Change-Id: I4136781005cbd0551461adecdd94d8e3cc8688d0
Reviewed-on: https://swiftshader-review.googlesource.com/8570Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

operator[] on a pointer was using the vector register size (16 bytes)
instead of the size of the (emulated) vector type.

Change-Id: Ice8677b41a3af81176fe9495fd02ede9f8b0482f
Reviewed-on: https://swiftshader-review.googlesource.com/8569Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Stores are dead if the address is from an alloca and all uses of the
address are stores, not just when it's a single store.

Change-Id: I5b65a64e21b9e398922e5440c35f0d318c03e911
Reviewed-on: https://swiftshader-review.googlesource.com/8574Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

If a store address has no definition, it must be a pointer argument.
Unlike stores to alloca's, these are not dead.

Change-Id: I66e50b14ffea1cfc61fc756b5545ab8ae28b4bf8
Reviewed-on: https://swiftshader-review.googlesource.com/8573Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

In order to add Subzero as a dependency in Chromium, Chromium's
DEPS file will be modified to use SwiftShader's DEPS file in
order to pull SwiftShader's desired revision of Subzero. This
will ensure that SwiftShader and Subzero are always in sync.

Change-Id: I0b24845c0ddbf974baac4ef6cc3ff6ab397d5f69
Reviewed-on: https://swiftshader-review.googlesource.com/8529Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Although SwiftShader was building properly on the build bots, it
wasn't exporting its symbols properly. I compared the flags used in
the BUILD.gn files and in the CMakeLists.txt file and attempted to
get both build systems to use the same flags and options.

Change-Id: Idc73d4ad5b45311881e88720bf611e25110dcc70
Reviewed-on: https://swiftshader-review.googlesource.com/8576Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Constant operands of commutative operations preferably go on the right
hand side to avoid requiring an extra register for two operand
instructions. Also, Subzero assumes constants in pointer arithmetic are
on the right hand side to consider optimizing it into an addressing
mode.

Change-Id: Ife5a471903d5f4bef0c19b6c908d75715f06bfec
Reviewed-on: https://swiftshader-review.googlesource.com/8548Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Change-Id: I40922c89056cacd00f9d728dfe1f3f12824c81a1
Reviewed-on: https://swiftshader-review.googlesource.com/3253Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

BUG: 31072273
Change-Id: I037505ad3ab1ba80aecab4e24ec8d1932df2dcf7
Reviewed-on: https://swiftshader-review.googlesource.com/7030Reviewed-by: Lingfeng Yang <lfy@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Lingfeng Yang <lfy@google.com>

MAP_ANON has been deprecated in favor of MAP_ANONYMOUS, but Mac OS only
defines MAP_ANON.

Bug chromium:630728

Change-Id: I7345a5f9227acbc0caa29d8103855ab531f16499
Reviewed-on: https://swiftshader-review.googlesource.com/8531Reviewed-by: Alexis Hétu <sugoi@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Fixed:
- type comparison mismatch
- redefined macros
- unused functions
- line endings

Change-Id: I965ca4b887aee968e6cf95feb36c3cbf90b85580
Reviewed-on: https://swiftshader-review.googlesource.com/8528Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Change-Id: I7979eb2557a6103ad10b7be3c274fadf12d27ae7
Reviewed-on: https://swiftshader-review.googlesource.com/8508Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Bug swiftshader:20

Change-Id: I67818bfe10cb29211559fb2ee047f6bec6ce46d4
Reviewed-on: https://swiftshader-review.googlesource.com/8451Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Bug swiftshader:20

Change-Id: I20c2ab7cb4c00c365520ff8b8500f7594127498b
Reviewed-on: https://swiftshader-review.googlesource.com/8468Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

When we have an uninitialized variable that is declared outside of a
loop and is read and written to inside the loop, we should not eliminate
the load and replace it with an undefined value, because in the next
iteration we want the value from the previous iteration.

Change-Id: Ic996c796307f62fc835079af6c58adf1eb288259
Reviewed-on: https://swiftshader-review.googlesource.com/8488Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Bug swiftshader:20

Change-Id: I81ad267d450713ffe2a5a84e1d7f7f140b515c85
Reviewed-on: https://swiftshader-review.googlesource.com/8454Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Provide default constructors for each type, and initialize the swizzle
parent in a less bug prone manner.

