Change-Id: I8158c0bb2c19d6cdee12a970f24785fae609d5fc
Reviewed-on: https://swiftshader-review.googlesource.com/5352Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

https://msdn.microsoft.com/en-us/library/windows/desktop/bb172511(v=vs.85).aspx

Change-Id: I2bc8251bce77aacb55fbad1c677f865395f5a986
Reviewed-on: https://swiftshader-review.googlesource.com/5351Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Change-Id: I192d8fe104c3734c9c37108312f1dc93470d8d9e
Reviewed-on: https://swiftshader-review.googlesource.com/5342Reviewed-by: Shannon Woods <shannonwoods@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Change-Id: I9fe9bd339f2aacb04394259ee8753dbc8e041430
Reviewed-on: https://swiftshader-review.googlesource.com/5340Reviewed-by: Shannon Woods <shannonwoods@google.com>
Tested-by: Nicolas Capens <capn@google.com>

When depth and stencil are both present, they must be the same image,
or return GL_FRAMEBUFFER_UNSUPPORTED, according to the GLES 3.0 spec,
section 4.4.4, Framebuffer Completeness.

Change-Id: I607c05617fb55743635eda987da6de523385d92b
Reviewed-on: https://swiftshader-review.googlesource.com/5262Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Change-Id: I4a7aeefedcd2d891093520d5a10ebefadcddb5be
Reviewed-on: https://swiftshader-review.googlesource.com/5320Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Change-Id: I72439a60850add0600face13c94cc481a7083792
Reviewed-on: https://swiftshader-review.googlesource.com/5221Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Bug 28616401

Change-Id: Ia54134e6e1dc6b101c7f81d9047e9bcc9fab3061
Reviewed-on: https://swiftshader-review.googlesource.com/5311Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Bug 28616401

Change-Id: I888ac3dd3186206c7745c23a542b91510a461f45
Reviewed-on: https://swiftshader-review.googlesource.com/5310Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Bug 28474248

Change-Id: I61bf28687623437c6b9b383dc189ffa28ad24c22
Reviewed-on: https://swiftshader-review.googlesource.com/5330Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Change-Id: I7f62eae32d2122ec14f400841eaefa3712cf582d
Reviewed-on: https://swiftshader-review.googlesource.com/5291Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

The spec states that "If <sync> is not a valid sync object for <dpy>,
EGL_FALSE is returned and an EGL_BAD_PARAMETER error is generated."

Change-Id: I4bbd2ddfefd7baba9c301decb644bfb545e66f01
Reviewed-on: https://swiftshader-review.googlesource.com/5290Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Implemented modf as trunc + sub
Passes all related dEQP tests

Change-Id: I43656c51a670d235153e5fac390a8db311b14f8d
Reviewed-on: https://swiftshader-review.googlesource.com/5280Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Bug 28474248

Change-Id: Id5b93f3494c918cb344e5da7010dfa48a25cdde1
Reviewed-on: https://swiftshader-review.googlesource.com/5263Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

This reverts commit 0637eac3.
It's considered a spec bug:
https://cvs.khronos.org/bugzilla/show_bug.cgi?id=8656

Change-Id: Id49121d5adf91ecab3d261c3c12e2f879f47e856
Reviewed-on: https://swiftshader-review.googlesource.com/5192Reviewed-by: Shannon Woods <shannonwoods@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Change-Id: I1d1a716bd9bd5b636c5b263a7094ff13ac9aa549
Reviewed-on: https://swiftshader-review.googlesource.com/5260Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Change-Id: Iddd9acf8496abe4dfa147f8a59b4cfb2f23df3a7
Reviewed-on: https://swiftshader-review.googlesource.com/5261Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

This cl adds a utility function that assigns the correct buffer
for each linked transform feedback variable. All the information
about location and size is sent to the VertexProcessor for use
during rendering.

Change-Id: I942805250804f56805de1fc117024c20976e83a0
Reviewed-on: https://swiftshader-review.googlesource.com/5174Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Previously we would mask instruction execution after hitting a LEAVE,
but still jump back on every loop iteration (even if infinite). This
change applies the enableLeave mask to the loop test condition so we
break out of the loop when all strands have hit LEAVE.

Change-Id: Ia331a91fb363d20743d5b87a468c3cfcaa7b481e
Reviewed-on: https://swiftshader-review.googlesource.com/5250Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

The not equal folding logic was wrong. It would end the loop
as soon as it found equal components, which was checking if
ALL components are not equal, but that's not how not equal
works, it should return true if ANY component is not equal.

I also refactored it because it was hard to follow, with the
boolNodeFlag variable being used upside down of the result.

Change-Id: I704f26f2fd31a1cc637f5d7601409de8161c5b80
Reviewed-on: https://swiftshader-review.googlesource.com/5211Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

The SRGB conversion code was already available, but wasn't used
specifically for the SRGB type framebuffers. Also, the SRGB
conversion should always be applied after blending.

According to the GLES 3.0 spec, section 4.1.8 - sRGB Conversion:
"the R, G, and B values after blending are converted
 into the non-linear sRGB color space by computing."

All related dEQP tests pass.

Change-Id: I9342d2f74aa650f28835a951bdfa8bd371bc6924
Reviewed-on: https://swiftshader-review.googlesource.com/5189Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Change-Id: I0bde678865902d549d4260bf8f6c1a368a46f4d4
Reviewed-on: https://swiftshader-review.googlesource.com/5191Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Change-Id: If90d45b5f0c83620486dd7b6decc2f0f5fc7e558
Reviewed-on: https://swiftshader-review.googlesource.com/5220Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Whenever texture projection is used with a constant texture
coordinate, folding allows to use the full precision reciprocal
and not generate the lower precision reciprocal operation.

