Astro's Playroom (PPSA01325)

[fma-rc]

Existing thread

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

Title

Astro's Playroom (PPSA01325)

Content version tested

01.905.000

AnyPS5 commit

main 8c4a295 with #2490, #2491 and PR 4162 merged on top (PR 4162 has been merged into main since). Last updated 2026-10-12.

System tested

Arch Linux, GCC 16.2.1, RTX 5080 (NVIDIA 615.78.08), 31 GiB RAM

Furthest point

In game

What is shown

The first level, cpuplaza_mini_04. Astro answers Cross (jump) and Square (punch). The level runs at about 5 FPS on the build above.

On the way there: the title screen (15 to 28 FPS), the main menu, the fly-through past older PlayStation hardware and the DualSense introduction (15 to 17 FPS). Title screen and menu are drawn completely since #3878, and the text is in the right language since #3933.

Not tested: audio and stick input. One screen after the main menu shows only a "close" prompt and no message; I have not looked at it again.

Current blocker

Frame rate in the level.

To reach the level the title still needs #425 or its successors #2490 and #2491. Without them it stopped on sampling the depth plane of depth surface 0x529430000 (2048x2048, vk format 124) as a 2048x2048 texture of guest format 7 (measured on main 7b5214d).

Stops from earlier in this thread that are fixed on main: the missing libSceJobManager (#3170), the heap error report from lazy binding on a fiber stack (#3214), the 2D image_sample on a 1D texture (#3222), the writing attribute in sceFontWritingLineWritesOrder (#3260), sceAudioInHqOpen (#3669) and the GNM subvector loop instructions (#3750).

How it was run

Substitute fonts in anyps5-fonts/ as docs/user/USAGE.md describes; without them the title asserts Out of Global Heap Memory. Environment: APS5_HOST_IMPORT_MIB=16384.

Buttons were sent through APS5_PAD_INPUT_FILE with tools/pad_input.py: Cross on the title screen, then "start a new adventure" (down, Cross, OK on the warning), then Cross through the introduction. From a new game the level starts after about 500 s.

[fma-rc]

What the title passes as the writing attribute, read from its only call site (app.elf 0x187d311 to 0x187d348): the low byte is the constant 0xFA, bits 8 to 14 come from a per-run flag word shifted left by 7 (the title fills that word from sceFontCharacterGetBidiLevel), and bits 30 and 31 are that word's bits 8 and 9, which the title sets from sceFontCharacterLooksWhiteSpace on the first and last character of the run. So 250 is level 0 with neither whitespace bit. The third argument can be NULL at that call.

After the write the title calls sceFontWritingLineGetOrderingSpace and uses the result to decide whether the run fits the line, so accepting the attribute without knowing what the console does with it would turn wrapping off silently. No module in the dumps I have exports the function and the public sources I found only log or stub it, so this needs the system libSceFont or its header.

[fma-rc]

Update, 2026-10-11: the title now gets through its menus. With #3669, #425, #3750 and #3878 merged onto main a11c8f7 (Linux, RTX 5080) it shows the title screen and main menu correctly, takes "start a new adventure", plays the fly-through past older PlayStation hardware and reaches the DualSense introduction pages. Nothing past those pages has been seen yet, so no gameplay. Frame rate by the window title is 15 to 30 FPS on the menus and introduction and 6 to 8 FPS in the fly-through.

Two things were in the way after the audio-in and depth-view stops. A wave64 mesh shader uses s_subvector_loop_begin / s_subvector_loop_end, which the recompiler refused (#3750). And the full-screen passes are 4-vertex strips over 3-record vertex buffers, so the fourth vertex is fetched as (0,0,0,0); NVIDIA draws that vertex at the screen centre, which put a wedge over the upper right of every scene (#3878).

Buttons were sent with APS5_PAD_INPUT_FILE and tools/pad_input.py. Still open: the text is Japanese with ANYPS5_LANGUAGE=1 too, and one screen after the main menu shows only a "close" prompt with no message. The first post has the full list of stops.

[fma-rc]

Update: the title is in its first level and Astro responds to the pad. With #3669, #425, #3750, #3878 and #3933 on main a11c8f7, plus one commit of mine on top of #425 (linked in my comment there), CONTINUE loads cpuplaza_mini_04: the floor hatch opens, a DualSense rises, Astro climbs out, and at about 210 s the camera settles behind Astro. A Cross press makes Astro jump and a Square press makes it punch (captures taken one second before and after each press). Astro also runs where the left stick points (forward, right, toward the camera), which gets it through the stick tutorial; scripting that needed stick fields in the test-run pad file (#4032).

It runs at 4 to 11 FPS there. Without the extra commit the level runs at about 1 FPS, because #425 redoes its layered depth copy on every draw; the measurements are on that PR. What limits it now is the driver worker thread, at 99% of one core over about 1700 draws per frame.

The Japanese text was the title asking sceSystemServiceParamGetInt for parameter 400, which returned 0 (#3933); the menus and the new-game warning are English with it. No error line in a 245 s run.

[fma-rc]

Working on optimization now, ~4.8fps to ~11 fps so far.