Emit the "[loop]" attribute on "for" and "while" so that FXC does not
attempt to unroll them. This is to work around an FXC bug where it fails
to unroll loops with gradient operations.
FXC ostensibly unrolls such loops because gradient operations require
uniform control flow, and loops that have varying iterations may
possibly not be uniform. Tint will eventually validate that control flow
is indeed uniform, so forcing FXC to avoid unrolling in these cases
should be fine.
Bug: tint:1112
Change-Id: I10077f8b62fbbb230a0003f3864c75a8fe0e1d18
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/69880
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
Use uintBitsToFloat.
Bug: tint:1306
Change-Id: Ie9a5e14c13c0d63b57c126f16c4e2a5c7a77e3f7
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/69740
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
In GLSL, all identifiers beginning with gl_ are reserved (not just those
explicitly named in the spec), and so any found in WGSL must be renamed.
Bug: tint:1304
Change-Id: I92ed7ec674620f67775378ecb8debcfdb4b5bbb4
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/69701
Reviewed-by: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
In GLSL, relational operators are only valid for scalar operands. For
vector operands, lessThan, greaterThan, etc must be used.
Bug: tint:1303
Change-Id: Ia800f89111630c756dc1b30ef0c6858fb520fb16
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/69561
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Stephen White <senorblanco@chromium.org>
Reviewed-by: Ben Clayton <bclayton@google.com>
To fix this, we trick the compiler by wrapping the function body with an
if (true) { <function body> } followed by returning an unused value of
the return type.
Bug: tint:1081
Change-Id: I763bf768f40d07a1045f0a70017bb40d488c8428
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/68822
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
Some pass DXC, but not FXC (e.g. continue_in_switch).
Some now pass (e.g intrinsics/gen/atan2/*.hlsl)
Some now fail gracefully instead of asserting (e.g. 807.spv)
Change-Id: I92b17fcadc7850de5bd93ff07507cea7e5487fc9
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/68820
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@chromium.org>
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
The semantic nodes cannot be fully immutable, as they contain cyclic
references. Remove Resolver::CreateSemanticNodes(), and instead
construct and mutate the semantic nodes in the single traversal pass.
Give up on trying to maintain the 'authored' type names (aliased names).
These are a nightmare to maintain, and provided limited use.
Significantly simplfies the Resolver, and allows us to generate more
semantic to semantic references, reducing sem -> ast -> sem hops.
Note: This change introduces constant value propagation across constant
variables. This is unlocked by the earlier construction of the
sem::Variable.
Change-Id: I592092fdc47fe24d30e512952511c9ab7c16d7a1
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/68406
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Update or remove tests that try to do this.
Fixed: tint:491
Change-Id: I1f351a4abf68ae9bc6b100885fb1bcea08b31211
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/68242
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
WGSL does not support pointer-to-vector-component, so the SPIR-V
reader needs to sink these pointers into their use.
Bug: tint:491
Change-Id: Ib5ae87d2f6bbac13280314ba11369d7ced505b56
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/68241
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: David Neto <dneto@google.com>
This removes a lot of awkward logic from the MSL writer, and means
that we now handle all module-scope variables with the same transform.
Change-Id: I782e36a4b88dafbc3f8364f7caa7f95c6ae3f5f1
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/67643
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
The `Ignore()` intrinsic is about to be deprecated, so don't use it for testing.
Bug: tint:1213
Change-Id: I314ecaeb9a9c337c7b6980189054120a74807ebd
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/67066
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: David Neto <dneto@google.com>
Reviewed-by: James Price <jrprice@google.com>
Metal 1.x does not support swizzling on packed_vec types.
Use array-index for single element selection (permitted on LHS and RHS of assignment)
Cast the packed_vec to a vec for multiple element swizzles (not permitted as the LHS of an assignment).
Fixed: tint:1249
Change-Id: I70cbb0c22a935b06b3905d24484bdc2edfb95fc2
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/67060
Commit-Queue: Ben Clayton <bclayton@google.com>
Kokoro: Ben Clayton <bclayton@google.com>
Reviewed-by: James Price <jrprice@google.com>
The intrinsics that did anything useful with this were deprecated
several releases ago.
