Regress is a tool that detects test regressions with SwiftShader changes.

Regres monitors changes that have been put up for review with Gerrit.
Once a new patchset has been found, regres will checkout, build and test the change against the parent changelist. Any differences in results are reported as a review comment on the change.

Change-Id: I0a848a56d99f020d6ffd48699a2277fa71526788
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/24929Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

At least for large formats, we assume 16 byte alignment for render
target writes. Requiring 16 bytes in general seems a good idea for
x86_64.

Bug: b/127721996
Change-Id: Ie4d2d9bf71fb8a3d8b58f099785e75a72ff0e181
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26448Tested-by: Chris Forbes <chrisforbes@google.com>
Presubmit-Ready: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Change-Id: I5c4e28ccd80163c03a7172b8db881ea28e9e0203
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26350Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>

This was just calling the wrong function.
Looks like a copy and paste bug.

Change-Id: Icde43940f17f467f7602b7a2ce0e5ac1e517c769
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26208Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Allows expressions like:

Array<Float4> arr;
Pointer<Float4> ptr = &arr[0];

Bug: b/126330097
Change-Id: I4ad95b46cfa1332eded0424ba4aeb7fe8642b71a
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/25709Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Some dEQP tests were not made for SPV_ENV_VULKAN_1_1, and error in the optimizer.
This produces 0 words of SPIR-V, that then goes and explodes in SpirvShader.

Add an assert to catch the 0-word case.

Bug: b/127454276
Change-Id: I60576e6691a9cf74656dfcd9c53aeed7ab578de1
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26188Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

MSVC does not like: sizeof(T[N])

Change-Id: If7586a0d29f609372e036fdb31481e923469a7ad
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26349Reviewed-by: Alexis Hétu <sugoi@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

- Fix broken use of VkImageCreateFlags in various places as an aspect mask
- Be consistent about layout of D+S images. Layout is now [aspect][layer][level].
- Allow fetching an offset into a particular aspect.

Fixes dEQP-VK.pipeline.depth.*

Bug: b/118619338
Change-Id: I46adc9c637882e7144945eaeacce9f087d53caf0
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26011Tested-by: Chris Forbes <chrisforbes@google.com>
Presubmit-Ready: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

tests/regres/full-tests.json contains a full test list that will be run
once every day. The results of this full test pass will update the
test lists binned by status (*-PASS.txt, *-FAIL.txt, etc).

tests/regres/ci-tests.json references just the *.PASS.txt lists, and
will be run for every change up for review.

The goal of this refactor is to significantly speed up the regres test
cycles. The side effect of this is that regres will no longer
automatically tell you when a single change passes new tests. It will
however tell you when you break tests. The daily test will show you new
tests that are passing, and gives a simple way to see all tests by
status.

Bug: b/127664487
Change-Id: I590378e0bf819211dbc4de85f63ce77a681f6e46
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26328Reviewed-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

Bug: b/126873455
Change-Id: I5df5206cb3743dd7b36d4ad521f9e9ddf81ae638
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26168
Presubmit-Ready: Chris Forbes <chrisforbes@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Bug: b/127282157
Change-Id: Id2c5b29d744f6f7ddcbb853db0a95228425d7f83
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26108Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

This initializes arrays to hold the descriptor sets in the routine. Nothing uses this yet.

Bug: b/126330097
Change-Id: If052d0b93e62e4f32e88ed02f9bc21f4203587f5
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/25553Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

We don't support any sparse features, so we can always just produce zero
properties structures here. The VK1.1/GPDP2 equivalent already forwards
to this function, so we also get that for free.

Bug: b/118429780
Change-Id: I7b43a291354f1ca2ae01d021dfa48263b08b57cc
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26308Tested-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Change-Id: I39608947651b855c6e2668de98ab5696e18f501f
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26310Tested-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

When the build fails, it's very hard to find the compiler error in the spam.

Change-Id: Ic975f0a075e4104db82a58ef03e1ba643e520fe2
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26031
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Replace the runtime `warning` with something that will actually crash
us, and add stub implementations of OpLabel and OpReturn to allow
trivial shaders to work.

There's little point in continuing beyond an unimplemented instruction
-- if you get lucky, you hit a crash on use of the instruction's result.
If you get unlucky, you wander off into undefined behavior.

