[zackboy]

@SpaceFolding
PGXP ins't finished yet don't forget this, for the moment the compatibility is variable (in some games memory+cpu is needed but it makes "beautiful" bugs)
Vertex Caching can correct and/or add bugs
Perspective correction has problems only with Spyro games
I think in Time Crisis it's not jitter problem but "bad" animations (for PS1 it's beautiful to see animation on the face and not just texture swap)
This game just relly need z-buffer, it's too visible for me..

 

[zackboy]

Spyro Games:
GTE Accuracy/Vertex caching: better but not perfect and bugs in some 2d elements.
PGXP memory only: better than GTE Accuracy/Vertex caching but some elements are in low precision.
PGXP memory only + Vertex caching: as PGXP memory only, some low precision values are replaced by Vertex caching values, the 2d bugs are back.
PGXP memory&CPU: better but strangly sometimes I have a texture corruption when the portails are in the camera in Spyro 1, some elements are always not in high precision.
PGXP memory&CPU + Vertex caching: as PGXP memory&CPU (rarely some low precision elements have Vertex caching values) + bugs in some 2d elements.

Perspective correction: Works good except on close polygons.

xBRZ:
Spyro 1/2 Good (black borders sometimes).
Spyro 3 Good but black borders sometimes + Bad ui (only the borders are xBRZ-ified).

For your problem I don't know.. Sorry.

 

[zackboy]

Have you enabled xBRZ+texture cache from the tweak? The 4Gb patch will correct this problem..

 

[zackboy]

@TheDimensioner I have already mentionned this texture corruption.. It appears when one of the portals is on the screen with Mem + CPU logic (in certain angle)...

With a recent nvidia card use shadowplay, it's better than Fraps (less FPS impact and better compatibility and quality)

 

[zackboy]

Oh! Hi @tapec !
Hope your health is better! ^^

 

[zackboy]

@shinra358
http://ngemu.com/threads/peteopengl2tweak-tweaker-for-peteopengl2-plugin-w-gte-accuracy-hack.160319/

 

[zackboy]

Yes @tapeq has alwready explain somewhere, but you can ask him (and why not ask him to add an explaination in the first post)
My post is mostly because there is too many off topic between PGXP and PeteOpenGL2Tweak.. Sorry

 

[zackboy]

Hi everybody, sorry for the double post but I have some "requests":
Of course if you want and have time.

@iCatButler About PGXP:
Do you think it's possible to correct the bug with 2D elements in Mem + CPU Logic?

Other things:
- what do you think about adding an "Demi-Widescreen Hack";
the actual Widescreen Hack change the game in 4:3/12:9 into 16:9
the Demi-Widescreen Hack change the game in 4:3/12:9 into 14:9, so less 2D stretched/misaligned, less clipping (like Crash and Tomb Raider 2)
(I know thats not real 4:3 and sometimes it depend to NTSC/PAL, but you see what I mean)

- @tapeq About PeteOpenGL2Tweak Is it possible to correct the Texture misalignment when using the texture filtering? (GSdx is better but bugged, Edgbla's GPUBladeSoft is the best)

- Why not adding CPU Overclock 33.8688 MHz to 37.5 MHz (Like PS2 in certain case), maybe the compatibility will be better.. But the result like nothing..

- if anybody has an idea for improve the games' Framerate without CPU overclock

PS: I know I spoken about the tweak, it is incoherent compared to what I was saying before but I think only one post is better, Sorry

 

[zackboy]

@iCatButler I already use the mem heck but I don't see a real difference in the games I tested, the problem with the cpu overclock is the speed in certain games and/or the controls that not respond..
You maybe saw Tomb Raider (1-5) are pretty fast with, and it's sad in particular with TR3 who has a bad Framerate sometimes..
But if it's impossible, IMO a perfect emulator has perfet compatibility, doesn't need game specific hack or game specific config. So no problem.

Thank you for considering my requests!

@SueSide known issue, no fix yet..

 

[zackboy]

Is there somebody who know the difference between PCSX-R/ePSXe and Mednafen's CPU overclock?
I tested Tomb Raider 3 and with Mednafen the framerate looks better without any speed issue.. :O

 

[zackboy]

Thanks, so in Mednafen it's more a "power hack" than an overclock, and theoretically there is no side effect, right?
Too many emulators have their specific features.. It's good to stand out from others, but not handy for the users..

 

[zackboy]

For PGXP's purposes the 'w' component is the depth information generated by the GTE when it transforms vertices into screen space. It's used for perspective correct texturing as it can tell the GPU the relative depth of all the vertices in a polygon.

The values in that image are about the best you can expect. It's clear the further points are toward the red end of the spectrum and nearer ones blue, with a smooth gradient in between. This means that there should be no problems with texturing and the values could even be used for to create a proper z-buffer (possibly at some point in the future).

