Arcade Game Zone (PPSA19015)

[JAB-dev]

Existing thread

• I searched this category for the title ID and no thread exists for this title on any platform.

Title

Arcade Game Zone (PPSA19015)

Content version tested

01.003.000

AnyPS5 commit

main bf2c63f (committed 2026-10-19 20:52 UTC) for the stop lines below, alone and with the two pull requests named under Current blocker on top. The in-game runs used a local build: main c4c2b3a with earlier versions of those two pull requests (without the refusals described below), plus the open pull requests and local changes listed under Current blocker.

System tested

Windows 11 x64, GCC 15.2 (WinLibs), NVIDIA GeForce RTX 4070 SUPER (driver 610.88). Linux not tested.

Furthest point

In game

What is shown

On main the title stops after 13.9 s (see Current blocker). Everything below is from the local build.

The publisher logo, the title screen, the join screen, the main menu, the mini-game carousel and the versus setup. Mini-games reached:

• Air Hockey: a full match against a bot, to the results screen.
• Basketball: a four-player bot match, to the results screen.
• Table Football: a live match against a bot.
• Darts: the 301 screen, waiting for input.
• Pool: the table with the aim line, waiting for input.
• Ball-Hop Bowling: the lane, waiting for input.
• The 8th and 9th carousel entries: each played to its results screen.
• The 10th carousel entry, a racing game: ran for over 100 s, with the race timer advancing from 01:14 to 01:29 in the frame dumps.

These come from successive local builds; not every mini-game has been rerun on the newest one. Keyboard input reaches the game. In the attached footage the window title's FPS counter reads about 8 in the menus and about 60 in the mini-game.

Earlier local builds stopped intermittently with a depth-tracking refusal at the first scene change, from the main menu into a mini-game (about 1 launch in 5). On the current local build 30 consecutive launches made that change without stopping.

Gameplay footage of the local stack is attached below.

Current blocker

On main the title stops after 13.9 s with AGC graphics: unsupported color format 19 number type 0 component swap 2. The title renders into a 4_4_4_4 UNORM colour target, and the RTX 4070 SUPER (driver 610.88) exposes none of the four Vulkan 4-bit formats as a colour attachment.

Pull requests for this title

PR	What it does	Where the title stops
main (no PR)		After 13.9 s: AGC graphics: unsupported color format 19 number type 0 component swap 2
#8563	Renders 4_4_4_4 UNORM colour targets through packed R16_UINT exports	After 14.5 s: AGC graphics: partial writes of a 4_4_4_4 unorm color target are unsupported. The draw it stops at writes only the top nibble of a 2-sample target.
#8564 (depends on #8563)	Merges partial writes into single-sample and multisampled 4_4_4_4 targets under pixel interlock	After 13.7 s, at the first 2-sample merge draw: a refusal, because that pixel-rate shader reads PERSP_CENTER and POS_Z_FLOAT on a 2-sample rasterizer (full text below)

The full stop line with #8564:
AGC graphics: partial writes of a multisampled 4_4_4_4 unorm color target run the pixel shader once per sample, so a shader the guest runs once per pixel cannot read the ancillary word or, unless every sample sits at the pixel centre, centroid, sample-located or pull-model barycentrics, Z, 1/W or a position at the centroid or a sample: they would follow the invocation's sample (SPI_PS_INPUT_ENA 0x402, SPI_PS_INPUT_ADDR 0x402, SPI_BARYC_CNTL 0x1000000, sample 1 location 0xc4)

To go further, someone needs to measure where the console evaluates Z for a pixel-rate shader under MSAA.

Open upstream pull requests the local build also carries

PR	What it does
#6300	Imports cleared native multisample D32S8 targets and whole HTILE fills
#6808	Only its depth part; the colour-guard change was left out because it fails a main test
#6727	Retires an unwritten native multisample depth surface on guest access

Two existing tests from the other two pull requests are adjusted locally to go with #6727:

• feat(agc): reuse a native multisample colour target and honour guest HTILE clears #6808's test of a guest read over a freshly bound, unwritten multisample depth surface now expects the surface to be retired instead of refused.
• feat(agc): import cleared native multisample D32S8 targets and whole HTILE fills #6300's D32S8 stencil-target test marks the surface written before its stencil-plane reads, so those reads still refuse.

Local changes, not proposed (some are inexact, so none can be upstreamed as they are):

• Before the refusal above existed, the merge let that shader read Z at each sample's position. Inexact.
• A merge that kills pixels under late tests in a depth-writing draw is drawn as a depth prepass followed by an EQUAL merge pass. Inexact on depth ties.
• A colour target imported over a host-written multisample depth surface retires that surface. Inexact: the colour target starts from stale guest bytes.
• A multisample depth surface is retired once the guest has overwritten the range being read. Driver writes are tracked per byte, CPU writes per 4 KiB page.
• A guest read that straddles the edge of a host-written depth or HTILE range retires the surface when every byte inside the range was overwritten since; otherwise it still throws. Local, not proposed.
• A depth range is re-watched when a live host import still covers it.
• Clean cached storage images over a depth range that lost write tracking are evicted before that re-watch. Local, not proposed.
• Texture reads of host-written depth ranges are checked against their texel footprint.
• A pending host import is settled before a depth import uses it.

How it was run

Converted with relinker --windows --module-dir Media/Modules --module-dir Media/Plugins app0/eboot.elf app.exe, then run as app.exe. The stop lines were run with ANYPS5_NO_SHADER_CACHE=1 and no other switches, with a fresh conversion and libs/ for each build; the seconds run from launch to exit. The mini-game runs and the 30 launches drove the menus with key messages sent to the window; the mini-game runs also set APS5_DUMP_FRAMES, APS5_DUMP_FRAMES_EVERY and APS5_DUMP_SCALE to save frames.

[JAB-dev]

AGZ-playthrough-edit.mp4