Change-Id: Ia7f406a66274c7b1737cbe82db733e2e3ab65175
Reviewed-on: https://swiftshader-review.googlesource.com/8453Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Change-Id: Ic8785bd56ac7303b0502f0303f9270a6999a1df9
Reviewed-on: https://swiftshader-review.googlesource.com/8452Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Subzero silently allows inserting a 32-bit integer into a vector with
16-bit elements. This generates a pinsrd instruction, which requires
SSE4.1, even when specifying to only support SSE2. This change emits
a pinsrw instruction instead, which is SSE2.

Bug swiftshader:20

Change-Id: I37bd0ebb6d9cfdd2cdef4788e89e9672386c685e
Reviewed-on: https://swiftshader-review.googlesource.com/8450Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Subzero returns vNi1 type results for vector comparisons, which we were
sign-extending to v4i32, even for smaller fields. We don't actually
ever have to sign-extend these results.

Change-Id: Ifdd30edd498d66fc4c557804035794a659eacd87
Reviewed-on: https://swiftshader-review.googlesource.com/8449Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Subzero uses a helper function to convert uint to float. It's faster to
just emit a sequence of operations to perform the cast manually.
This implementation converts the lower 31 bits as a signed integer, and
adds 0x80000000 as a floating-point value when the upper bit is set.
This approach does not produce the correct rounding in rare cases, but
should still be adequate. For consistency, we're also using this method
with the LLVM back-end.

Change-Id: Ic5d7b73cd2a9e154056365cdbe9af0962bdbe1cb
Reviewed-on: https://swiftshader-review.googlesource.com/8312Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Change-Id: I01d1e3a5464ba81440cc144ac50beae2c8b3f3c8
Reviewed-on: https://swiftshader-review.googlesource.com/8448Reviewed-by: Nico Weber <thakis@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Whenever we land a SwiftShader related change in Chromium,
it will look for owners of SwiftShader to make sure the cl
is properly LGTMed before landing, so the file was added
here.

Also fixed a case typo in an rc file.

Change-Id: I282a39764654a08b06320360dd66a5d877f368d3
Reviewed-on: https://swiftshader-review.googlesource.com/8428Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

Change-Id: Ie5f7d782764fbc40d4e3ed9cdce0110bbb9da80d
Reviewed-on: https://swiftshader-review.googlesource.com/8408Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

Chromium does not allow any third_party code to depend on base,
so the main test runner for swiftshader's tests in chromium can't
live inside the swiftshader repo. To solve this:

1) Added a test environment for swiftshader that will be used
for testing in chromium. Currently, it is simply empty.
2) Split unit tests in 2 files, unittests.cpp, which will
contain the tests ran by both chromium and swiftshader and
main.cpp, which will be used as the swiftshader test runner,
but NOT as the chromium test runner (that one will be added
in chromium at the same time I do the swiftshader update).
3) The BUILD.gn file was updated to make the chromium version
of the test depend on base and on the chromium version of the
swiftshader test runner (swiftshader_tests_main.cc, to be added).

Change-Id: I1128bb482565908bc285dcd6818d90d1d881a50e
Reviewed-on: https://swiftshader-review.googlesource.com/8409Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

PNaCl does not use the Microsoft x86-64 calling convention on Windows,
so the System V convention is the new default for Subzero. We need to
set a compile definition to explicitly request the use of the
Microsoft convention.

Change-Id: Ic70f5015d4b5753aa08a1247b6ac73eb9a9d278c
Reviewed-on: https://swiftshader-review.googlesource.com/8378Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

libX11 may have already been loaded (dynamically or statically) by the
application. Attempting to load it again could result in a different
version being loaded, causing various compatibility issues.

Bug b/32880157

Change-Id: Ica8e6b1c85b82469885a20dcef31e25e6b598132
Reviewed-on: https://swiftshader-review.googlesource.com/8348Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

CMake considers built-in configuration variables to be predefined, thus
setting CMAKE_BUILD_TYPE with a default value doesn't work because it
already exists. The FORCE attribute overrides it, but we have to guard
against overriding user-specified values.