Change-Id: I79e3c901b6a78583853a067ec46cfa98e346517c
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/66262
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: James Price <jrprice@google.com>
Update one test to avoid this error.
Fixed: tint:1179
Change-Id: Id41b0eb0f404648de4e86a835fe43f1729cb4696
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/64464
Auto-Submit: David Neto <dneto@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: David Neto <dneto@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Change ast::Array to use an ast::Expression for its `size` field. The
WGSL frontend now parses the array size as an `primary_expression`,
and the Resolver is responsible for validating the expression is a
signed or unsigned integer, and either a literal or a non-overridable
module-scope constant.
The Resolver evaluates the constant value of the size expression, and
so the resolved sem::Array type still has a constant size as before.
Fixed: tint:1068
Fixed: tint:1117
Change-Id: Icfa141482ea1e47ea8c21a25e9eb48221f176e9a
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/63061
Auto-Submit: James Price <jrprice@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: James Price <jrprice@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
In https://dawn-review.googlesource.com/c/tint/+/62444 the Resolver validated that there are no parameters of the same function with the same name, but this also introduced validation that errors if parameters shadow a module-scope variable.
The WGSL spec allows for shadowing, but Tint so far has not implemented this support.
There are transforms that generate functions that presume parameter <-> module-scope variable shadowing is okay. DecomposeMemoryAccess is one of these.
This fixes those transforms which could generate programs that fail validation.
Bug: chromium:1242330
Fixed: tint:1136
Change-Id: Id6ec59bbdb398b3b2a23312115a7c1dadf433e98
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/62900
Reviewed-by: James Price <jrprice@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
These operators are not defined in the metal namespace when the vector
operands are packed.
Fixed: tint:1121
Change-Id: I2e8f4302e08117ca41bac6c05fb24a70d1215740
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/62480
Kokoro: Kokoro <noreply+kokoro@google.com>
Auto-Submit: James Price <jrprice@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
MSL vectors with other widths already match WGSL's rules for alignment
and size.
Change-Id: I237052372463ea8323eab47c3b4ca90c6d8afcc3
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/62600
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Remap all resources into a flat namespace, to allow tests to pass when
multiple resources use the same binding number.
Fixed: tint:959
Change-Id: I58ed07c789e1ea90fc370ceba73b9d8292902549
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/61261
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: James Price <jrprice@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
This new test was added around the same time as the entry point IO
rework, so the (now incorrect) test output made it past the bots.
Change-Id: I89fc4041b9cd00cd363ba61d07371554263eca96
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/61460
Auto-Submit: James Price <jrprice@google.com>
Reviewed-by: Ryan Harrison <rharrison@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: James Price <jrprice@google.com>
SPIR-V spec states:
> Each structure-type member that is a matrix or array-of-matrices must have be decorated with a MatrixStride Decoration
As already pointed out in https://dawn-review.googlesource.com/c/tint/+/59840, we were not handling arrays-of-matrices.
To do this correctly, we need the ast::StrideDecoration to be placed on the Matrix type, which is a much bigger change to support.
For now, chase the type, and error if we have a custom MatrixStride on an array of matrices, otherwise drop the decoration.
Bug: tint:1049
Fixed: tint:1088
Change-Id: Idcb75b3df88040836a03a14e0ca402ebee7be9a7
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/60923
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: David Neto <dneto@google.com>
Commit-Queue: David Neto <dneto@google.com>
Auto-Submit: Ben Clayton <bclayton@google.com>
The refactored CanonicalizeEntryPointIO transform makes it much easier
to handle SPIR-V style IO as well, and doing this removes a lot of
duplicated code. Remove all of the SPIR-V transform code for shader IO
and vertex point size.
Bug: tint:920
Change-Id: Id1b97517619b4d2fd09b45d5aee848259f3dfa77
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/60840
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: James Price <jrprice@google.com>
Auto-Submit: James Price <jrprice@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
This is a major reworking of this transform. The old transform code
was getting unwieldy, with part of the complication coming from the
handling of multiple return statements. By generating a wrapper
function instead, we can avoid a lot of this complexity.