Change-Id: I3b222ea74446754276096c81fc3669c311872316
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26088Tested-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Bug: b/127282157
Change-Id: Icaec924ef011b42069d157bec2d76ae5df3eea46
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26048Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

This commit adds `-Wno-implicit-fallthrough` to disable implicit
fallthrough warnings.  This commit is required because the latest
LLVM/Clang toolchain emits more warnings.  Since `third_party/llvm-7.0`
is an upstream project, we don't want to locally patch the code.  Thus,
we are disabling implicit fallthrough warnings before it has been fixed
in the upstream.

Bug: b/115344057
Test: Build libGLESv2_swiftshader for Android
Change-Id: Ibde53e030ba66af20ba5281b33a684350ef4b35f
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/25529
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Logan Chien <loganchien@google.com>

`d` is used to drive some indexing math later.

Bug: b/127343247
Change-Id: Id9da208f5a7b678748f129f97de03f5bf04bcbe1
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26128Tested-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

dEQP tests passing:
dEQP-VK.wsi.xlib.swapchain.get_images

Bug: b/124265819
Change-Id: I4e2dc8dd48b93bbddba44654c72afb8f43bc3561
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/25789Tested-by: Hernan Liatis <hliatis@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Change-Id: I30738f487bfc1ac57ae3ef6265c013bb2865417f
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/25210
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Tested-by: Raymond Chiu <chiur@google.com>

- Fix mapping of B8G8R8A8 -- this is the same as the old D3DCOLOR.
- Remove all the other stream types that don't match anything in Vulkan.
- Remove unused resource pointer in stream

There is still future cleanup work to do here -- we should be able to
do vertex fetch purely based on the VkFormat rather than mapping it onto
the GL model.

Change-Id: Ia100ec68e8930acd17b6cba35a27403b4ef8e883
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/25768
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Bare bones implementation of a creatable and
destroyable swapchain

Bug: b/124265819
Change-Id: Ie8277184863ab6b7204b6c8f6fc2b2f86ad787c9
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/25509
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Hernan Liatis <hliatis@google.com>

- OpSConvert, OpUConvert, OpFConvert have no valid use, as we only have
  one value width (32 bits) and these instructions perform width
  conversions only.
- OpConvertFToU, OpConvertFToS, OpConvertUToF, OpConvertSToF implemented
- OpBitcast implemented. Note that the spec looks scary wrt pointer
  types in OpBitcast, but 2.16.1 Universal Validation Rules disallows
  this use in the Logical addressing model.

Bug: b/126952020
Change-Id: I6c1c95d5ad4e19177e40ead7713bf63ffe16c679
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26010Tested-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Bug: b/126873455
Change-Id: Ibe76dc02ab22903f1ad2d00685cd1855aaeb3338
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/25968Tested-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

- Make the existing LLVMReactor lowering support for MulHigh on non-x86
  available on x86 as well, as we don't have good intrinsics-based implementation
  of 4x 32bit mul highs. At some point in the future we can rework this
  to use some shuffles and a pair of pmuludq.
- Plumb through Int4 and UInt4 variants of MulHigh
- Implement SPIRV OpUMulExtended, OpSMulExtended in terms of MulHigh

Bug: b/126873455
Change-Id: I25ba0a69691e7a6f7a5542ec4a90a44ba8f68331
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/25929Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>

Memset the scalars to 0, and check their pointers before use. Catches cases where the intermediate was declared but not set.

Also switch from using assert() to ASSERT() as the latter still fires with DCHECK_ALWAYS_ON.

Change-Id: I81abb50aea41568f95c22a340b90b7c56839d3ce
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/25869Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

Previously we had generated this via deqp-vk --deqp-runmode=stdout-caselist
+ a sed one-liner to strip off the prefix. Unfortunately, that's not quite
good enough -- that output contains a `GROUP: ` line for every non-leaf in
the the test tree. deqp would then produce log spam when failing to find
those nonexistent tests.

New test list generated by also filtering those `GROUP: ` lines.

Change-Id: If3c97724988f94e2e127e9e091960e1f51e3599a
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26068Tested-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Correct the sign of the area calculation. The front facing determination
can then be written to exactly match the spec -- CCW means positive area
is front facing.

Change-Id: Ibed3e26b0cf77030325044c7dc9e818264a8750d
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/26009Tested-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Our command buffer playback mechanism is safe for multiple inflight
submissions of the same command buffer, so there is nothing special to
do here.

The problematic case of needing to carry some state from a primary
command buffer into a secondary is already caught by the slightly later
check for pInheritanceInfo.

