[pull] main from gpuweb:main #11
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
![pull[bot]](https://avatars.githubusercontent.com/in/12910?s=60&v=4){kind=small}
It's not guaranteed that the -0.0 value will survive once it is stored in a Scalar object Issue: #2901
* Prevent details toggle if user is selecting text * select all on click
This refactors the existing code to have a clearer separation from the non-AF test running code, and sets up for implementing vector support. Issue #1626
* Remove orientation from several texture copy tests The orientation argument for these tests was copied from tests which use an ImageBitmap source. Because CreateImageBitmap has an imageOrientation option it made sense to test against all variants of it. However, other tests, such as those that consume ImageData directly, don't have any native mechanism for Y-flipping the source. The data could be generated Y-flipped, but that doesn't increase coverage of any implementation features. Despite not flipping the source orientation, however, the test was still expected the data to be flipped. By removing this argument and no longer testing for flipped source data, these tests all begin passing with no loss of platform coverage. Bug: dawn:2017 * Lint fix
This is run by `npm test`, and should catch cache file collision issues which have caused reverts in the past.
This may be a temporary change if copying compressed textures get removed from core and moved to a feature.
This PR add execution tests for f16 built-in clamp. Issue: #1248
…2922) Matrix addition will be covered in future PRs. Issue #1626 Co-authored-by: jzm-intel <[email protected]>
…float32-filterable formats. (#2915) * Adds magFilter op tests * Adds inital minFilter tests * Adds remaining sampling tests * Fix formatting issue * Redo min/magFilter tests for mirror, and adds more doc.
And fix state tracking so that subcases are actually marked skipped in the logger
Add missing entry to `listing_meta.json`
Issue #1626 Co-authored-by: jzm-intel <[email protected]>
* depthCompare is not required for depth attachments if not used * Refactor for success * Refactor success definition
These were landed on the wrong branch, and are causing tests to fail to build, because some of the code that they depend on is implemented in a PR that hasn't landed yet.
Rewrites how test cases are generated for atan2, so that if running in const-eval unbounded results will not be generated, since those will cause compilation errors. Fixes #3088
This is only defined for f32, so doesn't really need to be defined in the common super class. This allows for removing the various stub references to it, that will never be implemented.
This was accidentally removed in 30c129e
* Slim down on typed array allocations in conversion.ts * Add comments explaining requirements of aliased working data
This removes a the need to create bunch of temporary JSON objects, reducing the amount of garbage collection we need to do. This change also changes the DataCache to be unbounded to a 4-element LRU cache, capping the amount of memory used.
Unlike doxygen, TSDoc doesn't support @p to link to a parameter. Use code backticks instead.
* Use DataView instead of a bunch of separate typed arrays. * Avoid small allocations where it's trivial to do so. Speeds up deserialization around ~10% based on profiling in Chrome.
Reduces thh number of cases by using sparse instead of full ranges, since there is going to be a cartesian product of input values when generating cases. Optimizes two quantization functions that had not been updated to re-used their TypedArray. Creation and then immediate destruction of TypedArrays are a type of hotspot we have encountered in other areas of the code base.
…ents tests (#3104) This patch uses the largest value of maxInterStageShaderVariables supported on current adapter in the tests about maxInterStageShaderComponents when creating devices so that when the value we use as maxInterStageShaderComponents is larger than the default one, we won't be limited by the default value of maxInterStageShaderVariables. This patch also removes the assertion that the value of maxInterStageShaderVariables must be larger than a quarter of maxInterStageShaderComponents as on many backends the largest value of maxInterStageShaderComponents is equal to 4x maxInterStageShaderVaraibles, so in "overLimit" tests the value of maxInterStageShaderComponents can be greater than 4x device.limits.maxInterStageShaderVaraibles.
To match all the other data types
* dev_server: serve on localhost only by default * Limit characters in route for /out/*/listing.js
Instead of passing the input through a F16Array, use the library provided function hfround. hfround is a fast look up table based rounding function for f16. Benchmarking locally this provides a ~20% improvement to fma interval calculations, which are particularly sensitive to quantization cost. Overall I was seeing more on the order of ~10% improvement.
Instead of passing the input through a F32Array, use the builtin Math.fround. This leads to a ~5% improvement benchmarking locally. This is less than the equivalent f16 change, because F32Array is provided by the runtime, whereas F16Array is being polyfilled, so is probably more efficient to begin with.
buffer() was offseting the array instead of truncating the returned array.
Another small bump (~5%) to be gained through using a builtin instead of trampolining through a TypedArray.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
See Commits and Changes for more details.
Created by
pull[bot]
Can you help keep this open source service alive? 💖 Please sponsor : )