Repository navigation
Hasbro Family Fun Pack (CUSA03312) #7198
savvaniss
started this conversation in
Game bring-up
Replies: 2 comments
|
Update, 19 Oct, for base 01.00 and patch 01.02 (my own dump). Conversion: complete without local experiments.
Startup: 20 unresolved imports.
Furthest point: still stops before the first frame, at the startup import check, and no window opens. AI-assisted: yes (Claude Code) |
0 replies
|
Update, 19 Oct: the title now plays its intro video and stops while loading its start menu (Windows 11, Intel UHD 770, my own dump). The post above is updated.
Current stop: the first vertex program of the menu reads a buffer whose V# is AI-assisted: yes (Claude Code) |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
Existing thread
Title
Hasbro Family Fun Pack (CUSA03312)
Content version tested
01.00 (base). The 01.02 patch converts the same way.
AnyPS5 commit
#6977 head
93eeee014with #7660 (bytes between unwind functions), #7780 (PS4 save-data imports) and #7869 (point and line polygon modes). On top: the local placeholders and stand-ins listed under Current blocker, which are experiments, not proposals. Inlibs/, a copy of the builtlibSceGnmDriver.prxis also namedlibSceGnmDriver_padebug.prx.System tested
Windows 11 Pro 26300, WinLibs GCC 15.2.0 r7. Runs on the Intel UHD Graphics 770 (driver 32.0.101.7088) of the same PC as the AMD Radeon RX 9070 XT (driver 32.0.31041.1004).
Furthest point
Logos or loading screen
What is shown
A window opens and the Hasbro intro video (
hasbro_emea.mp4) plays; the driver gets about 945 frame submissions in 20 s. The title then reloads its start-menu screens, fonts and music, and stops at the first vertex program of the menu.Current blocker
24abe8ae 02500000 00ffc1ff 90c00fac: "buffer descriptor uses an unsupported type", withADD_TID_ENABLEset and word 3 bits 31:30 = 2. An earlier run stopped at another vertex program'sADD_TIDbuffer ("buffer ADD_TID is only valid for compute shaders"). Both look like ring or scratch V#s of a PS4 vertex program; how the console encodes them is not established here.sceGnmDrawInitDefaultHardwareState200: emits feat: run PS4 games on the AGC driver — Limbo (CUSA01664) plays at 60 fps #6977's 1.75 sequence.sceGnmUpdatePsShader,sceGnmUpdateVsShader,sceGnmUpdateGsShader,sceGnmUpdateHsShader: emit thesceGnmSet*Shaderpackets.RELEASE_MEMwithDATA_SEL = 4(a 64-bit counter) is rejected as GDS data; the stand-in writes the 100 MHz counterDATA_SEL = 3writes.CB_COLOR_INFO.FAST_CLEAR) are refused in feat: run PS4 games on the AGC driver — Limbo (CUSA01664) plays at 60 fps #6977's path; the stand-in draws them without CMASK.sceGnmDrawInitDefaultHardwareState200,sceGnmMapComputeQueueWithPriority,sceGnmUpdateGsShader,sceGnmUpdateHsShader,sceGnmUpdatePsShader,sceGnmUpdateVsShader,sceNpCheckPlus,sceNpGetAccountCountry,sceNpGetAccountDateOfBirth,sceNpGetAccountLanguage,sceNpGetParentalControlInfo,sceNpMatching2AbortContextStart,sceNpMatching2SignalingGetConnectionStatus,sceNpProfileDialogTerminate,sceNpProfileDialogUpdateStatus,sceNpTrophyAbortHandle,sceSslGetNotAfter,sceSslGetNotBefore,sceInvitationDialogClose,sceKernelPrintBacktraceWithModuleInfo.Names are from
tools/nid_names.py.How it was run
My own dump of the title. The conversion command was
relinker --windows app0/eboot.bin app.exewith the bundledsce_module/beside the eboot.libs/comes from the same build, pluslibgcc_s_seh-1.dll,libstdc++-6.dllandlibwinpthread-1.dll.app.exewas started from its folder, laid out as in USAGE.md, withANYPS5_GPU=Intel.AI-assisted: yes (Claude Code)
All reactions