|
bb5a7e29
|
2017-08-30T13:03:12
|
|
Allow length() on arbitrary array expressions
This is required to pass some dEQP GLES 3.1 tests for arrays of
arrays, and WebGL conformance tests were also recently fixed to
require this behavior. The intent of the GLSL ES spec was not to
restrict usage of length().
In practice GL drivers don't implement array length() on expressions
with side effects correctly in all cases. HLSL doesn't have an array
length operator either. Because of this we always remove array length
ops from the AST before output.
BUG=angleproject:2142
TEST=angle_unittests, angle_end2end_tests, WebGL conformance tests
Change-Id: I863a92e83ac5315b013af9a5626348482bad72b3
Reviewed-on: https://chromium-review.googlesource.com/643190
Commit-Queue: Olli Etuaho <oetuaho@nvidia.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
9733ceef
|
2017-05-11T19:14:35
|
|
Initialize uninitialized locals in GLSL output
Guarantee that local variables are initialized before they are used
in GLSL output. In HLSL output all variables were already being
initialized.
Locals are initialized using an AST transform. The local variable init
can only be run after some simplification of the AST, so that it is
able to handle complex cases like:
for (int i[2], j = i[0]; i[0] < 3; ++i[0]) {
}
If we're dealing with ESSL 1.00 which lacks array constructors, in
this kind of case the uninitialized array initialization code needs to
be hoisted out of the loop init statement, and the code also needs to
make sure that j's initializer is run after i is initialized.
Another complex case involves nameless structs. This can be an issue
also in ESSL 3.00 and above:
for (struct { float f; } s; s.f < 1.0; ++s.f) {
}
Since the struct doesn't have a name, its constructor can not be used.
We solve this by initializing the struct members individually,
similarly to how arrays are initialized in ESSL 1.00.
Initializing local variables is disabled on Mac and Android for now.
On Mac, invalid behavior was exposed in the WebGL 2.0 tests when
enabling it. On Android, the dEQP test runs failed for an unknown
reason. Bugs have been opened to resolve these issues later.
BUG=angleproject:1966
TEST=angle_end2end_tests, WebGL conformance tests
Change-Id: Ic06927f5b6cc9619bc82c647ee966605cd80bab2
Reviewed-on: https://chromium-review.googlesource.com/504728
Commit-Queue: Olli Etuaho <oetuaho@nvidia.com>
Reviewed-by: Jamie Madill <jmadill@chromium.org>
|
|
|
1ecd14b8
|
2017-01-26T13:54:15
|
|
Fold user-definedness of function nodes into TOperator
Whether a function call is user-defined is not orthogonal to TOperator
associated with the call node - other ops than function calls can't be
user-defined. Because of this it makes sense to store the user-
definedness by having different TOperator enums for different types of
calls.
This patch also tags internal helper functions that have a raw
definition outside the AST with a separate TOperator enum. This way
they can be handled with logic that is easy to understand. Before this,
function calls like this left the user-defined bit unset, despite not
really being built-ins either. The EmulatePrecision traverser uses
this. This is also something that could be used to clean up built-in
emulation in the future.
BUG=angleproject:1490
TEST=angle_unittests
Change-Id: I597fcd9789d0cc22b689ef3ce5a0cc3f621d4859
Reviewed-on: https://chromium-review.googlesource.com/433443
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Olli Etuaho <oetuaho@nvidia.com>
|
|
1212bcac
|
2016-11-23T13:44:05
|
|
translator: separate declarations after rewriting loops
Otherwise when trying to add the declarations back, things might fail
because the loop initialization is a sequence and not a block.
BUG=668028
Change-Id: I8d84a25c25765e9655c16ce56604ae08f0f8176c
Reviewed-on: https://chromium-review.googlesource.com/414305
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Olli Etuaho <oetuaho@nvidia.com>
|
|
45bcc784
|
2016-11-07T13:58:48
|
|
translator: Scope all classes with "sh".
I was seeing an odd problem with our PoolAlloc conflicting with the
glslang/Vulkan TIntermNode, so the fix was to move everything to a
separate namespace.
The bison grammars are also regenerated. No functional changes.