Change-Id: Icf1cad6af15e788c5ae9ac8424f1ef0d29f9a1ff
Reviewed-on: https://swiftshader-review.googlesource.com/8377Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Bug b/34128224

Change-Id: I4aebcda840baa8ceef2ae99c98a80a83e32b89b3
Reviewed-on: https://swiftshader-review.googlesource.com/8376Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

This enables building with either LLVM or Subzero as the Reactor JIT
compiler back-end, on both Linux and Windows.

Change-Id: I4c052d224cd1302b7de412df58b07ee275ba924f
Reviewed-on: https://swiftshader-review.googlesource.com/8375Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

Change-Id: I82b6a3f5336ca64dea30a8d9541fcf43c7ead81a
Reviewed-on: https://swiftshader-review.googlesource.com/8373Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Change-Id: I3ad40f5c0f9faf8aa27356b72b4d98763b9711d7
Reviewed-on: https://swiftshader-review.googlesource.com/8372Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Change-Id: I03c5ebf799af288558d08190058d03f1ac2922b7
Reviewed-on: https://swiftshader-review.googlesource.com/8355Reviewed-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Change-Id: Ib1339f5a40523007ca5832ed1bc7b7d096e6c1f2
Reviewed-on: https://swiftshader-review.googlesource.com/8374Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

This silences a benign assert in the config sorting operator which
checks for unique configs. On platforms other than Android, all of the
configs are marked as conformant.

Bug b/34029810

Change-Id: I600d67747ba24f153a0316609e5a783991093ebd
Reviewed-on: https://swiftshader-review.googlesource.com/8388Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

The spec states that glViewport() silently clamps the viewport
width and height to GL_MAX_VIEWPORT_DIMS[0] and
GL_MAX_VIEWPORT_DIMS[1], respectively.

Bug b/34078120

Change-Id: Ifeec0d6b601ce8a3825796fa551eea1f46150002
Reviewed-on: https://swiftshader-review.googlesource.com/8371Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Nicolas Capens <capn@google.com>

Rasterization of very large triangles was causing signed 32-bit
integer overflow due to multiplying two unsigned 12.4 fixed-point
coordinates. The equations have been reworked to only require
multiplication of 12.4 by 0.4 fixed-point.

Bug b/34078120

Change-Id: I227b81254559af04baf50fbfec6a7f123bd230e3
Reviewed-on: https://swiftshader-review.googlesource.com/8370Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <capn@google.com>

Change-Id: Ibfd483ce3cba6ec036327b32cba1460d68f8c9e8
Reviewed-on: https://swiftshader-review.googlesource.com/3542Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

The index used to index into a structure (a scalar) used to have the
vector size of the resulting type. This changed recently to always be
1, so the field type size needs to be determined from the fields
themselves, which is also the type of the indexing result.

Change-Id: I2dc373d8d31b02c0f69879cb0a3feacd83d6e473
Reviewed-on: https://swiftshader-review.googlesource.com/8368Tested-by: Nicolas Capens <capn@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>

If VAOs are deleted starting from 0, then the
following happens when an app's context has
genned two VAO's that are both != 0,
and the first of the two nonzero VAOs
is the current one:

VAO 0
VAO 1 <- current
VAO 2

in context dtor loop over vao map:
delete vao 0 (done)
delete vao 1:
is current, so bind 0:
in bind 0, we create a new map entry since 0 was deleted
next loop iter deletes vao 0 again (done)
*** at this point, vao 0 pointer refers to freed memory ***
next loop iter wants to delete vao 2,
but the current vao is 0 and refers to freed memory.
then, in Context::deleteVertexArray:
if (getCurrentVertexArray()->name...) <- Segmentation fault (core dumped)

This CL deletes VAO 0 last.

Change-Id: Ifa606ad7517cd213f21606577d3bdd8d810b640d
Reviewed-on: https://swiftshader-review.googlesource.com/8350Reviewed-by: Nicolas Capens <capn@google.com>
Tested-by: Lingfeng Yang <lfy@google.com>