The original entry point function is stripped of all shader IO
attributes (as well as `stage` and `workgroup_size`), but the body is
left unmodified. A new entry point wrapper function is introduced
which calls the original function, packing/unpacking the shader inputs
as necessary, and propagates the result to the corresponding shader
outputs.
The new code has been refactored to use a state object with the
different parts of the transform split into separate functions, which
makes it much more manageable.
Fixed: tint:1076
Bug: tint:920
Change-Id: I3490a0ea7a3509a4e198ce730e476516649d8d96
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/60521
Auto-Submit: James Price <jrprice@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: James Price <jrprice@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
FXC errors on these, and they are undefined behavior in WGSL.
Bug: tint:1083
Change-Id: I7643fdc6991f8729f274535b603b761398412398
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/60500
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: David Neto <dneto@google.com>
Spread the array zeroing across as many workgroup invocations as possible.
Bug: tint:910
Change-Id: I1cb5a6aaafd2a0a4093ea3b9797c173378bc5605
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/60203
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: David Neto <dneto@google.com>
This reverts commit e5dbe24e94.
Reason for revert: Makes the Tint-Dawn roll fails because of
MSL compilation errors on as_type<uint>(-2147483648):
as_type cast from 'long' to 'uint' (aka 'unsigned int') is not allowed
as_type<uint>(-2147483647) compiles fine, so this is most
likely because the MSL compiler types the literal as a long
(since without the - it is larger than the max int32).
Original change's description:
> MSL writer: make signed int overflow defined behaviour
>
> Bug: tint:124
> Change-Id: Icf545b633d6390ceb7f639e80111390005e311a1
> Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/60100
> Kokoro: Kokoro <noreply+kokoro@google.com>
> Commit-Queue: Antonio Maiorano <amaiorano@google.com>
> Reviewed-by: David Neto <dneto@google.com>
TBR=dneto@google.com,bclayton@google.com,jrprice@google.com,amaiorano@google.com,noreply+kokoro@google.com,tint-scoped@luci-project-accounts.iam.gserviceaccount.com
Change-Id: I3e3384a9185013bb141a1b7b9b22bad8571bbc50
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: tint:124
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/60345
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Reviewed-by: Ben Clayton <bclayton@google.com>
Auto-Submit: Corentin Wallez <cwallez@chromium.org>
Commit-Queue: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
For loops only support assignments or function calls for the continuing statement.
Fixed: tint:1064
Change-Id: I07065b2119e7b9f97ca7e46b1464fd72333ca429
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/60212
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: David Neto <dneto@google.com>
The new vk-gl-cts tests have uncovered a whole bunch of FXC issues,
which have been filed as tint bugs.
Bug: tint:998
Bug: tint:1080
Bug: tint:1038
Bug: tint:1081
Bug: tint:1082
Bug: tint:1083
Change-Id: I0d14370f94647dfd9c7088e0b782c3b415c78ee7
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/60211
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
When building a vector via tint::writer::AppendVector, and the
vector argument is already a vector constructor, expand that
vector constructor into its components only when those components
are all scalars. This avoids a type breakage which can occur with cases
like this:
vector argument is:
vec2<i32>(vec2<u32>(0u,1u))
scalar argument is:
2
Before this fix, the result was:
vec2<i32>(0u, 1u, 2);
But should be this instead:
vec3<i32>(vec2<u32>(0u,1u),2)
This was noticed in SPIR-V writer output when forming a coordinate
vector from a an unsigned WGSL coordinate vector with a signed array
vector.
Fixed: tint:1048
Change-Id: Id46665739cc23da0ca58b9baabf7b4531b86350b
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/60040
Auto-Submit: David Neto <dneto@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: David Neto <dneto@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Also move:
test/fxc_bugs/vector_assignment_in_loop
to
test/bug/fxc/vector_assignment_in_loop
Change-Id: I7bbfc476fdb7a3296025609625e322fed8d16285
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/59444
Auto-Submit: Ben Clayton <bclayton@google.com>
Commit-Queue: James Price <jrprice@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: James Price <jrprice@google.com>
https://github.com/gpuweb/gpuweb/pull/1945 changes the SPIR-V mapping of this operator so that it now maps to OpFRem instead of OpFMod. Polyfill OpFMod with `x - y * floor(x / y)`
Also map the MSL output of this operator to use `fmod()`.
