[windows]: adjusts uniform buffers to hit hlsl optimization

[gaaclarke]

issue #188348

This removes the following class of HLSL warnings:

WARN: ProgramD3D.cpp:711 (rx::ProgramD3D::linkJobImpl): Uniform block 'FragInfo' will be slow to compile. See UniformBlockToStructuredBufferTranslation.md (https://shorturl.at/drFY7) for details.

By putting large uniform arrays into their own uniform buffer the HLSL compiler in ANGLE is able to optimize the compilation using a StructuredBuffer. This removes the warning about long compilation time.

testing: refactor, no logical change

Pre-launch Checklist

• I read the Contributor Guide and followed the process outlined there for submitting PRs.
• I read the AI contribution guidelines and understand my responsibilities, or I am not using AI tools.
• I read the Tree Hygiene wiki page, which explains my responsibilities.
• I read and followed the Flutter Style Guide, including Features we expect every widget to implement.
• I signed the CLA.
• I listed at least one issue that this PR fixes in the description above.
• I updated/added relevant documentation (doc comments with ///).
• I added new tests to check the change I am making, or this PR is test-exempt.
• I followed the breaking change policy and added Data Driven Fixes where supported.
• All existing and new tests are passing.

If you need help, consider asking for advice on the #hackers-new channel on Discord.

If this change needs to override an active code freeze, provide a comment explaining why. The code freeze workflow can be overridden by code reviewers. See pinned issues for any active code freezes with guidance.

Note: The Flutter team is currently trialing the use of Gemini Code Assist for GitHub. Comments from the gemini-code-assist bot should not be taken as authoritative feedback from the Flutter team. If you find its comments useful you can update your code accordingly, but if you are unsure or disagree with the feedback, please feel free to wait for a Flutter team member's review for guidance on which automated comments should be addressed.

[gaaclarke]

This looks good, we need to do this for KernelSamples too.

[gemini-code-assist]

Code Review

This pull request refactors several shaders and their corresponding host-side rendering code to move large arrays (such as colors, stop pairs, and kernel samples) into dedicated uniform blocks, preventing size and alignment issues on certain graphics platforms. Feedback recommends passing the temporary returned by GenerateBlurInfo directly to LerpHackKernelSamples to enable copy elision and avoid copying a large struct.

[walley892]

LG as long as the goldens don't change

[b-luk]

LGTM. Just a couple of nits, which I'll leave up to you to decide if you think it's anything worth changing.

[b-luk]

This is no longer true. GaussianBlurPipeline::FragmentShader::KernelSamples no longer contains sample_count, so this struct is no longer a mirror of it.

[gaaclarke]

I think the comment is still helpful despite that technicality since it convey's the relationship between them. I think it's better to keep it as-is.

[b-luk]

I think this is badly named, and it makes the code a bit harder to follow. I don't know the best way to improve it, but I'd be in favor of even just removing it entirely and using a pair<int32_t, GaussianBlurPipeline::FragmentShader::KernelSamples>.

[gaaclarke]

The name of the function that generates this is LerpHackKernelSamples. I wouldn't want to switch to unstructured data since we'd be losing named fields. It's common to have a struct declared for returning multiple values. We could name it LerpHackKernelSamplesResult to be more clear if you like. Dropping some of the name didn't seem to affect the readability I my eyes, let me know.

[b-luk]

Maybe it's not exactly a naming issue. I think what makes it harder to follow is that we no longer have a single struct that directly represents the shader input. The behavior used to be:

1. GenerateBlurInfo() generates KernalSamples, which is an expanded version of the FragmentShader::KernelSamples shader input.
2. LerpHackKernelSamples() converts the expanded KernalSamples into the input FragmentShader::KernelSamples, which is bound to the shader.

So there's a single KernalSamples struct which directly maps to a single shader input.

Now it's:

1. GenerateBlurInfo() generates KernalSamples, which is an expanded version of (FragmentShader::FragInfo::sample_count + FragmentShader::KernelSamples).
2. LerpHackKernelSamples() converts KernalSamples into LerpHackResult, which is the non-expanded version of (FragmentShader::FragInfo::sample_count + FragmentShader::KernelSamples).
3. The two parts of LerpHackResult get split up to be bound to two different inputs in the shader.

We end up with two intermediate KernalSamples and LerpHackResult structs, and neither of them map directly to a shader input field.

I don't know a good way to clean this up, so I'm not blocking submission. The new version isn't super complex. Once you look at it enough you can see what's going on. But I think it does force the reader to juggle a bunch more types/fields/subfields in their head to understand the flow of data compared to the old version. And for many of the new fields it's not immediately clear/intuitive what they represent within the data flow.

