#5356 Fix fast texture cache's mutex stall #2

[akleshchev]

First attempt didn't work well enough, prelocking mutex to make sure main thread gets priority and does not wait on the same mutex multiple times per frame.

[Copilot]

Pull request overview

This PR aims to reduce main-thread stalls in the fast texture cache path by prelocking cache-related mutexes so the main thread doesn’t repeatedly contend on the same locks multiple times per frame.

Changes:

• Prelock texture cache fast-cache and header ID map mutexes while draining mFastCacheList on the main thread.
• Expose LLTextureCache mutex accessors to support external prelocking.
• Add additional profiling zones in LLMeshRepoThread::notifyLoadedMeshes() and clarify comments around fast-cache image construction.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
indra/newview/llviewertexturelist.cpp	Prelocks texture cache mutexes while processing mFastCacheList to reduce per-frame lock contention.
indra/newview/lltexturecache.h	Adds public accessors for internal mutexes and documents a lock-ordering constraint.
indra/newview/lltexturecache.cpp	Comment-only clarifications around fast-cache open lifetime and LLImageRaw ownership.
indra/newview/llmeshrepository.cpp	Adds named profiling zones around mesh notification processing blocks.

---

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