Behavior of this operator is now consistent across all backends.
Fixed: tint:945
Fixed: tint:977
Fixed: tint:1010
Change-Id: Iefa009b905989c55ace24e073ab0e261c7cf69b0
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/58393
Auto-Submit: Ben Clayton <bclayton@google.com>
Commit-Queue: Ben Clayton <bclayton@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: David Neto <dneto@google.com>
Change-Id: Ied931db182c11f26bffda4900641e41a1e9b5ee8
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/59027
Commit-Queue: Ben Clayton <bclayton@google.com>
Commit-Queue: James Price <jrprice@google.com>
Auto-Submit: Ben Clayton <bclayton@google.com>
Kokoro: Ben Clayton <bclayton@google.com>
Reviewed-by: James Price <jrprice@google.com>
The TypeConstructorExpression logic that tested for splats was not considering references. This led to broken emission for the SPIR-V and HLSL backends.
Fixed: tint:992
Change-Id: I9824b71f526997f91d380c09b459f4fd73065b19
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/58397
Auto-Submit: Ben Clayton <bclayton@google.com>
Commit-Queue: David Neto <dneto@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: David Neto <dneto@google.com>
Arrays can be extremely large, and having the load and store functions unroll the elements can make the complier explode.
Fixed: chromium:1229233
Change-Id: Ieb5654254e16f5ce724a205d21d954ef9a0cd053
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/58382
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@chromium.org>
Reviewed-by: David Neto <dneto@google.com>
Auto-Submit: Ben Clayton <bclayton@google.com>
For loop initializers and continuing statements do not have a BlockStatement as their parent.
Handle removal of these statements with a new Transform::RemoveStatement() helper
Fixed: tint:990
Change-Id: I24e7b18dcf71d3ef0a4d3ee68b9f68518e0eb5e8
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/58063
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: James Price <jrprice@google.com>
CloneContext::Replace(T* what, T* with) is bug-prone, as complex transforms may want to clone `what` multiple times, or not at all. In both cases, this will likely result in an ICE as either the replacement will be reachable multiple times, or not at all.
The CTS test: webgpu:shader,execution,robust_access:linear_memory:storageClass="storage";storageMode="read_write";access="read";atomic=true;baseType="i32"
Was triggering this brokenness with DecomposeMemoryAccess's use of CloneContext::Replace(T*, T*).
Switch the usage of CloneContext::Replace(T*, T*) to the new function form.
As std::function is copyable, it cannot hold a captured std::unique_ptr.
This prevented the Replace() lambdas from capturing the necessary `BufferAccess` data, as this held a `std::unique_ptr<Offset>`.
To fix this, use a `BlockAllocator` for Offsets, and use raw pointers instead.
Because the function passed to Replace() is called just before the node is cloned, insertion of new functions will occur just before the currently evaluated module-scope entity.
This allows us to remove the "insert_after" arguments to LoadFunc(), StoreFunc(), and AtomicFunc().
We can also kill the icky InsertGlobal() and TypeDeclOf() helpers.
Bug: tint:993
Change-Id: I60972bc13a2fa819a163ee2671f61e82d0e68d2a
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/58222
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: James Price <jrprice@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
This uses FXC compilation failure mitigation for _any_ vector index assignment that has a non-constant index. FXC can still fall over if the loop calls a function that performs the dynamic index.
Use some vector swizzle logic to avoid branches in the helper.
Fixed: tint:980
Change-Id: I2a759d88a7d884bc61b4631cf57feb4acc8178de
Reviewed-on: https://dawn-review.googlesource.com/c/tint/+/57882
Auto-Submit: Ben Clayton <bclayton@google.com>
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Antonio Maiorano
Stephen White
Ben Clayton
James Price
David Neto
Corentin Wallez