[UNOFFICIAL] Ishiiruka-Dolphin Custom Version

[havardp]

(04-18-2016, 11:47 AM)Kamikaze_Ice Wrote: This isn't an issue with Ishiiruka, so you should ask about it in the PCSX2 forums (Assuming thats where you got it from, most development for DS3/4 comes from there.
I know of only two currently running forks of DS4Windows, and both are forks of electrobrains' fork of the original DS4Windows by InhexSTER, which was spiritually forked (read: inspired by) from Scarlet.Crush's DS3 tool. We need to Pimp a fork so we can fork while we fork to fork a fork of our fork.... maybe to many forks?
One is Jays2Kings, still using DS4Windows name, and the other is Input Mapper.
I recommend you seek further assistance from their forms.

But here's my 2 cents:
I bet your issue is with your laptops bluetooth stack. Not all are compatible with the controller. Now if you're using the same dongle on both, then it's probbably Windows screwing around with your driver without telling you. I ran into this issue after an upgrade from Windows 7 to Windows 8.1 (fresh install). I had to jump through a few hoops to make Windows stop installing their version of the driver included with Windows over my manually installed driver. This is another reason to seek further advice in their forums as they know more about it because that's where it's developed and would know of any other outstanding issues or weird behavior of drivers like this.

I'm using jays2kings fork, with a usb cable, so bluetooth isn't the problem. What's weird is that it works with dolphin on both PCs, but only the gaming pc works with ishiiruka, so i just wanted to hear if this was some known issue. the controller also works with normal games on the laptop. I did some weird multi monitor stuff with my laptop and TV, so that may have messed with things, i honestly have no clue. Either way i'm getting a steam link next week, so hopefully it will work with that appreciate the reply

[Kamikaze_Ice]

With all these new toys to play with, I mean the texture scaling modes, I decided to take screenshots so I can compare all modes under various settings to see which one I wanted to use.

All screenshots taken in the latest Ishiiruka (637 as of this post).
DirectX11 backend (OpenGL fails with the game's depth of field. and I can't use DirectX12... yet)
1920x1080 display resolution
3x Internal Resolution + Borderless Fullscreen (not exclusive fullscreen)
16x Anisotropic Filter
Texture Cache Accuracy: Normal (the middle position. Not "safe" or the default of "fast").  *Off-topic: Is this still a thing needed for this game? DoF blur works in DX11 at any setting but fails in OGL (works if you either start the game once, stop emulation and then restart the game without restarting the emulator or toggle Fast Depth Calculation. Well I say works, but what I really mean is that screwed up blur is not there. I don't know if the Dof is actually working or not). /off-topic*
No post-processing or scaling.
No texture packs.
Each mode is only at "2x", as I found some modes had no visible changes past 2x but the performance hit still increased.


Note #1: When dumping textures to make a different Playstation button texture pack I noticed a few bump maps were among what little I did dump (two sword, water and one enemy). I know I'm missing some textures I saw in the menu, which were obviously loaded in memory, so I'm assuming the environment bump maps exist because I've typically seen bump maps in the reverse order (environment and not weapons/enemies). Results of some scaling modes also supports this theory, like DDT-Sharp, but haven't been able to dump them. If I'm wrong, then I'll reassess and fix it later lol
Note #2: To lightly support the previous note, I can only dump 14 files total out of my FIFO file, and except for the text and sword none of them are shown on screen. It completely fails to dump the same number of textures as the actual game at the same point. If this is a FIFO limitation, then ignore this note.


My FIFO file. Used this for all screenshots for consistency and control so I can focus on the exact same frame without worry of minor engine calculation differences (because I'm an anal nitpicker).

Scaling Mode (OFF):No AA 2xMSAA 2xSSAA 2xSGSSAA
For reference.

xbrz: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
The HQx-ness of xbrz makes things look like it was made with a paintbrush, which is fantastic for 2d/sprite/non-realistic games, but not a good choice for this game..

Hybrid: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
I like this more than xbrz, but still not suited for this game. Also, what is this really? xbr/xbrz + reverse anti-aliasing?

Bicubic: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
Very nice. Look at the ground by the swords in this and compare how pixels transition from one to another. No more pyramid/cross blending! And because of that this deals with with texture aliasing/pixelation without becoming blurry.

