DEATHLOOP (PPSA01670)

[BunnyFl]

Existing thread

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

Title

DEATHLOOP (PPSA01670)

Content version tested

01.405.000

AnyPS5 commit

main 6a8f224 with two open pull requests merged on top (local merge 08f682d84b, no conflicts): #10414 (head d963d44) and #10507 (head f34c280). main alone, and main with #9694 (793e51a), were also run, see Current blocker.

System tested

Windows 11 Pro 10.0.26200, AMD Radeon RX 9070 XT (driver 32.0.32015.7029), Ryzen 7 5800X3D, GCC 15.2.0 (WinLibs 15.2.0posix-14.0.0-ucrt-r7), Release.

Furthest point

In game

What is shown

Three runs of 600 s each, from no save data, with scripted pad input (Cross, Cross, Cross, Circle, one press every 3 s) and APS5_FORCE_SDR=1. In all three, the dumped frames (one every 100 presents) show the Bethesda logo, the copyright and photosensitivity screens, the title card (Press ✕) and the main menu (BREAK THE LOOP / PROTECT THE LOOP), then the loading screen, the opening cutscene and first-person play on the beach, with subtitles. Some runs' dumps also caught the Arkane logo and the subtitle options screen. Two runs also reached the The Longest Day objective screen. No run stopped, and no log has an is not implemented line.

• Frame rate: in game (each run's last 50 s), 11-16 presents per 10 s. Per frame the log reports 96-129 ms of GPU work and 514-835 ms of GPU idle gaps.
• HDR display: this monitor reports HDR (720 nits). Without APS5_FORCE_SDR=1, the title presents in HDR10 and shows a different calibration screen (three panels: dark, snow, sunset) instead of the gamma one. feat(videoout): composite the overlay bus over the main output #10507 does not composite the overlay bus over an HDR10 presentation (it says so in its description and logs the overlay plane is not composited over an HDR10 presentation). DEATHLOOP draws its text and prompts there, so none of them were visible. With Cross only, it stayed on that screen for more than 5 minutes. The SDR gamma screen asks for Circle (○ SAVE AND CONTINUE), so the HDR screen may be waiting for the same button.
• Audio: works.
• Not checked: play beyond the first minutes on the beach.

Current blocker

• main stops while the key art splash is still on screen, on AGC graphics: CB metadata pass over part of a color target (State.cpp:2260): at about 50 s with a warm shader cache, 140 s on the first run. fix(agc): eliminate a fast clear over part of a CMASK color target #10414 lets the title past it. fix(agc): require full coverage only for metadata pass targets with CMASK or DCC keys #9694 alone does not: three runs with it stopped the same way as three on main, since the target has CMASK, as fix(agc): eliminate a fast clear over part of a CMASK color target #10414's description says.
• Text, menus and prompts need feat(videoout): composite the overlay bus over the main output #10507 (overlay bus), merged into main on 2026-10-26 as afc780e with the head tested here (f34c280). On an HDR display they also need APS5_FORCE_SDR=1 while the overlay isn't composited over HDR10.
• Open pull requests the title needs: fix(agc): eliminate a fast clear over part of a CMASK color target #10414.

How it was run

relinker --windows --registry --module-dir sys_module src/eboot.elf app.exe

The dump's eboot.bin and its 7 modules (sce_module: libc, libSceFace, libSceFaceTracker, libSceJobManager, libSceNpCppWebApi, libScePfs; sys_module: Party) are SELF containers, unwrapped with tools/unwrap_self.py first; fakelib is left out. On main, tools/import_audit.py (with --modules for both folders) reports 0 absent imports and 4 throwing-stub imports: sceVoiceWriteToIPort and sceAgcUpdateInterpolantMapping_0100, each imported twice. Environment: APS5_FORCE_SDR=1 on an HDR display; the runs also used APS5_NO_PAD_INPUT=1 with APS5_PAD_INPUT_FILE written by tools/pad_input.py, plus APS5_PROFILE_DRAW=1 APS5_LOG_TIMESTAMPS=1 APS5_TRACE_SUMMARY_EVERY=10 and APS5_DUMP_FRAMES for the pictures below.

Dumped frames of one run (main + #10414 + #10507, SDR), in order:

Conversion, the runs, log reading and this report were done with AI assistance (Claude Code); the screens above come from frames AnyPS5 dumped during the runs. The audio result was heard by me.