When `invariant` is enabled on MSL was was incorrectly setting
`@invariant` instead of `[[invariant]]`. We test with metal1.2 which
does not have invariant, so this only showed up when using metal2.1 or
higher.
Bug: chromium:1439273
Change-Id: Iab866608195e697b0370d465f350b25277d904a3
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/128880
Commit-Queue: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
D3D11 only supports HLSL SM5.0 which doesn't support `space`
(binding group in WGSL). So for D3D11, only one binding group will be
used, and tint will not emit `space` for HLSL, so shaders can be used
with D3D11.
Bug: dawn:1705
Change-Id: Ie0e9868137f10762c5243e188d76f5e41879c2bc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/125080
Commit-Queue: Peng Huang <penghuang@chromium.org>
Reviewed-by: Ben Clayton <bclayton@google.com>
Kokoro: Ben Clayton <bclayton@google.com>
min16float is not required to convert its input to an 16-bit float.
Instead, use f32tof16, and f16tof32.
This is also required to be able to enable the IEEE strictness option on
FXC, otherwise FXC emits an error X4028: expression has a minimum
precision type and cannot be marked precise
Bug: tint:1887
Change-Id: I2268c9336b8e873b36e6532984838054fe4dedbd
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/125920
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
In order to avoid declaring too many function parameters, we
previously modified this transform to redeclare private variables that
are only used inside a single function as function-scope
variables. This was broken as it meant that their values did not
persist across multiple calls to the same function.
Instead, wrap all private variables in a structure and pass it around
as a pointer.
Fixed: tint:1875
Change-Id: I83f5eb1071d57b9c6af56d6cf21b3a32c6e94260
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/124800
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: James Price <jrprice@google.com>
Conversion can happen for structure materialization (modf, frexp).
If both structure members are the same type and value, then a constant::Splat will be constructed, which needs to handle conversion.
Bug: chromium:1417515
Change-Id: Iadd14ce00b8d5c22226c601ec5af9a84e6c0c5cf
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/122900
Reviewed-by: James Price <jrprice@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
... overloads for texture cube arrays
FXC is telling porkies, when it says "The array element count
of GetDimensions on TextureCubeArray objects is unavailable
on ps_5_1".
The actual issue, as identified by Teodor Tanasoaia at Mozilla,
is that the argument needs to be unsigned.
In fact, *all* the texture queries should have used an unsigned
scalar or vector for the output value. This has been broken
forever!
This reverts commit bd9f6e6684.
Change-Id: I3e217bec17c6fd203cff618b143ebef3d8a61927
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/122980
Auto-Submit: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Dan Sinclair <dsinclair@chromium.org>
Commit-Queue: Dan Sinclair <dsinclair@chromium.org>
This change overhauls the Robustness transform to support three modes,
per address space:
* ignore - Disable robustness checks for the address space
* clamp - Clamp indices / texture args to ensure they're in
bounds. This was the old behavior, and continues to
be the default.
* predicate - Condition all indexing / textureLoad / textureStore /
atomic* operations on the bounds check. If any
dependent value is out of bounds, then the operation
is skipped.
This change also fixes multiple expression evaluation of the texture
builtin 'level' argument.
Change-Id: I2e300ddff2c8d3183a9701f06985ce1b262baf2c
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/122343
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: James Price <jrprice@google.com>
Kokoro: Ben Clayton <bclayton@chromium.org>
Commit-Queue: Ben Clayton <bclayton@chromium.org>
Recent changes to DecomposeMemoryAccess meant we lost the dependency information between the user of a module-scope variable of the storage / uniform address space and the variable.
Add dependency information to ast::InternalAttribute so this can be tracked.
This change also means that symbol renaming after the DecomposeMemoryAccess should work.
Fixed: tint:1860
Change-Id: Icfa2925f95c2ac50702522df514cd11bde727546
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/122660
Reviewed-by: James Price <jrprice@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
This CL updates the templates in the StringStream to match more types.
