#6055 Optimize LLImageGL::analyzeAlpha

[akleshchev]

Some of the image get created on a thread, but some large images (uploads, picker) get created on main thread and do alpha analyzis. Alpha calculations are pricey, so when large textures goes through main thread, move alpha calculations to a worker thread.

[Copilot]

Pull request overview

This PR optimizes LLImageGL::analyzeAlpha() by moving expensive alpha-mask analysis off the main thread for large textures, reducing main-thread stalls during texture creation paths that still run on the main thread (e.g. uploads / picker).

Changes:

• Refactors the alpha-mask histogram logic into a new static helper (analyzeAlphaData()).
• Defers alpha analysis to a WorkQueue worker when running on the main thread for larger images, applying results back on the main thread.
• Adds a request-serial (mAlphaAnalysisSerial) to ignore stale async results and invalidates pending jobs on certain state changes.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File	Description
indra/llrender/llimagegl.h	Adds the new static helper declaration and introduces mAlphaAnalysisSerial for async request tracking.
indra/llrender/llimagegl.cpp	Implements the helper and adds async deferral + request tracking/invalidation in alpha analysis flow.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (2)

indra/llrender/llimagegl.cpp:2362

• Profiler label typo: "Deffered" -> "Deferred".

            LL_PROFILE_ZONE_NAMED("Deffered alpha mask analysis");

indra/llrender/llimagegl.cpp:2388

• In the synchronous fallback, use the captured alpha_offset/alpha_stride rather than reading mAlphaOffset/mAlphaStride again. This keeps the analysis inputs consistent with what was captured for the request (and avoids potential races if those members are changed before this line executes).

        delete[] data_copy;
        mIsMask = analyzeAlphaData(data_in, w, h, mAlphaOffset, mAlphaStride);
        unref();

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.

Suppressed comments (1)

indra/llrender/llimagegl.cpp:2362

• Profiler zone label has a spelling error ("Deffered"). This makes profiling views harder to search/aggregate consistently.

            LL_PROFILE_ZONE_NAMED("Deffered alpha mask analysis");