[gaaclarke]

yea, it's less clear, our hand was forced based on hlsl's requirements. It looks like c++26 will have std::inplace_vector which may clean this up a bit. We won't need the struct anymore after that.

[auto-submit]

autosubmit label was removed for flutter/flutter/188538, because - The status or check suite Linux linux_unopt has failed. Please fix the issues identified (or deflake) before re-applying this label.

[b-luk]

Maybe it's not exactly a naming issue. I think what makes it harder to follow is that we no longer have a single struct that directly represents the shader input. The behavior used to be:

1. GenerateBlurInfo() generates KernalSamples, which is an expanded version of the FragmentShader::KernelSamples shader input.
2. LerpHackKernelSamples() converts the expanded KernalSamples into the input FragmentShader::KernelSamples, which is bound to the shader.

So there's a single KernalSamples struct which directly maps to a single shader input.

Now it's:

1. GenerateBlurInfo() generates KernalSamples, which is an expanded version of (FragmentShader::FragInfo::sample_count + FragmentShader::KernelSamples).
2. LerpHackKernelSamples() converts KernalSamples into LerpHackResult, which is the non-expanded version of (FragmentShader::FragInfo::sample_count + FragmentShader::KernelSamples).
3. The two parts of LerpHackResult get split up to be bound to two different inputs in the shader.

We end up with two intermediate KernalSamples and LerpHackResult structs, and neither of them map directly to a shader input field.

I don't know a good way to clean this up, so I'm not blocking submission. The new version isn't super complex. Once you look at it enough you can see what's going on. But I think it does force the reader to juggle a bunch more types/fields/subfields in their head to understand the flow of data compared to the old version. And for many of the new fields it's not immediately clear/intuitive what they represent within the data flow.

[flutter-dashboard]

CI had a failure that stopped further tests from running. We need to investigate to determine the root cause.

SHA at time of execution: 87c4efc.

Possible causes:

• Configuration Changes: The .ci.yaml file might have been modified between the creation of this pull request and the start of this test run. This can lead to ci yaml validation errors.
• Infrastructure Issues: Problems with the CI environment itself (e.g., quota) could have caused the failure.

A blank commit, or merging to head, will be required to resume running CI for this PR.

Error Details:

FormatException: ERROR: Linux test_ownership is a new builder added. it needs to be marked bringup: true
If ci.yaml wasn't changed, try `git fetch upstream && git merge upstream/master`

Stack trace:

#0      CiYaml._validate (package:cocoon_service/src/model/ci_yaml/ci_yaml.dart:395:7)
#1      new CiYaml (package:cocoon_service/src/model/ci_yaml/ci_yaml.dart:128:7)
#2      new CiYamlSet (package:cocoon_service/src/model/ci_yaml/ci_yaml.dart:48:23)
#3      CiYamlFetcher._getCiYaml (package:cocoon_service/src/service/scheduler/ci_yaml_fetcher.dart:124:12)
<asynchronous suspension>
#4      Scheduler.getPresubmitTargets (package:cocoon_service/src/service/scheduler.dart:1084:20)
<asynchronous suspension>
#5      Scheduler._getTestsForStage (package:cocoon_service/src/service/scheduler.dart:1368:14)
<asynchronous suspension>
#6      Scheduler._runCiTestingStage (package:cocoon_service/src/service/scheduler.dart:1408:30)
<asynchronous suspension>
#7      Scheduler.proceedToCiTestingStage (package:cocoon_service/src/service/scheduler.dart:1516:7)
<asynchronous suspension>
#8      Scheduler._closeSuccessfulEngineBuildStage (package:cocoon_service/src/service/scheduler.dart:1341:5)
<asynchronous suspension>
#9      Scheduler.processCheckRunCompleted (package:cocoon_service/src/service/scheduler.dart:1273:11)
<asynchronous suspension>
#10     PresubmitLuciSubscription.post (package:cocoon_service/src/request_handlers/presubmit_luci_subscription.dart:176:9)
<asynchronous suspension>
#11     RequestHandler.service (package:cocoon_service/src/request_handling/request_handler.dart:42:20)
<asynchronous suspension>
#12     SubscriptionHandler.service (package:cocoon_service/src/request_handling/subscription_handler.dart:134:5)
<asynchronous suspension>
#13     createServer.<anonymous closure> (package:cocoon_service/server.dart:448:7)
<asynchronous suspension>
#14     main.<anonymous closure>.<anonymous closure> (file:///app/app_dart/bin/gae_server.dart:187:9)
<asynchronous suspension>

[auto-submit]

autosubmit label was removed for flutter/flutter/188538, because - The status or check suite Linux linux_fuchsia_tests has failed. Please fix the issues identified (or deflake) before re-applying this label.

[gaaclarke]

this may benefit #188338