options so 'sourceFormat' is 32-bit aligned.

Also adds is_memcmparable<T> for checking if memcmp() can be used to
implement operator==() for cache keys. It's equivalent to
std::is_trivially_copyable except it provides a fallback for STL
implementations that don't support it.

Also fix class-memset violations in LLVM 7.0 with their solution from
a later version.

Bug: b/135744933
Change-Id: Ic1e5c2c6b944a5133fc55840c0508bc2cdd1d66a
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/33488
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>

Bug: 139469122
Change-Id: I0f319952b00fcca4165bf3f59221ebaf1ba12045
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35189Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

vk::warn etc expects the client to include the newline
Change-Id: Ib0f808d7d397c37e8c52d7ce1977551a206dbc40
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35251Reviewed-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Newer Vulkan CTS has started providing this extension structure during
device creation.

Bug: b/139528538
Change-Id: Ie5c72dcc0b96b4d9e98a06a36fdecbdb25304e49
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35250Tested-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Newer Vulkan CTS has started providing this extension structure during
device creation.

Bug: b/139528538
Change-Id: Ied17274d95acb8be1766f62e21fcc1c6b7594015
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35249Tested-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Now that the GLES unit tests can cause the float calculations in
these functions to result in inf and NaN, some platforms are throwing
errors when those floats are converted to integers. The Clip*Rect
functions no longer continue processing if any operation results in a
non-finite value, and tell their callers that they have failed to clip
properly.

Make Context.cpp set the error state to GL_INVALID_OPERATION when it
fails to blit.

Bug chromium:979986

Change-Id: I8e11e77b1e25eab37dee97aa616a908737b16036
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35269
Presubmit-Ready: Sean Risser <srisser@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Sean Risser <srisser@google.com>

Bug: b/139010488
Change-Id: I6a152c3df27e89c38ee89d57270b604bbe6d5acf
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35212Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Chris pointed out that ".S" assembler files hint that they should be preprocessed.
Now these files have the correct extension, we don't need to be explicit about running assembly files through the preprocessor.

Bug: b/139010488
Change-Id: I1399e884071c7fc2f169e93809a9f15204e41eff
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35211Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

This can be #defined in the assembly files, reducing per-build-system complexity.

Bug: b/139010488
Change-Id: Ided03000dc5e93e4d1e04ce6f0b57c6e45565545
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35210
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

This is backed by a regular heap allocation, so we can read and write to it.

Bug: b/139479506
Change-Id: I4a5ef04386120667a27c817c4ac0b59b8b44cd0b
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35030Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Change-Id: I7ae85ca4b77ad9d90fb2f30230f1b74e5b1f665a
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35248Reviewed-by: Ben Clayton <bclayton@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>

LLVM 7.0 properly supports ARM (unlike the LLVM 3.0 version we used
before), including the features needed by our Vulkan implementation
(unlike Subzero).

Bug: b/115344057
Change-Id: Ic9155b644ab649597219093c3d757862c60a9fa0
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35228
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>

Pastel proved that the LLVM path works well for all the architectures we
care about for Android. Converted all aspects of the build to Android.bp

Bug: b/139147241
Change-Id: I8ad8c2cc816fbb313feaf4a0c4d3dcf40cff37af
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35168
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>

- Had to rename .c of .c/.s pair to avoid Android's build system
  producing name clashes
- Rename .s to .S to get assembler-with-cpp behavior

Bug: b/139010488
Change-Id: Ib5ee9355e60fe52b1caedfc6ec1b456296d519f5
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35188Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>

If the merge block M contains phis, and M is a loop header of a second loop, we will get assertions that the back edges of the second loop have not yet been generated.

We're only trying to update the phis of M for the lane exits of the first loop, so filter the phi updates to edges from those blocks.

Bug: b/139424793
Change-Id: Idc197fb93c28cdc4b91174a687179f9fc498344a
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35208
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

vk::trace is pretty much a no-op, and hides warnings and errors.
If we do get an error, spirv-opt tends to eat the SPIR-V, leading to a cryptic 'ASSERT(code.size() > 0)' crash.

Bug: b/139424793
Change-Id: Ie6f3819265ad0e32e85d4c6b1824d263d9fde02f
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35209
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

Change-Id: Ie3fdd1532a899562736b0b9497a7e6ca588ca4e2
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35154
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

If glBlitFramebuffer tries to blit massive rectangles, it can result in
NaNs. This change ignores those rectangles instead of trying to blit
them.

Added a unittest for blitting a framebuffer.

Also fixed a typo in a debug message.

Bug chromium:979986

Change-Id: Ifd8e379f27b6882044bb93c6a68647b8474e7afc
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/34428
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Presubmit-Ready: Sean Risser <srisser@google.com>
Tested-by: Sean Risser <srisser@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

This adds a new header <vulkan/vulkan_fuchsia_extras.h> which
provides Fuchsia-specific definitions that are not part of the
upstream Khronos headers yet, and are still only available from
the Fuchsia platform tree.

Putting these into a separate header should make comparisons
and updates to the rest of the Khronos headers much easier in
the future. Once everything has been upstreamed properly, the
extra header will be removed.

NOTE: vulkan.hpp is not updated because doing so is tricky and this
      header is not included by any SwiftShader code.

Change-Id: Ifac0f4fde305156dfe6b349935bab321b2f21622
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/35130Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: David Turner <digit@google.com>

Currently only implements the parallelize() function - useful for evenly distributing workloads across threads.

Bug: b/139010488
Change-Id: I7c5705a46afadf137a0cf7fd517b61027ac3a239
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/34819Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Ticket is a synchronization primitive used to serially order execution.

Bug: b/139010488
Change-Id: I7f34556609b0206c87f57be9c6bc14b21096f8f4
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/34818Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

BoundedPool<> is a fixed capacity pool which blocks when attempting to borrow an item when the pool is empty.

UnboundedPool<> is a automatically growing pool that never blocks.

Useful for building data pipelines.

Bug: b/139010488
Change-Id: Ie1ceec970a866d9d0107247ad2baaced7f024973
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/34817Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>