Development Thread Testers wanted - tev_fixes_new

[neobrain]

Hello everyone,
You may recall my tev_fixes branch testing request. Well, it turned out the branch was ill-designed and different approaches had to be taken - more details will appear on our blog soon-ish.

Now, I finally finished the tev_fixes_new branch to a point where it's ready for general consumption. JMC has tested the branch extensively, but I'd like to see how it performs on other systems. The testing requirements outlined in tev_fixes still hold for the most part.

Basically, test as many obscure games as you can and check if they have any sort of pixel processing defects, e.g. transparent objects which shouldn't be transparent, incorrect colors, etc. To give you an impression of what issues are affected, see the list of issues known to be fixed by the branch.

If you report any issues, make sure to provide:
- a screenshot that shows the issue (preferrably marked with a red circle or something like that)
- a fifo log with which I can reproduce the issue.
Please do not report issues without providing these two things, it will effectively make the whole report useless! On the other hand, if I get these two things I'm sure I'll be able to debug the issue and at least say if it will get fixed or not.

Builds available here (last updated: March 4, 2014, 23:49 UTC):
32 bit
64 bit
Linux and OS X users can get the source from https://github.com/neobrain/dolphin/tree/tev_fixes_new .

Regards,
neobrain

[neobrain]

Bump!

Did we run out of testers or do I need to explain anything better? Fwiw, a "Tested X games, found no bugs!" response is as good as an actual issue report, because it improves confidence in having a wide range of systems without regressions.

[JMC47]

One thing I'd be interested to see is performance regressions/improvements on various games. Games like Super Smash Bros. Melee and Wind Waker run much faster for me. But Spyro: A Hero's tail runs like crap even in the menus on this branch, when it runs mostly okay on master even in game.

Here are a list of games with (known) regressions

NBA Live 2006 - Court Graphics are a fail (I got the fifo log from 2005)
NBA Live 2005 - Court Graphics are a fail - Fifo: https://dl.dropboxusercontent.com/u/484730/NBALive2005.7z Screenshot: https://dl.dropboxusercontent.com/u/484730/GLYE69-1.jpg

Here's a list of games I've at least played through one level/area on.