BUG=angleproject:1576
Change-Id: I959c7afe4c092f0d458432c07b4dcee4d39513f3
Reviewed-on: https://chromium-review.googlesource.com/408267
Reviewed-by: Yuly Novikov <ynovikov@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
6d40bbdd
|
2016-09-30T13:49:38
|
|
Split TIntermBlock from TIntermAggregate
The new TIntermBlock node class replaces TIntermAggregate nodes with
the EOpSequence op. It represents the root node of the tree which is
a list of declarations and function definitions, and any code blocks
that can be denoted by curly braces. These include function and loop
bodies, and if-else branches.
This change enables a bunch of more compile-time type checking, and
makes the AST code easier to understand and less error-prone.
The PostProcess step that used to be done to ensure that the root node
is TIntermAggregate is removed in favor of making sure that the root
node is a TIntermBlock in the glslang.y parsing code.
Intermediate output formatting is improved to print the EOpNull error
in a clearer way.
After this patch, TIntermAggregate is still used for function
definitions, function prototypes, function parameter lists, function
calls, variable and invariant declarations and the comma (sequence)
operator.
BUG=angleproject:1490
TEST=angle_unittests, angle_end2end_tests
Change-Id: I04044affff979a11577bc1fe75d747e538b799c8
Reviewed-on: https://chromium-review.googlesource.com/393726
Commit-Queue: Olli Etuaho <oetuaho@nvidia.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
|
|
|
d0bad2c7
|
2016-09-09T18:01:16
|
|
Split ternary node class from TIntermSelection
Ternary operator nodes are typed parts of expressions, they always
have two children and the children are also guaranteed to be
TIntermTyped. "If" selection nodes can't be a part of an expression,
they can have either one or two children and the children are code
blocks. Due to all of these differences it makes sense to store these
using two different AST node classes.
BUG=angleproject:1490
TEST=angle_unittests
Change-Id: I913ab1d806e3cdb5c21106f078cc9c0b6c72ac54
Reviewed-on: https://chromium-review.googlesource.com/384512
Commit-Queue: Olli Etuaho <oetuaho@nvidia.com>
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
|
|
|
666f65a1
|
2016-08-26T01:34:37
|
|
Revert "Revert "Cover vector dynamic indexing case in SplitSequenceOperator""
This reverts commit d2f59bb6dda4f3548e158a09540829f9ff56bba4.
Change-Id: If2842bce17a0c085e2bc913ff120083fbe90497c
Reviewed-on: https://chromium-review.googlesource.com/376189
Reviewed-by: Jamie Madill <jmadill@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
d2f59bb6
|
2016-08-17T11:50:52
|
|
Revert "Cover vector dynamic indexing case in SplitSequenceOperator"
This CL was causing inverted rendering in a WebGL application.
This reverts commit 7da9850643f55335a13a4663d226c73d0ac4d3b1.
Vectors or matrices that are dynamically indexed as a part of an
l-value generate new statements in the RemoveDynamicIndexing AST
transformation step. SplitSequenceOperator needs to detect this case
and split the sequence operator before statements are generated from
its operands to ensure the correct order of execution.
BUG=angleproject:1341
TEST=angle_end2end_tests
Change-Id: I854f8cce2d46107afa62f48edf3d32c6d5c97eda
Reviewed-on: https://chromium-review.googlesource.com/371643
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Commit-Queue: Jamie Madill <jmadill@chromium.org>
|
|
|
7da98506
|
2016-07-20T18:45:09
|
|
Cover vector dynamic indexing case in SplitSequenceOperator
Vectors or matrices that are dynamically indexed as a part of an
l-value generate new statements in the RemoveDynamicIndexing AST
transformation step. SplitSequenceOperator needs to detect this case
and split the sequence operator before statements are generated from
its operands to ensure the correct order of execution.
BUG=angleproject:1341
TEST=angle_end2end_tests
Change-Id: I84e41a59c88fb5d0111669cab60312b930531a22
Reviewed-on: https://chromium-review.googlesource.com/361695
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Olli Etuaho <oetuaho@nvidia.com>
|
|
|
00f6fbbe
|
2016-07-20T16:32:29
|
|
Add IntermNodePatternMatcher helper class
This will enable sharing code between different AST traversers that
apply transformations on similar node structures. This will make the
code more maintainable.
For now the helper class is used in UnfoldShortCircuitToIf and
SeparateExpressionsReturningArrays.
BUG=angleproject:1341
TEST=angle_end2end_tests, WebGL 2 conformance tests
Change-Id: Ib1e0d5a84fd05bcca983b34f18d47c53e86dc227
Reviewed-on: https://chromium-review.googlesource.com/361693
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Olli Etuaho <oetuaho@nvidia.com>
|