Jak Collection/inFamous 1 Broken Refracting Particles

[JimScript]

Throughout the Jak collection, the particles that refract the scene are broken, even though other refracting effects like water and the scattershot work fine. Some parts of the particles mesh with the scene which leads me to believe this is a uv problem.

Jak and Daxter:

BCUS98281_20211112114854_capture.zip

Jak II:

BCUS98281_20211112115137_capture.zip

System:

• SYS: RPCS3 v0.0.19-13006-8b3ed7b5 Alpha | HEAD
• SYS: AMD Ryzen 5 3600 6-Core Processor | 12 Threads | 15.95 GiB RAM | TSC: 3.600GHz | AVX+ | FMA3
• SYS: Operating system: Windows, Major: 10, Minor: 0, Build: 19043, Service Pack: none, Compatibility mode: 0
• RSX: Found vulkan-compatible GPU: 'NVIDIA GeForce RTX 2070 SUPER' running on driver 471.68.0.0
• SYS: Firmware version: 4.85

[JimScript]

I tried to get a photo of a real ps3, but when the effect is working correctly it is hard to see.

[kd-11]

Maybe you can circle or highlight the difference? I have no idea what this is about otherwise (never played jak)

P.S By highlight I mean, with like a red circle and arrows and such.

[MSuih]

The Jak II scene is supposed to be an air vent blasting hot air out. In his last image you can see it slightly distort the image. In the first post the same distortion effect looks like some ice cubes were stacked on top of the vent or something. You're not supposed to see the edges of the effet so sharply.

[JimScript]

And because of that it is hard to put a red circle around it, it is basically the warped image behind jak above the vent. What it is suppose to look like is a smudge like effect, like the image is bulging out.

[kd-11]

Oh ok, so its just the infamous distortion shader glitch.
For deformation shaders the UVs are from 2 sources, fixed UV in screen space and a deformation vector usually gleaned from dFdx and dFdy + a perturbation ref texture. The functionality is implemented but doesn't work right. I haven't had time to look into the distortion issues, they're fairly low priority all things considered, but if more games are affected I can move it up the queue. Currently I've only heard of it in infamous and now Jak games.

[RinMaru]

(never played jak)

shame on you >.<

but yea i get this error too. ill be upgrading from RX 480 to RX 6700XT soon so i can test on that it that would help with figuring it out

[Asinin3]

FYI, One other game I saw someone have this issue in was Persona 5, #6993 but I couldn't reproduce it and they didn't update the issue so it was closed.

[JimScript]

Is the Persona 5 issue the same, because in both inFamous and Jak the image in the shader is offset to the left and up a bit. Here is a picture of it in action in inFamous:

You can also see it in the Jak II image and a little bit in the Jak I image above, but it is hard to see in the Persona 5 examples in that thread.

[JimScript]

Do you have any examples of fire distortion being broken? It is a little hard to see and the blocking kinda reminds me of anti-aliasing breaking.

[JimScript]

Wow that is a lot more broken than your first post implied, but looking at it closely it looks like MLAA or something is broken. If you look around 4:45 you see the squares form around the flagpoles and the edge of the sculpture. If this error is in effect too, I can hardly tell.

[Darkhost1999]

If you ignore obvious SPU issues and watch the smoke I think BF3 has same issue as infamous with same effect as infamous vents.

bf3.mp4

[JimScript]

Yeah it kinda looks like MLAA or MSAA is drawing over the games particle systems. Jak and inFamous are both ingame shader problems, this looks like a post-processing problem.

[Darkhost1999]

Yeah it kinda looks like MLAA or MSAA is drawing over the games particle systems. Jak and inFamous are both ingame shader problems, this looks like a post-processing problem.

It'd make more sense if the issue went away disabling this and RPCS3 AA.
This being disabled in all videos and screenshots above. RPCS3 being on.

[JimScript]

Well now that is strange, maybe it is the same issue, maybe it's just that Battlefield 3 uses the function in a different way than Jak or inFamous. But it is hard to see the similarities, as both Jak and inFamous have it as a glassy image that is offset to left. What makes you think that this is the same issue?

[Darkhost1999]

What makes you think that this is the same issue?

Lack of better knowledge

[JimScript]

Well I guess we won't know for sure if this or the Persona 5 issue is related until it is fixed.

[Codex-]

Came to open the same issue, happy to try or test things.

So far I've tried many combinations and permutations of settings to no avail.