Animal Crossing
Alien Hominid
Batman Vengeance
Battalion Wars
Crash Bandicoot: Wrath of Cortex
Disney's Tarzan Unleashed
Disney's Cars
Donkey Kong: Jungle Beat
Donkey Konga 2
Ed, Edd 'n Eddy: Misedventures
F-Zero GX
F-Zero X
Final Fantasy Crystal Chronicles
Fortune Street
Four Swords Adventures
Godzilla: Destroy all monsters-melee
Harry Potter and the Goblet of Fire
Harry Potter and the Chamber of Secrets
Harvest Moon Magical Melody
James Bond 007 in Agent Under Fire
Kirby Air Ride
Lego Star Wars II
Lego Star Wars: The Video Game
Luigi's Mansion
Nintendo Puzzle Collection (Japanese)
Madden 2004
Madden 2006
Mario Golf
Mario Golf Toadstool Tour (at least, no worse than master)
Mario Kart Wii
Mario Kart Double Dash
Mario Party 2
Mario Party 4
Mario Party 5
Mario Party 6
Mario Party 7
Mario Party 8
Mario Party 9
Mario Power Tennis
Mario Power Tennis - New Play Control!
Mario Super Sluggers
Mario Superstar Baseball
Mario Tennis (VC)
MegaMan X Command Mission
MegaMan Anniversary Collection
MegaMan Network Transmission
Metroid Prime
Metroid Prime 2: Echoes
Metroid Prime 2: Echoes Bonus Disc
Metroid Prime Trilogy (Mostly played Metroid Prime 3)
Metal Gear Solid: The Twin Snakes
Midway Arcade Treasures 2 (all games)
Monopoly Party
Monster Hunter Tri
MVP Baseball 2005 (As good as master, at least)
Namco Museum
NCAA Football 2003
NCAA Football 2004
Need for Speed Underground
New Super Mario Bros. Wii
NHL 2003
Okami
One Piece Pirate Carnival
Pac-Man Fever
Pac-Man Vs (sorta, it's just like Master)
Pac-Man World 2
Paper Mario
Paper Mario: The Thousand Year Door
Pikmin 1 New Play Control!
Pokemon Channel
Pokemon Colosseum
Pokemon Puzzle League
Pokemon Snap
Pool Paradise
Prince of Persia: The Sands of Time
Phantasy Star Online: Episodes 1 & 2
Resident Evil 0
Resident Evil Code Veronica
Robots
Samurai Jack: The Shadow of Aku
Shrek 2
Skies of Arcadia Legends
Sonic Adventure 2 Battle
Sonic Adventure DX
Sonic Heroes
Sonic Riders
Sonic Riders Zero Gravity
Soul Calibur 2
Spiderman
Spongebob Squarepants: CFTKK
Spyhunter
Spyro: A hero's tail
Spyro: Enter the dragon fly
Starfox Adventures
Starfox Assault
Star Wars: Rogue Squadron
Super Bubble Pop
Super Mario Galaxy
Super Mario Galaxy 2
Super Mario Strikers
Super Mario Sunshine
Super Monkey Ball
Super Monkey Ball 2
Super Smash Bros (VC)
Super Smash Bros. Brawl
Super Smash Bros. Melee
Swingerz Golf
Tales of Symphonia
Tales of Symphonia - Dawn of the new World
Teenage Mutant Ninja Turtles
Tetris Party
Tetris Worlds
Texas Hold'em
The Ant Bully
The Chronicles of Narnia
The Hobbit
The Legend of Zelda: Twilight Princess (Wii)
The Legend of Zelda: Twilight Princess (GC)
The Legend of Zelda: The Wind Waker
The Legend of Zelda: Ocarina of Time Master Quest
The Legend of Zelda: Collector's Edition
The Legend of Zelda: Four Swords Adventures
The Legend of Zelda: Skyward Sword
The Legend of Zelda: Wind Waker (Japanese)
The Simpson's Hit and Run
Tiger Woods PGA Tour 2004
Tiger Woods PGA Tour 2005
Tiger Woods PGA Tour 2006
Tiger Woods PGA Tour 2009
Tomb Raider: Legend
Tony Hawk's Pro Skater 3
Tony Hawk's Pro Skater 4
True Crime: Streets of LA
Turok: Evolution
Valhalla Knights
Viewtiful Joe
Viewtiful Joe 2
Wario World
WiiPlay
Wii Sports
Wii Sports Resort
Worms 3D
WrestleMania X8
Xenoblade Chronicles
You Don't Know Jack

Tons of NES/SNES Virtual Console Games as well. A lot of these games have improvements on Tev Fixes New. Check the spreadsheet for noticeable ones.

[Link_to_the_past]

(01-31-2014, 09:18 PM)neobrain Wrote: Bump!

Did we run out of testers or do I need to explain anything better? Fwiw, a "Tested X games, found no bugs!" response is as good as an actual issue report, because it improves confidence in having a wide range of systems without regressions.

I couldn't find anything that JMC hasn't found, he did a very good test on his own, no wonder no one answers.

[KHg8m3r]

Using the same settings as the latest dev build, I'm getting some massive slowdowns using OpenGL on Nvidia. (I'm at school, and I'll update this later with what games)

[neobrain]

(01-31-2014, 09:54 PM)JMC47 Wrote: One thing I'd be interested to see is performance regressions/improvements on various games. Games like Super Smash Bros. Melee and Wind Waker run much faster for me. But Spyro: A Hero's tail runs like crap even in the menus on this branch, when it runs mostly okay on master even in game.