Hybrid Bicubic: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
Midrange crush/level flattening = loss of detail = image degradation. Very noticeable on the ground around the swords.
Is this literally "Hybrid + bicubic" or "xbr/xbrz + bicubic"?

Out of the four original modes, I would use Bicubic everywhere, with special exceptions for hybrid/xbrz depending on a game's art direction.
Now for the new toys!

Jinc: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
Also, very nice. Color samples in addition to what bicubic does (much smoother). Very close to Bicubic, so this could probably use some extra sharpness added in.

Jinc-Sharper: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
Ack. Too much sharpness! Also, tattoo now has visible ringing (darker edge on tattoo side with lighter edge on skin side).

Smoothstep: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
Note: I have no clue what "smoothstep" is or what it's goal is, and my googlefu only gives me text explanations that I'm too stupid to understand (I'm a visual person, all the text, formulas and graphs explaining the concept of smoothstep tell me nothing without a picture to act as a relative anchor to tie in to).
This is probably a good one to use with if you use any FSAA, downscaling, and/or with other shaders, but by itself it appears to be as close as one can get to being unfiltered without actually being unfiltered.

3-point: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
Note: Again, no clue about this or it's goal.
Do you have motion sickness? Yes? Well, go on and look! Dat motion blur! *barf* Heh, sorry. Like I said, I don't know anything about this one, but it's clearly not suited for this game.

DDT: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
Somewhat similar to xbr but vastly superior for this art direction--No more paintbrush appearance! I really like it vs xbrz/hybrid on the ground texture, but there is a few out-of-place pixels in the tattoo here.
@Hyllian, do you think this one could be refined a bit more? I think it could use two things: I) A small amount of "rounding" (like how bicubic "rounds" bilinear) and II) if you could preserve some of the original contrast. Look at the Chest's lid brackets and those single brightest pixels of the ground around the swords to see what I mean (can't tell if caused by bump map, which may complicate things). Do remember I don't know anything about what's technically involved here. I just wanted to share my opinion because quite this one

DDT-Sharp: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
Yuck, look at what this does to the ground bump map. Reminds me of a metal plate floor. Arm tattoo, actually anything without a bump map looks much cleaner.

My favorites for this game: Bicubic, DDT and Jinc. Still undecided what I want to do with this game, "improved and preserve" the original look (Bicubic/Jinc) or "enhanced" it (DDT), and what post-processing shaders I want to use. I'll wait til I get to some town where it supposedly has the worst performance before touching the shaders though.


OK class, what did we learn today?
That Hyllian has made a bunch of damn good shaders. Seriously, thanks man.
Didn't know you had anything besides xbr and what you've posted in here until I used my googlefu for xbr, xbrz and hybrid.

[Kamikaze_Ice]

While doing my comparisons I was going all OCD only on two areas: Tattoos, and the ground around the swords. At the time I was looking to find out which scaling mode best complimented SGSSAA and cross referencing the modes with SSAA to check for any artifacts that might have been caused by forcing SGSSAA in The Last Story.
By pure happenstance my cat decided he needed some attention and jumped onto my keyboard which changed where I was looking at in GIMP.
I would NOT have noticed it otherwise, but thanks to my cat I did notice and now I can't unsee it.
This is present in all the screenshots in my previous post, to varying degrees.
I really have no idea on how to describe what I'm seeing, so bear with me.

First, I want to point out that there is difference between SSAA and SGSSAA, but I strongly feel that this is not the problem and is also not what I'm trying to explain.
No scaling mode: 2xSSAA vs 2xSGSSAA - Even at 1600% zoom it's extremely hard to see the difference in the circled areas, so keep this in mind.
Hybrid-bicubic mode: 2xSSAA vs 2xSGSSAA - Now you can see the difference. On the right it's obvious enough to notice without magnification and even more obvious if you zoom in.
The first comparison shows how similar both FSAA methods are. This holds true for most of the frame in the second comparison, except there are a few areas that are suspiciously different.