• SYS: RPCS3 v0.0.22-13693-03df3c52 Alpha | master
• SYS: AMD Ryzen 9 5900X 12-Core Processor | 24 Threads | 31.91 GiB RAM | TSC: 3.700GHz | AVX+ | FMA3
• SYS: Operating system: Windows, Major: 10, Minor: 0, Build: 22621, Service Pack: none, Compatibility mode: 0
• SYS: Current Time: 2022-06- 1T22:55:11
• RSX: Found vulkan-compatible GPU: 'NVIDIA GeForce RTX 3070 Ti' running on driver 512.95.0.0
• SYS: Firmware version: 4.88

Current (as of v0.0.22-13693-03df3c52, 2022-06-01):

Expected:

Can see a video of how the particles are meant to look here at 3:48

[Codex-]

It sounds like you already have a strong suspicion as to what the issue is, kd-11, I was having a read through the jak and daxter decompilation project and thought it might help provide some more concrete insight to the implementation details for these. Albeit for the ps2 version but who knows: https://github.com/open-goal/jak-project/tree/master/goal_src/engine

Here's to hoping it helps a little 😅

[JimScript]

Well I imagine that Mass*Media(the company that ported the collection) might have changed a few things behind the scenes for it to work on the PS3. But then again, the Jak project could have too. The PS2 and PS3 are two very different consoles, so the work might not be transferable.

[JimScript]

Thanks for bumping the issue, maybe I should change the name to include inFamous 1 so these issues stop popping up.

[JimScript]

For some reason, inFamous 1 fades into the broken refraction and at a certain distance some of it is fine, with it getting progressively worse when it gets closer to the camera:

inFamousHazeFade.mp4

[JimScript]

Okay, so I tried dissecting the scene and I found this normal that dictates the refraction:

I then dissected it further to see what the problem was, and found that the working parts are all increments from the 128 base value while the broken parts are all solidly 255.
Here is an example I made for the Red Channel, exaggerated for easy viewing:

The red parts go from 129-131 red while the blue parts are all 255 red, interesting that they are on opposite sides. Which makes me think there might be an integer underflow problem when it tries to become 127 or lower, so it loops back to 255.
Here is the example for Green Channel too, also exaggerated:

Don't know how this affects Jak though.
Here is the RSX capture of the scene:
BCUS98119_20230128235239_capture.zip

[JimScript]

Well, I decided to look back into how the Jak issue relates to the inFamous issue:

With the normal that the refraction uses of course:

And it's similar in that below its midway point it whites out, but this midway point is 127 instead. It also looks like it's more based off of polygons instead of textures, and when a any pixel in a triangle goes below 127, then that whole triangle for that channel whites out. That's probably why some of the polygons flicker ingame, they're going in and out of the broken range. Here are the channels re-rendered in between 127-140 for better viewing:
Red Example:

Green Example:

[KieronJ]

For reference, this shader works by rendering the pentagons with additive blending. The vertex "color" seen above comes from vert_uv - vert_xy. A second pass then uses the RG channels of the pentagon buffer as a screen space displacement.

[AniLeo]

I'm locking this due to all the off topic spam, if you have new information to add ping me on Discord to have it reopened, otherwise kd will reopen it when he works on it

[Hasster1]

Some more info from the InFamous' side of things:
While trying my luck at finding some interesting stuff with Cheat Engine, i've stumbled upon one opcode, that re-activated the bug with the glitched borders of the screen, that was fixed in #4973
I also noticed, that with that bug present on the screen, the broken reflections did not show up (The fire usually produced the reflection particles, but they were hidden)
I've traced that opcode in ghidra, and found the value, that controlled the amount of shifting for both particles and the glitched border. Default amount is set to 0.1.

I hope this info helps a bit with fixing this bug.

[Hasster1]

can you produce a test build with this workaround? would be much appreciated.

There's already a patch for InFamous that sets the shifting to 0, disabling the effect.

[DarkPhobia0]

Just to understand, this doesn’t fix the effect, just disables it, right? Thanks! El El mié, 11 sept 2024 a la(s) 5:56 a. m., Hasster < ***@***.***> escribió:

…

can you produce a test build with this workaround? would be much appreciated. There's already a patch for InFamous that sets the shifting to 0, disabling the effect. — Reply to this email directly, view it on GitHub <#11149 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A566LCAIB3D2Z6SUHVJQNB3ZWAAVHAVCNFSM5H5TNNQKU5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TEMZUGMYDKMJUGYYA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[Hasster1]

Just to understand, this doesn’t fix the effect, just disables it, right?

Yes, that's right.

[Hasster1]

I've never used the patch updater before... im just lost sorry.

https://discord.gg/RPCS3

[AniLeo]

Locking the issue again since the spam is back, which is a shame.