All of the internal `operator<<` methods have been converted over to
StringStream. The precision was increased in order to better match the
precision needed to read back as a double.
Bug: tint:1686
Change-Id: Iaa15cf247f174967dd1014647ba5a74804997c22
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/122080
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: Dan Sinclair <dsinclair@chromium.org>
This CL switches the GLSL Generator to use utils::StringStream. The line
writer is converted internally as well, although it converts to
`std::ostream` for now. This caused a couple of the MSL, HLSL and GLSL
tests to generate slightly fewer decimal points in a couple tests.
Bug: tint:1686
Change-Id: I9ec8c1a5ef49679fc1c9a9aece86ab3390e103fc
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/121880
Commit-Queue: Dan Sinclair <dsinclair@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
In order to preserve padding properly for MSL, we need to use its
packed_vec type for all vec3 types in storage buffers, not just struct
members. This commit includes a complete rewrite of the PackedVec3
transform to achieve this. The key details are:
* An internal `__packed_vec3<>` type was added, which corresponds to a
`type::Vector` with an additional flag to indicate that it will be
emitted as packed vector.
* The `PackedVec3` transform replaces all vec3 types used in
host-shareable address spaces with the internal `__packed_vec3`
type. This includes vec3 types that appear as the store type of a
pointer.
* When used as an array element, these `__packed_vec3` types are
wrapped in a struct that contains a single `__packed_vec3`
member. This allows us to add an `@align()` attribute that ensures
that `array<vec3<T>>` still has the correct array element stride.
* When the `vec3<T>` appears as a struct member in the input program,
we apply the `@align()` to that member to ensure that we do not
change its offset.
* Matrix types with three rows that are used in memory are replaced
with an array of columns, where each column uses a `__packed_vec3`
inside an aligned wrapper structure as above.
* Accesses to host-shareable memory that involve any of these types
invoke a "pack" or "unpack" helper function to convert them to the
equivalent type that uses `__packed_vec3` or a regular `vec3` as
required.
* The `chromium_internal_relaxed_uniform_layout` extension is used to
avoid issues where modifying a type in the uniform address space
triggers stricter layout validation rules.
Bug: tint:1571
Fixed: tint:1837
Change-Id: Idaf2da2f5bcb2be00c85ec657edfb614186476bb
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/121200
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: James Price <jrprice@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Change the DecomposeMemoryAccess to behave more like the DirectVariableAccess transform, in that it'll inline the access of buffer variable into the load / store helper functions, instead of passing the array down.
This avoids large array copies observed with FXC, which can have *severe* performance costs.
Fixed: tint:1819
Change-Id: I52eb3f908813f72ab9da446743e24a2637158309
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/121460
Kokoro: Kokoro <noreply+kokoro@google.com>
Auto-Submit: Ben Clayton <bclayton@google.com>
Reviewed-by: James Price <jrprice@google.com>
Commit-Queue: James Price <jrprice@google.com>
This CL removes the following AST nodes:
* ast::Array
* ast::Atomic
* ast::Matrix
* ast::MultisampledTexture
* ast::Pointer
* ast::SampledTexture
* ast::Texture
* ast::TypeName
* ast::Vector
ast::Type, which used to be the base class for all AST types, is now a
thin wrapper around ast::IdentifierExpression. All types are now
referred to using their type name.
The resolver now handles type resolution and validation of the types
listed above based on the TemplateIdentifier arguments.
Other changes:
* ProgramBuilder has undergone substantial refactoring.
* ProgramBuilder helpers for type inferencing is now more explicit.
Instead of passing 'nullptr', a new 'Infer' template argument is
passed.
* ast::CheckIdentifier() is used for more tests that check identifiers,
including types.
Bug: tint:1810
Change-Id: I8e739ef49435dc1c20a462f3ec5ba265661a7edb
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/118723
Reviewed-by: Dan Sinclair <dsinclair@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: James Price <jrprice@google.com>