I don't think a mipmap texture or Anisotropic filtering is the culprit as it is a very sharp transition and doesn't apply to other parts of the texture at the same distance and angle from the camera. Not to mention that the game's mipmaps are all the same resolution.
I don't think it's the scaling modes either, as there is still a difference, an extremely SMALL difference, in the same areas without a scaling mode.
I don't think AA is it either, since you can still see changes if you compare No AA to MSAA.(Bth circled areas are not an edge so MSAA does not touch it.
I'm unable to confirm/deny my stance on AA, since enabling wireframe to check geometry edges looks like THIS. There appears to be something rendering below the UI that blocks my view of the environment wireframe (happens in Master, too). Possible relation?


Interestingly, if you set the external frame buffer to "real" there is a green line on the screen. This reminds me of old, badly encoded videos which means something it not quite right.
Green line is not there if set to "Virtual", and when disabled the screen gets stretched vertically by however thick that green line is. I wonder which is the more accurate aspect (virtual or disabled). It's only like 10 pixels but still

Other comments regarding Ishiiruka 637:
1) Thanks for continuing to share your work with us.

2) The texture overlay is different from Master (4.0-9202). One corrupt tag? See: Master vs Ishiiruka
Note: Only tried 637 and 528.

3) Ishiiruka's FIFO player has problems recording The Last Story. No problem recording and playing with Xenoblade on 637, or even playing/recording from different builds or even between Master/Ishiiruka, so it must be something with the game in general.
"Dual Core" mode is not a problem, although it used to be. Enabled/disabled makes no difference. Just in case I went ahead and have done everything in this and the previous post was done with it disabled.
Tried current (637), 628, 570, 541, 528 and 524. Same thing--Unknown Opcodes.

Despite the errors, it does not crash or hang Ishiiruka and appears to work just fine (can make changes to settings or load/unload custom textures just fine). Although I can't dump all the textures (which may or may not be a problem limited to the FIFO player).
My FIFO recording (same as previous post)

Here is what current (637) tells me:

Code:

52:13:990 Boot\Boot.cpp:248 N[BOOT]: Booting C::\SUPERSECRETPATHTOGAMESFORCOOLPEOPLES\Dolphin\FIFOs\last story (1 frame for SS).dff
52:15:371 BPStructs.cpp:725 W[Video]: Unknown BP opcode: address = 0x00000046 value = 0x00000273
52:15:371 BPStructs.cpp:725 W[Video]: Unknown BP opcode: address = 0x00000069 value = 0x000004ed
52:15:371 MsgHandler.cpp:88 E[*]: Warning: GFX FIFO: Unknown Opcode (0x72 @ 0000000017644E68, preprocessing=preprocess=false).
This means one of the following:
* The emulated GPU got desynced, disabling dual core can help
* Command stream corrupted by some spurious memory bug
* This really is an unknown opcode (unlikely)
* Some other sort of bug

Further errors will be sent to the Video Backend log and
Dolphin will now likely crash or hang. Enjoy.
53:34:607 MsgHandler.cpp:88 E[*]:Warning: Illegal command 72
CPBase: 0x00c14720
CPEnd: 0x00c94700
CPHiWatermark: 0x0005ffe8
CPLoWatermark: 0x00000000
CPReadWriteDistance: 0x00000020
CPWritePointer: 0x00c195a0
CPReadPointer: 0x00c19580
CPBreakpoint: 0x00000000
bFF_GPReadEnable: true
bFF_BPEnable: false
bFF_BPInt: false
bFF_Breakpoint: false
bFF_GPLinkEnable: true
bFF_HiWatermarkInt: false
bFF_LoWatermarkInt: false

53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x72 @ 0000000017644E68, preprocessing = no)
53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x65 @ 0000000017644E6A, preprocessing = no)
53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x6f @ 0000000017644E71, preprocessing = no)
53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x66 @ 0000000017644E73, preprocessing = no)
53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x6d @ 0000000017644E7A, preprocessing = no)
53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x72 @ 0000000017644E7C, preprocessing = no)
53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x6e @ 0000000017644E83, preprocessing = no)
53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x5c @ 0000000017644E8A, preprocessing = no)
53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x65 @ 0000000017644E8C, preprocessing = no)
53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x63 @ 0000000017644E8E, preprocessing = no)
53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x77 @ 0000000017644E95, preprocessing = no)
53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x5e @ 0000000017644E9C, preprocessing = no)
53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x72 @ 0000000017644E9E, preprocessing = no)
53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x5f @ 0000000017644EA5, preprocessing = no)
53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x69 @ 0000000017644EA7, preprocessing = no)
53:54:086 OpcodeDecoding.cpp:352 E[Video]: FIFO: Unknown Opcode(0x6e @ 0000000017644EA9, preprocessing = no)