Most games produce 'w' components that are suitable for perspective correct texturing. All that this requires is that for any given polygon the further vertices have a greater depth than those nearer. The most notable exception is Gran Turismo 2, which stitches the environment together from patches of geometry that have different depth information. This results in some border triangles having depth values that stretch from the furthest limit to the nearest in the wrong direction resulting in incorrect texturing.

Spoiler

As you can see some triangles are shaded from red/yellow to blue in the wrong direction, resulting in oddly stretched textures.

More commonly the games will process elements at different scales but won't stitch them together like this. The result is that textures appear correct but any attempt to build a Z-buffer from the values would be incorrect.

Spoiler

As you can see, while the main level geometry appears correct, the sky and repeating or dynamic objects appear out of order in terms of depth.

Hey @iCatButler I have an idea for correcting the perspective correction bug, but I don't know if it's feasible..
Why not adding a "perspective decorrection" when the depth difference is too brutal (we can see Blue to Red and that the problem), so I think normal perspective in this type of case is better than buggy correction.

You don't want to waste time with z-buffer if I understand? I know some games will don't work but they are exceptions, the others use the OT In a traditional way.

Spoiler

 

[zackboy]

There is also Spyro, but the depth difference isn't as brutal

Spoiler

(we can also see no z-buffer possible for this game too by using OT, it's like there is multiple OT :/)

So z-buffer looks like waste of time except if an other way than using the OT is found

 

[zackboy]

Tomb Raider III has some small bugs with pgxp and CPU Logic, Some other 2d problems with vertex cache, and not completely all vertice corrected with memory only.. But nothing so bad.
You don't use the last pgxp version:
WIP_PGXP_build_16_08_31.zip

 

[zackboy]

Hi @iCatButler
Just now I wanted to test PCSX with PGXP in memory+CPU but by missclick I launched the Bios, and there is some false positives in SCI logo, there is few vertices so I think test the bios for your project will be easier than any other games

Overclock in x1.25 is just perfect (but the SPU have to be correctly configured):
No Gamepad break, No Speed issue (or anecdotal, plus generally games with speed depending to the CPU's clock aren't optimised for PAL, 60/50 > 1.2≈1.25 that's correct ) and less framerate drop.
I can't imagine this with mednafen's "CPU Overclock" instead of just "WipEout MemHack"
I know that's not your speciality but do you thik it can be implemented ?

 

[zackboy]

Here is a short video for watching PGXP Mem+CPU Logic "Funny" issues:

 

[zackboy]

After looking w-component of some games I saw Harry Potter, Metal Gear Solid, donald duck quack attack, Rayman 2 and Tomb Raider Games who will work correctly with instead of OT. (enable CPU Logic help sometimes)

I have an idea for helping some games, make a "Depth mode" list:
- Ordering Table (Original, more compatible less precise)
- w-component (Less compatible, buggy)
- Hybrid (Use w-component in Ordering Table)

For the Hybrid the OT stays the main "Z-buffer" but vertices with the same(similar not identical) depth value will be also sorted by w-component (like an OT in an OT/ except verry first and last drawed values in the OT for protect eventual UI/skybox/background), with that method only "z-fighting" will be corrected.
For exemple with your crash bandicoot 3 screens in your 1st post the levels should display correctly
Toy Story 2 and pandemonium are good exemples.
Maybe Gran Turismo 1 (the colors are limit limit)

I think Spyro will not work... Again I'm not sure, with OT values I only see Red and sometimes Yellow or cyan, but alway one color... :/
I'm not sure for Crash 3 warp room..
Gran Turismo 2 too
Or maybe there are just many "same" colors with small variants (there are not all identical when I use a magic wand in an image editor (and the repartition seems Logic), Spyro is the only one who all colors are identical)

Excuse me to talk again for z-buffer, I promise it's the last time


EDIT: I tested many games more or less popular in my search of obscure engines, I found some Games who have also multiples problems with the Ordering Table view:
Spyro 1/2/3, Rugrats: Search for Reptar, Stuart Little 2, Disney's Treasure Planet, Tom Clancy's Rainbow Six series

 

[zackboy]

Some games with strange Ordering Table Visualisation. (build_17_03_17)
Spyro1/2/3, Rugrats: search for reptar, Stuart Little 2, Disney's treasure planet > generally only one color diplayed (they are identical when using magic wand without tolerence in an image editor)
Tom Clancy's Rainbow Six seems incoherent > orange can't be in front of blue..

My PGXP phantasm (warning that's hot)

Spoiler

It should be feasible..

 

[zackboy]

It looks really smooth for a PlayStation game, I'm impressed.

