Tekken 5 Dark Resurrection [NPUB30009] - Game Freezes During FMV Playback

[Fukuzatsu]

Quick summary

Tekken 5 Dark Resurrection [NPUB30009] plays an FMV intro at the title screen and uses FMV for its title logo as well; after allowing the game to sit at the title screen for a short time (in this test, around 10 minutes), the FMV will freeze, preventing the user from proceeding. This also occurs during gameplay, as any time the user returns to the title screen, an FMV will also be loaded for the title logo.

Details

Similar but different to #12497, this issue occurs while FMV is playing (rather than failing to load it at the beginning), either on the title screen or the intro video itself. There also isn't much consistency about when or how it freezes, after many attempts it appears to be totally random as to when it freezes, though it is always during/related to FMV playback.

The log file for this game gets very long even on default settings, but the main message I notice a lot of when it crashes is this one:
·E 0:07:47.037146 {PPU[0x1000000] main_thread} SYS: Stat: 'cellVdecGetPicItem' failed with 0x80610104 frame=0, sequence_id=1, seq_id=0 [x553]

It's not immediately apparent to me how to fix it, but as with the other FMV related issue I've reported, I tried the following:
Toggling libvdec.sprx for LLE/HLE
Toggling libavcdec.sprx for LLE/HLE
Toggling Accurate RSX Preservation, setting RSX FIFO Accuracy to 'Atomic', Driver wakeup delay to 40

I have also verified my copy of the game using PSN PKG Checker.

For the log attached below, I have run the game using version 0.0.23.14014 with default settings, aside from having the Resolution Scale set to 200%.

Attach a log file

RPCS3.log.gz

Attach capture files for visual issues

No response

System configuration

GPU: NVIDIA GeForce RTX 3050 8GB DDR6
11th Gen Intel(R) Core(TM) i5-11400F @ 2.60GHz | 12 Threads | 15.84 GiB RAM | TSC: 2.592GHz | AVX-512+ | FMA3
Operating system: Windows, Major: 10, Minor: 0, Build: 22000, Service Pack: none, Compatibility mode: 0

Other details

No response

[Fukuzatsu]

I've been doing some more thorough testing on this, observing behaviour during different crash instances to see if there were any patterns.

For the purposes of describing this, each intro "cycle" is comprised of what I'll say are 5 component parts.

1. Bandai Namco Logo and Intro FMV
2. Title Screen
3. Demo Fight Versus Screen (the "Get ready for the next battle" screen)
4. Demo Fight
5. Rankings Screen

At the end it will of course loop back around to the logo/FMV at the beginning.

In testing this, I've noticed some unique entries in the log, which I'll share here:
On a regular "cycle", the in-app log shows the following:
At FMV start - U sys_prx TODO: _sys_prx_get_module_id_by_name(name=“cellLibprof”, flags=0, pOpt=*0x0)x2
At Demo Versus Screen - E SYS: 'cellVdecGetPicItem' failed with 0x80610104 frame=0, sequence_id=1, seq_id=0 [1]
At Demo Versus Screen - E SYS: 'cellVdecGetPicItem' failed with 0x80610104 frame=0, sequence_id=1, seq_id=0 [2]
Demo Fight Begins - U sys_prx TODO: _sys_prx_get_module_id_by_name(name=“cellLibprof”, flags=0, pOpt=*0x0)
Ranking Screen Displays - E SYS: 'sys_memory_free' failed with 0x80010002 : CELL_EINVAL, 00 [1]

On a crash cycle, the in-app log instead shows:
At FMV start - U sys_prx TODO: _sys_prx_get_module_id_by_name(name=“cellLibprof”, flags=0, pOpt=*0x0)x2
E SYS: 'cellVdecGetPicItem' failed with 0x80610104 frame=0, sequence_id=1, seq_id=0 [1]
E SYS: 'cellVdecGetPicItem' failed with 0x80610104 frame=0, sequence_id=1, seq_id=0 [2]
E SYS: 'cellVdecGetPicItem' failed with 0x80610104 frame=0, sequence_id=1, seq_id=0 [3]
Crashes during playback of intro FMV

In essence, on a crash cycle it reports 3 errors with cellVdecGetPicItem during the FMV intro despite only reporting them on the versus splash screen on a regular cycle (and in the regular cycle there are 2 errors rather than 3).

