video_out: SubmitEopFlip adding mutex and lockguard also fixing reject of full flips by adding a wait

[diegolix29]

This stops the flip bug we have on video out on my end while testing dark souls 3 fire fades
updating buffer labels, and adding buffer id for logging on error
let me know if its ok.

Also fixed a discrepancy on driver.cpp

[GHU7924]

[CUSA03745] The Last Guardian

Errors appeared:

[Render.Vulkan] vk_pipeline_cache.cpp:493 CompileModule: Compiling vs shader 0xc8e8ce26
[Render.Vulkan] vk_pipeline_cache.cpp:493 CompileModule: Compiling fs shader 0xf694dd0e
[Render.Vulkan] vk_pipeline_cache.cpp:493 CompileModule: Compiling vs shader 0xd41f6d2b
[Tty] logger.cpp:62 log_flush: [stdout] Garlic AllocatedMemorySize: frame 00000003, 671.8M, 23.9% used
[Tty] logger.cpp:62 log_flush: [stdout] Peak volatile Garlic: 60 KBytes
[Tty] logger.cpp:62 log_flush: [stdout] Garlic AllocatedMemorySize: frame 00000004, 710.3M, 25.2% used
[Tty] logger.cpp:62 log_flush: [stdout] Peak DCB: 26.53 KBytes
[Tty] logger.cpp:62 log_flush: [stdout] Peak CCB: 3.62 KBytes
[Lib.VideoOut] <Error> driver.cpp:224 SubmitFlip: Flip queue is full
[Lib.VideoOut] <Error> video_out.cpp:354 operator(): EOP flip submission failed for buffer 2
[Lib.VideoOut] <Error> driver.cpp:224 SubmitFlip: Flip queue is full
[Lib.VideoOut] <Error> video_out.cpp:354 operator(): EOP flip submission failed for buffer 0
[Lib.VideoOut] <Error> driver.cpp:224 SubmitFlip: Flip queue is full
[Lib.VideoOut] <Error> video_out.cpp:354 operator(): EOP flip submission failed for buffer 1
[Tty] logger.cpp:62 log_flush: [stdout] Garlic AllocatedMemorySize: frame 00000011, 726.3M, 25.8% used
[Lib.VideoOut] <Error> driver.cpp:224 SubmitFlip: Flip queue is full
[Lib.VideoOut] <Error> video_out.cpp:354 operator(): EOP flip submission failed for buffer 2
[Lib.VideoOut] <Error> driver.cpp:224 SubmitFlip: Flip queue is full
[Lib.VideoOut] <Error> video_out.cpp:354 operator(): EOP flip submission failed for buffer 0
[Lib.VideoOut] <Error> driver.cpp:224 SubmitFlip: Flip queue is full
[Lib.VideoOut] <Error> video_out.cpp:354 operator(): EOP flip submission failed for buffer 1
[Tty] logger.cpp:62 log_flush: [stdout] exec at score engine
[Lib.VideoOut] <Error> driver.cpp:224 SubmitFlip: Flip queue is full
[Lib.VideoOut] <Error> video_out.cpp:354 operator(): EOP flip submission failed for buffer 2
[Lib.VideoOut] <Error> driver.cpp:224 SubmitFlip: Flip queue is full
[Lib.VideoOut] <Error> video_out.cpp:354 operator(): EOP flip submission failed for buffer 0
[Lib.VideoOut] <Error> driver.cpp:224 SubmitFlip: Flip queue is full
[Lib.VideoOut] <Error> video_out.cpp:354 operator(): EOP flip submission failed for buffer 1

These lines didn't exist before. My game stopped working properly. It doesn't show any critical errors.

[diegolix29]

does the game works on main but this pr break it?

[GHU7924]

does the game works on main but this pr break it?

I have YES, but let someone else write who has this game.

[kalaposfos13]

P.T. crashes, TLG hangs, and Journey and Minecraft spam the same lines as in the comment above, but otherwise work

CUSA03745.log
CUSA00265.log
CUSA02172.log
CUSA01127.log

[diegolix29]

@kalaposfos13 @GHU7924 can u try TLG with the new commit?

[GHU7924]

@diegolix29 I have no changes, also hangs. I also think that the log will be useless here and it will be better to check through Visual Studio. But I won't be able to check this.

[kalaposfos13]

You just changed two s32-s to int-s, this will have exactly zero effect and if I'm not mistaken will even compile to the exact same executable as s32 is basically just a typedef for an int

[diegolix29]

You just changed two s32-s to int-s, this will have exactly zero effect and if I'm not mistaken will even compile to the exact same executable as s32 is basically just a typedef for an int

yes that was the discrepancy i find on driver.cpp, i revert it as s32 as it should be, on both header and constructor, the logging i place it to fetch info from testing as i only have darksouls 3 that triggers that crash. Rn im doing some changes to both logging and the registeronce hoping it will help, after local testing ds3 i will push a commit.

[diegolix29]

can u try again with the new commit ds3 still is safe with this change.

[GHU7924]

CUSA03745.log

P.S. I'm not sure if there are any other problems associated with this PR, but at least now it is possible to get into the game.

[diegolix29]

Just need to see if this dosent regress any other game, but it should be ready after no regressions are confirmed

[diegolix29]

Forgot to add this PR fixes out of order flip IRQ

[kalaposfos13]

Now none of those four games spam the error anymore, and P.T. went back to only occasionally crashing with EOP flip submission failed, so whilst this PR did not improve that, it now doesn't regress it either.
CUSA01127.log

[diegolix29]

Now none of those four games spam the error anymore, and P.T. went back to only occasionally crashing with EOP flip submission failed, so whilst this PR did not improve that, it now doesn't regress it either. CUSA01127.log

It should be fixed now, i changed the log error too cause now we wait for space for flip instead of rejecting the flip when queue is full, let me know if u still randomly hit that crash

[diegolix29]

So if PT and TLG are ok and were the only regressed games this should be ok now 💪

[GHU7924]

[CUSA30992] Teenage Mutant Ninja Turtles: Shredder's Revenge version 1.00 works for me in Main.
With this PR she doesn't even get to the menu.

At the end of the log there is a spam error:
[Core] <Error> stubs.cpp:42 CommonStub: Stub: select (nid: T8fER+tIGgk) called, returning zero to 0x814159deb

In general, here are the logs, but I don’t know if they will help.

Main: CUSA30992.log (I skipped the opening cutscenes and ended up in the game's main menu)

PR: CUSA30992.log (Always black screen)

UPD: PR Old Build ddefcaf the game worked.

[diegolix29]

@GHU7924 could u try it again? i added something that should help if what i think its happening is correct.

[GHU7924]

Now it works.

[diegolix29]

Niceee thanks hoping i have no more games regressing with this changes. gonna do a rebase

[GHU7924]

[CUSA03745] The Last Guardian

2025-05-29.22-34-04.mp4

CUSA03745.log

---

[CUSA00076] The Order: 1886 (this game doesn't work yet but it has a line)

[Lib.VideoOut] <Warning> video_out.cpp:159 sceVideoOutSubmitFlip: flipmode = 3

I don't know if this information is necessary for this PR, but just in case I'm attaching the game log for study, if the information is useful, then good, if not, then I apologize.

CUSA00076.log

[diegolix29]

this is not related to the pr but i can take a look at it no problem but will be for the weekend

[GHU7924]

There are no changes. But I think @kalaposfos13 should also recheck his games, maybe there will be some differences.

this is not related to the pr but i can take a look at it no problem but will be for the weekend

There is no need for this, I only reported it because I thought it might be related, but I was wrong.
You still have other open pull requests, so you shouldn't waste your time on this.