Oh I'm sorry it's just a concept, I have a programmer diploma but I'm only very high level, I wanted to be more but for french university logic I haven't the skills for that, only because I was bad in economics...there is only computer science in my choice.. French logic: If you are good but verry bad in one subject(without any relation): no way, if you are just a little good: no problem
I will pass another diploma to be "BAC + 4", I'm currently only "Bac+2", but with my current skills; hardware programming, emulator programming and 3D programming aren't easy for me...
I'm already happy to understand a little the PS1 Documentation/logic ^^"

BUT I really think my propositions will work (for the hybrid depth there is many ways to implement it, correct the vertices with the new order, implement a real drawing with z value (a fake Z value because there is two values (OTZ/Wcomonent) for exemple it can be based on the index value if we don't want a propre code.. that's a hack it can be unpropre ))
It's not a perfect z-buffer, some games will need small modes that have anecdotal correction, but it's better than nothing, no?
We can also made an enableable "automatic Depth configuration" choice with a downloadable file list from our reports..
For quad render you can see an exemple by testing edgbla gpubladesoft* implementation it looks great with Tomb Raider, Spyro and many others games (there is an exemple in his youtube channel)
* this plugin need a good PC , so I'm not sure for you

In the emulators the only cool thing that I can MAYBE do is porting the mednafen's CPU Overclock, I understand how the code is made, but it seems too simple, I think I forgot something..

I have many many ideas for PCSXR emu:
-Memory Folder Like PCSX2 (because I don't like always change memory cards or have per game memory card):
Make per Profile memory folder.. when we start a game a virtual memory card is created and the saves for this game are included in.. (maybe add an option for including MGS compatible games easter egg???)
Because it's profile memory folder, the second memory card can be an other profile
-Texturing downscaling after the scaller:
many games downscale their textures with distance, the problem is that they use the original textures, and this new small texture is scalled by the emu: it looks disgusting (Tomb Raider in venice is a good exemple) track that and downscale the scalled textures (xbrz-ified) will be way better
I think it's impossible or it will be really buggy...
-Correct SBI compatibility:
Some sbi files works with ePSXe while not with PCSXR.. But they are read, I don't understant why it crash.. (These games are Spyro 2 and 3 PAL versions)

It makes me think that Pete, edgbla and ePSXe team aren't really cool with the emu community, ePSXe has the best compatibility and the best sound "plugin" (eternal works great but Sled Storm's musics are broken, or maybe it's due to PCSXR) but their emulator is rarelly updated.., Pete don't give us OpenGL 2 source code, I can understant but after all this years and tapec's tweaks it's not cool, and there is some gliches to fix.

Sorry I talked a bit too much about my life.

EDIT: I think 64bit version is for Linux

 

[zackboy]

@iCatButler Oh you already testing something like that..
I also imagined a precise OTZ but even if it's possible it will not correct the main issue of the OT: This is an average value of multuples vertice..
we have at worst the possibility to multiply the number of polygons.. (it's just a reduce leak of z-buffer)

Actually we can't just use w-component, the values will break background and skyboxes in almost all games and some games like Gran Turismo 2, Crash 3 and Spyro 1/2/3 have incoherent values..
For missing w-component values PGXP has just to be completed, but is it possible that some w-component values are wrong or there is no magic to hope for? (I'm talking about Gran turismo 2)

Did you see my video about Ordering Table, the values are just extremly close or it's an issue?
What's the problem with Tom Clancy's rainbow six ?

You say it's better to focus on w-component, basicly that's what I propose, The OT is just a base, I've said OT but it's the OTZ, using x values to y values to reorganise them by w-component, if we isolate only skyboxes and UI elements it's my "complete mode", the main issues with that is we need many configurations to prevent breaking some games and we can't have real Z value at the end but a fakeZ can be created.

@Mikhail000 @TheDimensioner The problem with the texture is the scaler used, the results are verry good but has some issues..
In Tomb Raider 3 and Spyro 1 some textures have black borders, but it's considered as corrected in the forums.. https://sourceforge.net/p/xbrz/forums/general/thread/59b98fd5/
Some type of textures are partially scaled (generally only the borders like CTR FF7...)

I think the scalling have a too high priority, I don't have framerate problems with PCSXR (exept a little with Silent Hill) but with PPSSPP I don't know what they made this lasts months but I have MANY frame drops with scalers..
I prefer a slow scaling than something that break my sound and speed..

In my opinion the best scaler is PPSSPP's Hybrid+Bicubic: based ,if i remember correctly, on xBRZ, HQx and bicubic..
The results are verry verry good but seems a little slower than xbrz...

Actually "Mem + CPU logic" distort or displace some "2D" elements the lasts builds from this year reduce a lot the distortion, but it stays not perfect yet...