Attached is a truncated log file which includes the first cycle (which runs correctly) and the last cycle (where the intro crashes). Feel free to let me know if there's any better data I can provide.
Tekken5DR-Crash.gz

Incidentally, I noted down the parameters of the demo fights that occurred prior to the crash to see if maybe a particular arrangement of characters/stages caused it, but it didn't seem to have any connection. For example, even though the intro crashed following a demo fight of Julia v. Baek on Urban Jungle; setting up the same match in say, VS mode does not cause the crash when the cycle is left to run again.

[Fukuzatsu]

As mentioned in #12497, I've been continuing to test for this issue across various updates, and as of my last test session with version 0.0.23-14088, the issue is still persisting in T5DR, though the similar FMV-related problem in Tag Tournament 2 has since been resolved.

[ChwaraeTeg]

I can consistently repeat this issue on 0.0.25-14407 with all settings left on default - it will freeze usually within 10 minutes, and the same 3 "cellVdecGetPicItem" errors appear in the log at the point of freezing.

However, I cannot reproduce the issue with libvdec.sprx toggled to LLE, even when leaving it running for over an hour at a time.

[Fukuzatsu]

Coming back to this issue, seems like the problem was improved somewhat for a while (or at least the game had to run much longer to hit a freeze state, but as of version 0.0.27-14876 I'm able to once again reproduce the freeze in less than 10 minutes, even with libvdec.sprx toggled to LLE, though the prior error message regarding "cellVdecGetPicItem" no longer appears. While the game does crash with solely libvdec.sprx set to LLE, I did also test it in conjunction with libatrac3plus.sprx set to LLE, which did not change the behaviour.

The in-application log reads as follows:
E sys_fs: 'sys_fs_stat' failed with 0x80010006 : CELL_ENOENT, “/dev_hdd0/game/NPUB30009/USRDIR/caption_/T55DATA1.BIN” [1]
U sys_prx TODO: _sys_prx_get_module_id_by_name(name=“cellLibprof”, flags=0, pOpt=*0x0) x3
E SYS: '_sys_lwmutex_lock' failed with 0x80010005 : CELL_ESRCH [1]
E SYS: '_sys_lwmutex_lock' failed with 0x80010005 : CELL_ESRCH [2]
E SYS: 'sys_lwmutex_lock' failed with 0x80010005 : CELL_ESRCH [3]
E SYS: 'sys_ppu_thread_set_priority' failed with 0x80010005 : CELL_ESRCH [1]
E SYS: 'cellAtracIsSecondBufferNeeded' failed with 0x00000000 : 00 [1]
E SYS: 'sys_memory_free' failed with 0x80010002 : CELL_EINVAL, 00 [1]
E VM: Access violation reading location 0x36200000 (unmapped memory)
E sys_fs: 'sys_fs_stat' failed with 0x80010006 : CELL_ENOENT, “/dev_hdd0/game/NPUB30009/USRDIR/caption/T55DATA1.BIN” [2]
U sys_prx TODO: _sys_prx_get_module_id_by_name(name=“cellLibprof”, flags=0, pOpt=*0x0) x3
E SYS: 'sys_memory_allocate' failed with 0x80010004 : CELL_ENOMEM [1]
E SYS: Stat: '_sys_lwmutex_lock' failed with 0x80010005 : CELL_ESRCH [x24094]
E SYS: Stat: '_sys_lwmutex_lock' failed with 0x80010005 : CELL_ESRCH [x23660]

The full log from this test is attached RPCS3.log.gz

[FlexBy420]

I left the game running for a bit and I cannot reproduce this bug, so it may have got fixed at some point.
RPCS3.log.gz