[mrfelek]

What

Three additions to docs/dev/GAME_TESTING.md (#6627), from runs since it merged:

• Memory budgets (under "Check the GPU and the caches"). The host import budget and the texture cache budgets come from each machine's memory and can change a title's frame rate several times over, so results from two machines don't compare until these are known. The section names the log lines to read and report: [gpu] host import budget: ... refused so far (printed only when an import is refused, at most every 10 s), [gpu] sampled texture cache budget, and the [texture-cache] (10 s) line with APS5_PROFILE_DRAW=1. It links APS5_HOST_IMPORT_MIB (ENVIRONMENT) and APS5_TEXTURE_CACHE_MIB (USAGE, Texture cache budget). Examples from the discussions: ASTRO BOT is about 2.5× slower on 32 GiB machines at fix(agc): keep half the host heap out of the host import budget #6472's import budget (EdouardSence on Astro Bot (PPSA21564) #357 and fix(agc): keep at most 4 GiB of the host heap from the host import budget #7844), and Sky Garden on a 12 GB card loses about two thirds of its frame rate at the 2048 MiB texture budget floor (Astro Bot (PPSA21564) #357). "Reporting a result" item 6 now asks for these lines too.
• Sleeping display (under "Power and display"): a title started while the display sleeps stops with SDL_GetWindowDisplayIndex failed: Couldn't find any displays (VideoOutDriver.cpp); keep the display awake for long or unattended runs.
• tools/check_main_sync.py --fetch (under "Before a run"): reports whether the branch is in step with upstream main, so a result isn't taken on an old base by mistake.

Tested

Not tested beyond the docs: tools/check_conventions.py --base origin/main passes (it checks the links), and tools/pr_overlap.py reports no overlap. The log line formats are copied from GuestBufferMemory.cpp and ShaderResources.cpp on main, and the display message from VideoOutDriver.cpp. The memory-budget lines and the display stop were seen on Linux, RTX 3080 Ti, in ASTRO BOT (PPSA21564) and Diablo II: Resurrected (PPSA03912) runs this week.

Checklist

• Based on current main; no other open PR implements the same functions
• One topic per PR; follow-ups go in a new PR
• No comments in code except technical debt
• Unimplemented paths throw (NotImplemented_nid_no_patch); silent stubs are listed under Silent stubs in their library's technical debt file
• No notes, investigation .md files or images added to the repository (attach them to this PR)
• New third-party code is a submodule built from source
• Generic behaviour, not specific to one title
• Depends on: none
• Fixes: none
• Unblocks: none
• GPUs tested: n/a
• AI-assisted: yes

AI-assisted; the runs are from my machine(s).

[oneandonlydean]

Thanks, this is useful: the budget lines and when to report them belong in the guide. One change, please: we don't name specific titles in the docs, so could you drop the sentence with the game and level example? The general description of thousands of evictions per 10 s at the 2048 MiB floor already says it. Per-game findings go in the Game bring-up discussions.

[mrfelek]

@oneandonlydean done in bb864e8: the sentence with the title example is gone; the general description of the evictions at the 2048 MiB floor stays.

AI-assisted; the runs are from my machine(s).

[MarcZX10]

The additions match current source and documentation: the host-import refusal line is emitted only from the refusal path, texture-cache budget lines are produced by the documented profiling path, and the sleep-related SDL display-index error is formed at the VideoOut call site. The check_main_sync.py --fetch command exists and fetches the selected remote-tracking main before reporting ahead/behind/conflicts. The maintainer’s request to remove title-specific examples from the guide has been addressed. Conventions, eligibility and labels pass; the product jobs are correctly skipped for this docs-only change. I reran the overlap checker against the current open PR queue (1,105 PRs); it reports no overlap. No documentation defect found; approved.

[oneandonlydean]

@mrfelek thanks for dropping that sentence. The guide reads well now, and every variable and log line it names still matches current main. I've passed it on for merge.