[*]


I only tried one version of Master (4.0-9202), which has no outstanding issues (warning/error notification boxes) and has this to me:

Code:

57:34:462 Boot\Boot.cpp:248 N[BOOT]: Booting C::\SUPERSECRETPATHTOGAMESFORCOOLPEOPLES\Dolphin\FIFOs\last story (1 frame for SS).dff
57:35:887 BPStructs.cpp:684 W[Video]: Unknown BP opcode: address = 0x00000046 value = 0x00000273
57:35:887 BPStructs.cpp:684 W[Video]: Unknown BP opcode: address = 0x00000069 value = 0x000004ed

[Blitzkri3g]

This project is amazing! I have a suggestion.

Do you think you can port the "Exclusive fullscreen mode" from DX11 to DX9? DX9 has a noticeable ammount of Input lag compared to DX11 (even though it works better on low specs PC).

Thanks in advance

[Hyllian]

Very good impressions, Kamikaze_Ice! Thanks! I'll comment some of them below:

(04-18-2016, 01:05 PM)Kamikaze_Ice Wrote: Bicubic: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
Very nice. Look at the ground by the swords in this and compare how pixels transition from one to another. No more pyramid/cross blending! And because of that this deals with with texture aliasing/pixelation without becoming blurry.

Out of the four original modes, I would use Bicubic everywhere, with special exceptions for hybrid/xbrz depending on a game's art direction.
Now for the new toys!

Did you notice speed improvements from older versions?

(04-18-2016, 01:05 PM)Kamikaze_Ice Wrote: Jinc: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
Also, very nice. Color samples in addition to what bicubic does (much smoother). Very close to Bicubic, so this could probably use some extra sharpness added in.

Jinc-Sharper: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
Ack. Too much sharpness! Also, tattoo now has visible ringing (darker edge on tattoo side with lighter edge on skin side).

I made some changes to the original implementation. I made an anti-ringing, and toned down the sharpness a bit, so In the end I made two versions of the same scaler: jinc and jinc-sharper. The last one is for those who like sharp contrasts.

(04-18-2016, 01:05 PM)Kamikaze_Ice Wrote: Smoothstep: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
Note: I have no clue what "smoothstep" is or what it's goal is, and my googlefu only gives me text explanations that I'm too stupid to understand (I'm a visual person, all the text, formulas and graphs explaining the concept of smoothstep tell me nothing without a picture to act as a relative anchor to tie in to).
This is probably a good one to use with if you use any FSAA, downscaling, and/or with other shaders, but by itself it appears to be as close as one can get to being unfiltered without actually being unfiltered.

Smoothstep is a known function by shader/3d developers. Visually, It's in the middle between nearest neighbor and bilinear. It's called smoothstep because it smoothly change from one pixel to the next without introducing blurriness (bilinear does that). And it isn't as sharp as nearest neighbor, so it doesn't hurt your eyes. Smoothstep as a texture scaler is explained here: http://www.iquilezles.org/www/articles/texture/texture.htm

(04-18-2016, 01:05 PM)Kamikaze_Ice Wrote: 3-point: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
Note: Again, no clue about this or it's goal.
Do you have motion sickness? Yes? Well, go on and look! Dat motion blur! *barf* Heh, sorry. Like I said, I don't know anything about this one, but it's clearly not suited for this game.

It's a bilinear using only 3 points (standard bilinear samples four points). It's the original N64 texture scaler (yes, try using it with N64 games and you'll be reminded of the games you played in your childhood!). Try Zelda Ocarina of Time with 3-point and you'll remember how the HUD and textures looked like.

