This patch add f16 support for a major part of numeric built-in, and
implement corresponding unittests for resolver and backends. This patch
also enable f16 constant evaluation for unary minus operator, `atan2`
and `clamp`.
The following numeric built-ins are not supported yet:
* frexp
* modf
The end-to-end tests for f16 built-in are not added yet.
Bug: tint:1473, tint:1502
Change-Id: If807185617b21c510a1a9c371179a60800c4f875
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/96722
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: Zhaoming Jiang <zhaoming.jiang@intel.com>
* Ensure that copyright years are preserved when regenerating.
* Add 'Split' function which maps to strings.Split.
* Add 'Scramble' function with messes with a string.
* Add 'Import' function which allows templates to be imported, allowing
for reusable macros.
Change-Id: Ib77f59a989cf55addcced3e337c9031062a83470
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/97149
Commit-Queue: Dan Sinclair <dsinclair@chromium.org>
Reviewed-by: Dan Sinclair <dsinclair@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Previously enum members were added to the global namespace, so that
overload template parameters could be constrained to a single
enum-entry without the need to declare a matcher. While this was a minor
convenience feature, it means that you cannot declare an enum with
members that share the same name as a type. This will be very common for
extensions, like 'f16' where 'f16' is the name of an extension and a
type name.
Change scoping so that enum members need to be fully qualified. Also
change the intrinsic syntax so that enums always need to use a matcher
for enums.
Change-Id: Ided91130e9df537d38dc8ecb41325c0992dea14b
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/97146
Reviewed-by: Dan Sinclair <dsinclair@chromium.org>
Commit-Queue: Ben Clayton <bclayton@google.com>
Change tint's `--fxc` flag to take the path of the FXC compiler DLL.
Have tint attempt to validate with both FXC and DXC if `--validate` is
passed.
Fix the 'dirsWithNoPassExpectations' logic which looks like it got
broken with the tint -> dawn merge. It also incorrectly applied
filepath.FromSlash() on windows.
Change-Id: I0f46aa5c21bc48a2abc48402c41f846aff4a8633
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/96800
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Commit-Queue: Ben Clayton <bclayton@chromium.org>
This patch normalize the abstract interger and abstract float type name
to `ia` and `fa` and the type matcher name to `fia_fiu32_f16` format in
intrinsics.def, and also rename `aiu32Dispatch` and `afi32f16Dispatch`
in resolver/const_eval.cc to `Dispatch_ia_iu32` and
`Dispatch_fia_fi32_f16`.
Bug: tint:1581
Change-Id: Ia48e150d6028e4a7b999f72e57508df369b3e2d7
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/96721
Reviewed-by: 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>
Use the @const annotations to more efficently dispatch to the right
method of the ConstEval class.
Reduces a whole lot of dynamic casting logic.
Change-Id: I6791aac51b935b46d63af29abd1e577b3306a0ff
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/95950
Commit-Queue: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Always report the Javascript duration from the page instead of
the total duration from ResultDB. Reporting of the total duration
could be added back in the future, but usages of the tooling are
currently interested in the Javascript duration.
Bug: chromium:1336333
Change-Id: I74e91e7a9c29f3a4c45ed67898cc6f2de62e19a4
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/94762
Reviewed-by: Ben Clayton <bclayton@google.com>
Commit-Queue: Austin Eng <enga@chromium.org>
These sources need to be generated in Dawn's source tree
so that Chromium's //third_party/webgpu-cts DEP can be
removed.
Bug: chromium:1333969
Change-Id: I03c8cba691bcbfac00839f0cdd40fab6198ec83f
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/91060
Commit-Queue: Austin Eng <enga@chromium.org>
Reviewed-by: Ben Clayton <bclayton@google.com>
Add support for @const to builtins in intrinsics.def.
Propagate this flag through to the intrinsic table.
Handle builtins that are @const annotated in the resolver.
Currently no intrinsics are decorated with @const, so there's nothing to
test (yet).
Bug: tint:1504
Change-Id: I172483688617782bd7c58b70e3f38d0222a5d1af
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/92323
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Add support for a [[precedence(N)]] decoration on intrinsic table type
declarations. This will be used to ensure the type with the lowest
conversion rank is matched when a matcher could match multiple types
for a given abstract numeric argument type.
Bug: tint:1504
Change-Id: I96475b000c0917bbfa4e2873b1731ce048b96a7d
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/90664
Reviewed-by: David Neto <dneto@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
The concept of 'closing' an open type or number made sense when these
were immutable once the first type/number had been matched during
overload resolution.
In order to support abstract numerics, these template parameters need to
be constrained as the arguments are evaluated, so there's no longer a
binary open / closed state.