Here are a list of games with (known) regressions

NBA Live 2006 - Court Graphics are a fail (I got the fifo log from 2005)
NBA Live 2005 - Court Graphics are a fail - Fifo: https://dl.dropboxusercontent.com/u/484730/NBALive2005.7z Screenshot: https://dl.dropboxusercontent.com/u/484730/GLYE69-1.jpg

Here's a list of games I've at least played through one level/area on.
--- snip ---

Quite the comprehensive list, thanks!
As stated on IRC already, I debugged the NBA issue a bit but couldn't find a good fix, unfortunately.

(02-01-2014, 02:43 AM)Link_to_the_past Wrote: I couldn't find anything that JMC hasn't found, he did a very good test on his own, no wonder no one answers.

So everything runs fine for you? If so, that's also good to know, thanks!

(02-01-2014, 05:15 AM)KHg8m3r Wrote: Using the same settings as the latest dev build, I'm getting some massive slowdowns using OpenGL on Nvidia. (I'm at school, and I'll update this later with what games)

I'm aware of a speed decrease in this branch. We do think the slowdown is justified compared to the benefits which this branch gives us, though.
That said, having any actual numbers on the slowdown is nevertheless well-appreciated. Make sure to also post your internal resolution and AA settings, though (as well as whether or not per-pixel-lighting is enabled, and which other things from the "Enhancement" tab in the graphics configuration).
EDIT: Also, you only said the slowdowns where happening in OpenGL. What about D3D?

[JMC47]

D3D suffers from even worse slowdowns. It's very very worrisome and varies per game. Games like Melee run faster, but then Four Swords Adventures loses like a thousand FPS in the menus

[drhycodan]

I've done some testing on this and while I haven't come across any bug that isn't in the master branch, there seems to be a 10-15% performance drop in general, more on certain games. The Last Story seems to have gotten hit pretty hard on the performance. FMVs no longer play smooth like in the master branch

[KHg8m3r]

Update with some testing:
On both: Renderer to Main Window, Scaled EFB copy, Per-pixel Lighting, OpenMP, Fast Depth Calculation, Enable Progressive scan, all else default
FPS/VPS used to show speed

Using D3D @ 3x IR, 8 samples (quality 32) AA, 16x AF:

Games that run fine:
-LEGO Star Wars II: 60/60 all times
-Mario Kart DD: 60/60
-One Piece Grand Adventure: 60/60 all
-Paper Mario Thousand year door: 60/60 all times
-Smash Bros Brawl: 60/60 all
-Sonic Colors: 60/60 all, 30/60 level (thats regular)
-Tales of Symphonia: 60/60 all,
-LoZ WW: 30/60 all

Games with few FPS drops:
-Pokemon Colosseum: 1-2 fps drops ever now and then, makes walking lag and stutter (not as bad as OpenGL)
-Pokemon XD: same
-Super Mario Sunshine: 25/52 all
-Smash bros Melee: 60/60 all, some stuttering randomly
-LoZ TP: 28/59 all


Using OpenGL @ 3x IR, 4x SSAA AA, 16x AF:

Games that run fine:
-Paper Mario Thousand year door: 60/60 all

Games with few FPS drops:
-Pokemon Colosseum: 2-3 FPS drop walking around, makes walking lag and stutter
-Pokemon XD: Same as Colosseum
-Tales of Symphonia: 20/20 main menu, 45/45 battle, 60/60 walking

Massive FPS drops:
-LEGO Star Wars II: around 25/32 all the time
-Mario Kart DD: 50/50 menus, 30/30 to 45/45 races
-One Piece Grand Adventure: 30/30 menus, 35/35 fights
-Smash Bros Brawl: fluctuates around 20/20
-Sonic Colors: 20/60 title screen, 14/60 level
-Super Mario Sunshine: 15/30 all (fluctuates)
-Smash bros Melee: around 18/18 menus/stage select, 25/25 to 30/30 char select/battle
-LoZ WW: 20/40 all
-LoZ TP: 12/25 all


No bugs besides what's in master

[JMC47]

Your testing is kinda flawed. OGL is hit harder by higher internal resolutions. you need to do a control test at 1x IR with no enhancements to see the whole story.