(04-18-2016, 01:05 PM)Kamikaze_Ice Wrote: DDT: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
Somewhat similar to xbr but vastly superior for this art direction--No more paintbrush appearance! I really like it vs xbrz/hybrid on the ground texture, but there is a few out-of-place pixels in the tattoo here.
@Hyllian, do you think this one could be refined a bit more? I think it could use two things: I) A small amount of "rounding" (like how bicubic "rounds" bilinear) and II) if you could preserve some of the original contrast. Look at the Chest's lid brackets and those single brightest pixels of the ground around the swords to see what I mean (can't tell if caused by bump map, which may complicate things). Do remember I don't know anything about what's technically involved here. I just wanted to share my opinion because quite this one

DDT means Data Dependent Triangulation. It is a step ahead of 3-point interpolation. In 3-point, the triangles formed by the three points are always the same, formed by the triangles slices by the diagonal upper-right and down-left. In DDT, it compares the two main diagonals to see where the edge is, and then chooses the 'correct' diagonal to slice the square and form the two triangles from where it'll sample the three points necessary for interpolation.

(04-18-2016, 01:05 PM)Kamikaze_Ice Wrote: DDT-Sharp: No AA, 2xMSAA, 2xSSAA, 2xSGSSAA
Yuck, look at what this does to the ground bump map. Reminds me of a metal plate floor. Arm tattoo, actually anything without a bump map looks much cleaner.

Well, this uses some more complex heuristic to decided which diagonal should be chosen to slice the square. I like it more than plain DDT. Probably your example isn't very good for this texture scaler, but I wouldn't discard it yet. Try in other games.

(04-18-2016, 01:05 PM)Kamikaze_Ice Wrote: My favorites for this game: Bicubic, DDT and Jinc. Still undecided what I want to do with this game, "improved and preserve" the original look (Bicubic/Jinc) or "enhanced" it (DDT), and what post-processing shaders I want to use. I'll wait til I get to some town where it supposedly has the worst performance before touching the shaders though.

I think it depends very much on the kind of graphics. There are some games where 3-point does the job very well (N64 games). Overall, I think that Jinc has a more general usage. DDT-Sharp has some potential with some games (the ones full of HUDs).

(04-18-2016, 01:05 PM)Kamikaze_Ice Wrote: OK class, what did we learn today?
That Hyllian has made a bunch of damn good shaders. Seriously, thanks man.
Didn't know you had anything besides xbr and what you've posted in here until I used my googlefu for xbr, xbrz and hybrid.

You're welcome!

[Kamikaze_Ice]

I wasn't planing on playing with the post-processing shaders yet for The Last Story, but I was curious as to how depth buffer shaders would work.

I'll try to keep this post short and to the point.
I just reused my shader settings from Xenoblade for these screenshots, except where Barrel was used (which was used alone, with no other shaders).

Trigger=Swap
http://i.cubeupload.com/j1RVqB.png
http://i.cubeupload.com/x1gOw3.png
Man, even in Swap SSAO is screwed, making some "scanline" aftifacts.
I compared disabled vs virtual extrnal frame buffer, thinking the difference I saw between them would close the gaps.
Nope, but it does shift where these "scanlines" appear.
Also, a few tips of Dagran's hair pokes "through" the dialog window (I found it amusing)

Trigger=Projection
http://i.cubeupload.com/dkutAV.png
http://i.cubeupload.com/LogQqB.png
Shaders are broken. Loading any just removes the games bloom effect. Even ze Barrel glasses, zey do nothing!.

Trigger=EFB Copy
http://i.cubeupload.com/Qp05v9.png
http://i.cubeupload.com/UrCBWy.png
Here's what I used for Xenoblade. This game has a box stuck on screen (the/a frame buffer). This looks like something that happened to Mirrors in Luigi's Mansion a few months ago. I don't know what frame buffer effect belongs in that box. In game bloom works outside of it, but any post-processing shaders fail to work inside the box (like it's a mini Projection trigger zone), which explains what happens if you were to use the BARREL shader.

Trigger=After Blit
http://i.cubeupload.com/33Gaq4.png
http://i.cubeupload.com/YsaEG3.png
Barrel shape is different (swap and projection have the same shape) and yes thats with the same settings.
The "scanline" effect is stronger and larger in "After Blit" vs "Swap".


Such disapointment. This game really hates shaders

[Kamikaze_Ice]

(04-18-2016, 01:45 PM)Hyllian Wrote: Very good impressions, Kamikaze_Ice! Thanks! I'll comment some of them below:

Did you notice speed improvements from older versions?