Give up on this concept, and rename everything to 'template type' and
'template number'. This is likely easier for people to understand
anyway.
Also fix a small typo in the ICE message printed when there's an
ambiguous overload resolution (should never happen with the current
entries in the table).
Bug: tint:1504
Change-Id: I2bf043c71e5afa757259968eae4af830c50f38e0
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/90662
Commit-Queue: Ben Clayton <bclayton@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Reviewed-by: David Neto <dneto@google.com>
These are currently not used, but the first step towards moving type
constructors and converters over to using the intrinisc table.
This will simplify maintenance of type functions, and will greatly
simplify the [AbstractInt -> i32|u32] [AbstractFloat -> f32|f16] logic.
Bug: tint:1504
Change-Id: I15526670a6ff801e66551ab5adc37b1570ac49de
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/90242
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Instead of Failure.
If we have duplicate results (same query, same tags), but with different results, then we have to merge them. Previously we'd merge Pass and Slow as Failure, which was just odd. Fix this so the combined result is Slow.
Bug: dawn:1342
Change-Id: If9b035bb3c1e59b7a93a5c8b194f854405e656d6
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/89960
Reviewed-by: Corentin Wallez <cwallez@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Corentin Wallez <cwallez@chromium.org>
Auto-Submit: Ben Clayton <bclayton@google.com>
• Declare all the unary ops in the intrinsics.def file.
• Reimplement the bulk of Resolver::UnaryOp() with the IntrinsicTable.
This will simplify maintenance of the operators, and will greatly
simplify the [AbstractInt -> i32|u32] [AbstractFloat -> f32|f16] logic.
Bug: tint:1504
Change-Id: Ifc646d086fc93cfbe3f3f861b8c447178664c1f7
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/89028
Reviewed-by: James Price <jrprice@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
Commit-Queue: Ben Clayton <bclayton@chromium.org>
Adapt the builtin parsing and resolving to also support operators.
Will be used to generate intrinsic table entries for operators.
This will simplify maintenance of the operators, and will greatly
simplify the [AbstractInt -> i32|u32] [AbstractFloat -> f32|f16] logic.
Bug: tint:1504
Change-Id: Id75735ea24e501877418812185796f3fba88a521
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/89026
Commit-Queue: Ben Clayton <bclayton@chromium.org>
Kokoro: Kokoro <noreply+kokoro@google.com>
Reviewed-by: Dan Sinclair <dsinclair@chromium.org>
• Rename 'builtin-gen' back to 'intrinsic-gen', as 'intrinsics' now
include both builtins and operators.
• Move the intrinsic definitions, and IntrinsicTable to the resolver
package, where it belongs.
Bug: tint:1504
Change-Id: I5ad5c285c1e360a224ee1235c293ccd55eb2693d
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/89025
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Commit-Queue: Ben Clayton <bclayton@google.com>
• Add the included trybots in the CL description. All of these trybots are tested by the roll, but the final CQ-submit wouldn't necessarily test all of the variants before landing. This would mean that the 'cts export' could miss some results, as it takes the last PS with any results.
• Add --force flag to cts roll to force a roll. Useful for testing.
• Emit timing diagnostics for tests labelled 'Slow' instead of unhelpfully stating they pass.
• Enable the --cl and --ps flags for cts export
• Export with the most recent data to the top of the spreadsheet
Bug: dawn:1401
Change-Id: Id926367ab805bfb9f3032fce9cce7f00daf7a5d4
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/88661
Reviewed-by: Dan Sinclair <dsinclair@chromium.org>
Commit-Queue: Ben Clayton <bclayton@google.com>
Kokoro: Kokoro <noreply+kokoro@google.com>
• Strip 'release-x64' tags.
These are always paired with 'release', and all
other tests just have the 'release' tag. This
means that results with the 'release' +
'release-x64' tags are a subset of just the
'release' tags, leading to many broken
assumptions in the expectation update.
• Don't consider tests with the error reason
'asyncio.exceptions.TimeoutError' as slow, but
as a failure. The CTS expectation updater will
already mark tests exceeding the config
threshold as slow. This reduces the amount of
data requested from resultdb, and prevents
failed rolls when results are actually
deadlocking.
• Fix inferring of patchset in GetResults().
LatestPatchest() was called but was assigned to
a nested scope.
Bug: dawn:1401
Change-Id: I57b2c9a029b31430d63a056f7b13c4bf1bc5b437
Reviewed-on: https://dawn-review.googlesource.com/c/dawn/+/88450
Reviewed-by: Dan Sinclair <dsinclair@chromium.org>