Bug: b/118619338
Change-Id: I782285fbc9127d59eeac7cbf79fea3f1952e3cde
Reviewed-on: https://swiftshader-review.googlesource.com/c/25950Tested-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Hernan Liatis <hliatis@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Bug swiftshader:124

Change-Id: Ied13e859cc8984e59af393de6ca1a93428a0e30e
Reviewed-on: https://swiftshader-review.googlesource.com/c/25949Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>

This cl adds proper storage and update of descriptor sets.
The added functionality includes:
- The descriptor pool now allocates the proper larger required
  storage needed to actually store descriptor set data.
- Descriptor sets, when allocated, also get initialized with some
  basic header data (set layout) and also some Descriptor data
  when available (a.k.a: immutable samplers)
- Descriptors are currently bindless, since it is simpler as a first
  implementation, but can easily be modified, which is intended to
  be done in the near future. For now, each descriptor set is either
  a VkDescriptorImageInfo, a VkDescriptorBufferInfo or a VkBufferView
- Descriptors can be updated from either a VkWriteDescriptorSet or a
  VkCopyDescriptorSet structure. The update supports writing to
  multiple descriptor sets in a single operation and supports array
  sizes mismatch properly according to the spec

Bug b/123244275

Change-Id: I1e0430e0014e26a304632a4b2b10ad0f69b06180
Reviewed-on: https://swiftshader-review.googlesource.com/c/24910Tested-by: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Reviewed-by: Corentin Wallez <cwallez@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

This change requires include directives to include the Vulkan/ path so
it's clearer when we have dependencies across layers/folders.

This was already the case for the other build files, so it avoids
build breakage as well.

Bug b/126557661

Change-Id: I4d8773e8c28e6677fd1e93ca9c9d2d843431343c
Reviewed-on: https://swiftshader-review.googlesource.com/c/25948Reviewed-by: Alexis Hétu <sugoi@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Far from a complete set, but both Chris and I are looking at tests that require these opcodes.

Bug: b/126870789
Bug: b/126873455
Change-Id: Idb2109cfc3352da83aa38d42eed5d15537dfb6b7
Reviewed-on: https://swiftshader-review.googlesource.com/c/25868Tested-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: 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>

I incorrectly (9a16248d) added an UNIMPLEMENTED to variables that have a storage type that is neither Input or Output.  However, other storage types do actually work.

Revert that change.

Change-Id: I5ec36e9d6d536dcdd085c93e9f00782310e118d4
Reviewed-on: https://swiftshader-review.googlesource.com/c/25848Reviewed-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

Bug: b/124265819
Change-Id: I19bbaeefe19ade1afeb6764d4f5576a29334d8bd
Reviewed-on: https://swiftshader-review.googlesource.com/c/25490Tested-by: Hernan Liatis <hliatis@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Should pass the following deqp tests:
dEQP-VK.wsi.xlib.surface.create
dEQP-VK.wsi.xlib.surface.query_supportS
dEQP-VK.wsi.xlib.surface.query_capabilities
dEQP-VK.wsi.xlib.surface.query_formatsW
dEQP-VK.wsi.xlib.surface.query_present_modes
dEQP-VK.wsi.xlib.surface.destroy_null_handle
dEQP-VK.wsi.xlib.surface.initial_size
dEQP-VK.wsi.xlib.surface.resizeSW

Bug: b/124265819
Change-Id: I92da1cc8d60923ea97aa26d3d6a098274c6e06b7
Reviewed-on: https://swiftshader-review.googlesource.com/c/25308Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Corentin Wallez <cwallez@google.com>
Tested-by: Hernan Liatis <hliatis@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Pastel 1.0 does not need to implement logic ops. From discussion with
capn@, the existing support was only ever partial, and covered the exact
needs of a special project.

Change-Id: I3003227c4d29ab5d39e0ae6385207c87d22931e2
Reviewed-on: https://swiftshader-review.googlesource.com/c/25828Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Color attachment 0 (or any color attachments!) may not exist;
rasterizer behavior is supposed to be completely independent of this.

Bug: b/126417154
Change-Id: If16307a7964c267af254e2a62822ac38cf58c48c
Reviewed-on: https://swiftshader-review.googlesource.com/c/25809Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Tested-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>

Even in the GLES backend, we support RENDERTARGETS=8. This check only
ever considered the first 4.

Bug: b/126719156
Change-Id: I0084aa005f342f06373c53c007bee33268466b92
Reviewed-on: https://swiftshader-review.googlesource.com/c/25808Tested-by: Chris Forbes <chrisforbes@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>