It's hard to say as my system is being bottlenecked by my ram speed (DDR3 at a lowly 1600), so as soon as I reach that point everything just falls on it's face instead of getting slower.
I generally notice one of two things with Dolphin/Ishiiruka. I get either a huge impact (only from AA/SSAO/SSGI) or a super minor impact (1-2 frames or so) to actual framerates. And the wild card is some combinations with some games are nearly impossible for me to measure reliably. Texture scaling mode 2x vs 5x is a good example. My framerates, hardware sensors, activity monitors and other things all show good results on both 2x and 5x, but I think I get more stuttering at 5x despite what everything else tells me. However when I experience it, it also happens the same time as the emulator is loading something (texture, zone, etc), compiling a shader, or being caused by background processes (like firefox as it does a garbage collection to free up ram, which I never close).

(04-18-2016, 01:45 PM)Hyllian Wrote: I made some changes to the original implementation. I made an anti-ringing, and toned down the sharpness a bit, so In the end I made two versions of the same scaler: jinc and jinc-sharper. The last one is for those who like sharp contrasts.

I assumed as much. I just thought that somewhere in between those would be more appropriate well for The Last Story. Just sharing my opinion is all. I wonder if we will at some point get an options menu to the Scaling modes like what the post-processing "scaling" shaders have. Then we could fine tune them for each game (and to each user's aesthetic taste).

(04-18-2016, 01:45 PM)Hyllian Wrote: Well, this uses some more complex heuristic to decided which diagonal should be chosen to slice the square. I like it more than plain DDT. Probably your example isn't very good for this texture scaler, but I wouldn't discard it yet. Try in other games.
...
I think it depends very much on the kind of graphics. There are some games where 3-point does the job very well (N64 games). Overall, I think that Jinc has a more general usage. DDT-Sharp has some potential with some games (the ones full of HUDs).

I agree, the art direction of a game very important as different methods achieve different things. I didn't intend to come across as saying otherwise.

[Tino]

Tanks for the review nice work. about your questions:

(04-18-2016, 01:11 PM)Kamikaze_Ice Wrote: 2) The texture overlay is different from Master (4.0-9202). One corrupt tag? See: Master vs Ishiiruka
Note: Only tried 637 and 528.

that is because master and ishiiruka works diferently when using cmpr textures, master transforms everything to rgba, Ishiiruka transform cmpr(dxt1) to dxt3 that transcoding allow for a faster loading and a reduced memory usage with compressed textures, I only use rgba when a texture scalling algorithm is selected because scalling spects rgba data.

(04-18-2016, 01:11 PM)Kamikaze_Ice Wrote: 3) Ishiiruka's FIFO player has problems recording The Last Story. No problem recording and playing with Xenoblade on 637, or even playing/recording from different builds or even between Master/Ishiiruka, so it must be something with the game in general.

"Dual Core" mode is not a problem, although it used to be. Enabled/disabled makes no difference. Just in case I went ahead and have done everything in this and the previous post was done with it disabled.
Tried current (637), 628, 570, 541, 528 and 524. Same thing--Unknown Opcodes.

Thanks for the report and the fifolog i will try to debug the problem to see what is corrupting the fifolog.

(04-18-2016, 01:11 PM)Kamikaze_Ice Wrote: Man, even in Swap SSAO is screwed, making some "scanline" aftifacts.

the scanline artifact is a known issue about ssao, you can fix it tunning the settings, specially  filter radius and quality/sample count

(04-18-2016, 01:11 PM)Kamikaze_Ice Wrote: Here's what I used for Xenoblade. This game has a box stuck on screen (the/a frame buffer). This looks like something that happened to Mirrors in Luigi's Mansion a few months ago. I don't know what frame buffer effect belongs in that box. In game bloom works outside of it, but any post-processing shaders fail to work inside the box (like it's a mini Projection trigger zone), which explains what happens if you were to use the BARREL shader.

the game is doing some weird things with the efb and it seems to  overwrite part of it, the bad news it will be really dificult to fix this because we can not change how the game works.

[Tino]

@Radu91: thanks for the report, will take a look at the changes in the version after 584 too see if i found a change that can explain the difference.

[Tino]

updated latest folde with a new version.
@Radu91 the issue you reported should be fixed now.
Testing is welcome.
Enjoy