Change-Id: I6cab6567d63ecd9abe1cedbd7e46e1fd9099a3d3
Reviewed-on: https://swiftshader-review.googlesource.com/5210Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Added constant folding for:
rad, deg, sin, cos, tan, asin, acos, atan, sinh, cosh, tanh,
asinh, acosh, atanh, log, log2, exp, exp2, sqrt, inverstsqrt

Change-Id: Ifc6b089e5981da9a3739d4ed7636953a682ce220
Reviewed-on: https://swiftshader-review.googlesource.com/5200Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

There's an LLVM bug currently that prevents us from using ICMP_EQ,
but ICMP_NE works fine, so the "equality" test has been changed to
"!inequality" to fix the issue. The fix should be reverted once
LLVM is updated to a version where the ICMP_EQ issue is fixed.

Change-Id: I79d6ca99554317cc64ffa2905ae2804bd4805e2a
Reviewed-on: https://swiftshader-review.googlesource.com/5193Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>

Bug 27690616

Change-Id: I28b62fc56b0b572c1c9e94499d8a0c4c56737f3c
Reviewed-on: https://swiftshader-review.googlesource.com/5190Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Addition evaluates before the ternary operator, which resulted in the wrong
sizes being computed.

Bug 28346104

Change-Id: Id28843132097a4fb1599eed19c05f479f16f1315
Reviewed-on: https://swiftshader-review.googlesource.com/5187Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Change-Id: I8974a6bb4137d1dfad9f5dd3a92cc9306c38294c
Reviewed-on: https://swiftshader-review.googlesource.com/5184Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

dEQP tests enforce that the defined operator should be parsed even when
it is generated as a result of macro expansion, even though this is
undefined according to the C++ preprocessor spec.

Implement support for this by putting the parsing for the defined
operator inside MacroExpander. The operator gets processed right after
it is generated by macro expansion. Parsing the defined operator is
toggled with a boolean according to the context where MacroExpander
is used.

Change-Id: I8557e829f4278ab6cb27eb4a0f84ca0c0dd18d1a
Reviewed-on: https://swiftshader-review.googlesource.com/5182Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Change-Id: I49d6d36ecf9a5f0d3d333422b3e4970192617c05
Reviewed-on: https://swiftshader-review.googlesource.com/5180Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Change-Id: Idc8c2a241af91916857ba015b061ce655b33e866
Reviewed-on: https://swiftshader-review.googlesource.com/5179Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Change-Id: I29172c0c5423750db2665c350746b82fd2a34653
Reviewed-on: https://swiftshader-review.googlesource.com/5181Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

In a lot of cases, int was being used instead of size_to represent
sizes. That led to some warnings about inconsistencies between int
and size_t usage. While this cl doesn't solve all warnings, it tries
to use size_t and int where it should be appropriate to use them.

Change-Id: Id760df1360f65b2bba60f4075cdf4954fc6bbaf3
Reviewed-on: https://swiftshader-review.googlesource.com/5177Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>

The spec states that GL_INVALID_OPERATION is generated when a shader or
program function is called on the wrong object type. GL_INVALID_VALUE is
generated when there's no shader nor program object by that name.

Change-Id: Id9d922b390ca5d2db233eb60b6a44992105cb933
Reviewed-on: https://swiftshader-review.googlesource.com/5176Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

When textures have padding after a row of pixels, we can't use the sampler's
fast path which uses shift operations. We have to treat it as an NPOT texture.
Previously only mipmap level 0 was checked for padding, but when a texture is
also a render target we allocate 2x2 pixel blocks so a 1x1 mipmap has stride 2.

Change-Id: I7421fddbe5ed0d330d881f09c2161d9c42348600
Reviewed-on: https://swiftshader-review.googlesource.com/5175Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Proper validation of varyings used for transform feedback was added.
Simple types, as well as arrays and array elements are allowed.

Change-Id: I83ceb5eb19bf5691264ab0e142c3cc6a7debda4c
Reviewed-on: https://swiftshader-review.googlesource.com/5058Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

We used to compute the cube map LOD by projecting all sampling rays onto the
same face (arbitrarily chosen). This changes it to intersect the rays with the
cube and use the 3D distance between them.

Change-Id: If90b29cb4c13e6128ce6642eb28bb92602e783df
Reviewed-on: https://swiftshader-review.googlesource.com/5160Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

From the ES3 spec, section 4.4.2 (about DeleteRenderbuffers):
"Unused names in renderbuffers are silently ignored, as is
 the value zero."

And from section 6.1.5:
"IsSampler will return TRUE if sampler is the name of a
 sampler object previously returned from a call to
 GenSamplers and FALSE otherwise".

Also, glIsSampler can use Context::isSampler(), which uses
ResourceManager::isSampler(), which should only look whether a
name has been reserved, not whether the object is allocated.

Note: Even though the spec mentions (about Sampler objects):
"they acquire state only when they are first used as a parameter
to BindSampler, SamplerParameter*, GetSamplerParameter*, or
IsSampler", it shouldn't hurt to not allocate the object when
calling IsSampler, since a subsequent call to any other Sampler
object related function should allocate it.

Change-Id: I69c9d8bbcf93231747b913c0afbd970c78d02cf1
Reviewed-on: https://swiftshader-review.googlesource.com/5161Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Change-Id: Id82e8dba0d934f5c4bb2eefe1647839f25917898
Reviewed-on: https://swiftshader-review.googlesource.com/5150Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>