[Soft-Float] - Initial Interpreter Implementation of Ps2's floating point unit specification #12001

GitHubProUser67 · 2024-11-12T20:12:27Z

This Pull Request implements the first take ever on real Soft-Float support in PCSX2.

This work is a combination or several efforts and researches done prior.

Credits:

https://www.gregorygaines.com/blog/emulating-ps2-floating-point-nums-ieee-754-diffs-part-1/
https://github.com/assumenothing/Tommunism.SoftFloat/tree/main
https://github.com/GitHubProUser67/MultiServer3/blob/main/BackendServices/CastleLibrary/EmotionEngine.Emulator/Ps2Float.cs
https://github.com/Goatman13/pcsx2/tree/accurate_int_add_sub
PCSX2 Team for their help and support in this massive journey.

This pull request should be tested with every games requiring a clamping/rounding mode/float patches (cf: GameDatabase).

Currently, this PR fixes on the interpreters:

Any other games using the NegDiv hack or any other FPU rounding mode.

This sets the floor for Soft-Float in PCSX2, a long awaited contribution.

github-actions

Thank you for submitting a contribution to PCSX2

As this is your first pull request, please be aware of the contributing guidelines.

Additionally, as per recent changes in GitHub Actions, your pull request will need to be approved by a maintainer before GitHub Actions can run against it. You can find more information about this change here.

Please be patient until this happens. In the meantime if you'd like to confirm the builds are passing, you have the option of opening a PR on your own fork, just make sure your fork's master branch is up to date!

seta-san · 2024-11-12T20:46:56Z

Does this work on the recompilers or just interpreter?

refractionpcsx2 · 2024-11-12T20:48:09Z

You should try reading the title.

…oint unit specification. This Pull Request implements the first take ever on real Soft-Float support in PCSX2. This work is a combination or several efforts and researches done prior. Credits: - https://www.gregorygaines.com/blog/emulating-ps2-floating-point-nums-ieee-754-diffs-part-1/ - https://github.com/GitHubProUser67/MultiServer3/blob/main/BackendServices/CastleLibrary/EmotionEngine.Emulator/Ps2Float.cs - https://github.com/Goatman13/pcsx2/tree/accurate_int_add_sub - PCSX2 Team for their help and support in this massive journey. This pull request should be tested with every games requiring a clamping/rounding mode (cf: GameDatabase). Currently, this PR fixes on the interpreters: - PCSX2#354 - PCSX2#11507 - PCSX2#10519 - PCSX2#8068 - PCSX2#7642 - PCSX2#5257 This is important to note, that this implementation, while technically fixing Gran Turismo 4 and Klonoa 2, makes the games crash due to very high floats being passed in the emu code, and failing at some points later in the process. This has not yet been ironed-out. Other than that, this sets the floor for Soft-Float in PCSX2, a long awaited contribution.

seta-san · 2024-11-12T20:51:02Z

You should try reading the title.

I don’t know why my brain just skimmed over that.

MrCK1 · 2024-11-12T23:33:42Z

I ran a bunch of tests for "Test Drive Unlimited" AI during the demo scene after sitting idle on the menu. No combination of settings/interpreters seems to have any effect on behavior. There must be something else going on

AmyRoxwell · 2024-11-13T04:55:20Z

I can confirm with the multiplication setting that MHG and Dos not longer have bugged bounces! (Plus the build works on linux well! [as well as running interpreter can do, of course :'D])

Shoegzer · 2024-11-13T05:01:11Z

Shouldn't this help with #2990 as well?

AmyRoxwell · 2024-11-13T05:12:18Z

Shouldn't this help with #2990 as well?

I remember seeing on the public dev channel that stuntman not longer had AI issues with this and not longer the car AI failed? it's been a while

Blackbird88 · 2024-11-13T08:37:52Z

Tourist Trophy NTSC works now too! For the first time License Tests in it work!
Requires: EE MUL/DIV

Some of the tests still don't work however.

Works

Hangs

Tokman5 · 2024-11-13T08:40:25Z

I tested the demo replays of Tokyo Xtreme Racer Zero (see issue #5597 ) and noticed that the car movement in interpreter mode is now closer to the movement in recompiler mode.
However, there are still slight differences between the two, and both are far from matching the console playback.

comp.mp4

Goatman13 · 2024-11-13T09:51:24Z

Nice work!

I tested only 2 games for now. Hype Time Quest which if i remember correctly needs accurate fpu mul, and it works fine without patch.

Second game is THPS4. While here accurate VU1 add/sub fix issue which is normally fixed by VU1 rounding in database, accurate mul/div breaks game graphics. Of course, game don't need accurate VU1 mul/div to work correctly, but something seems to be off when enabled.

Accurate vu1 add/sub/sqrt:

Accurate vu1 mul/div (same place in game):

Edit: Maybe THPS4 is just passing some big float value to GS with accurate mul/div. That may not necessarily be an issue with the float operation itself.
Additionally, I tested Burnout 2 and accurate VU1 add/sub fixed issue that was previously fixed by VU1 rounding mode (white car parts).

Edit2: Freaky Flyers is fixed with accurate mul/div on VU1. Awesome! We weren't even sure if that's floating points issue until now.

The game sends some super low floats to the Mul unit. On PS2, floats with exponent zero should return zero, but this is not the case in Mul, the multiplier can work with denormals internally. I love when undocumented stuff is used by some games for their 3D engine ^^.

GitHubProUser67 · 2024-11-13T15:10:48Z

The Tony Hawk case is fixed, the game uses an un-documented behaviour in it's 3D engine.

The PS2 has no denormals support .... except in the Mul unit apparently.

The behaviour is now emulated properly.

Shoegzer · 2024-11-13T16:33:00Z

I remember seeing on the public dev channel that stuntman not longer had AI issues with this and not longer the car AI failed? it's been a while

@AmyRoxwell Can you provide a reference to this? There's no indication of it being fixed in #2990, and I assume the devs would have closed it if it were. In any event it involves pathing in Driver 3 as well.

weirdbeardgame · 2024-11-13T17:10:03Z

2024-11-10.14-28-32.mp4

This also affects the Fatal Frame 1 issue. Meaning this + the current GameDB patch will end up being the ultimate fix

More accurate approach to compare.

Goatman13 · 2024-11-13T18:25:53Z

#3200 is fixed when tested with b7f3806
Require accurate mul/div for FPU.
I didn't tested, but Krome studio games should profit from this pr too, ones that have patches in game db. Like Spyro, Star Wars, one of Transformers game.

ghost · 2024-11-13T19:15:33Z

This pr's EE interpreter fixes #11636 's gamedb issue.

Implements accurate SQRT options, also removes Tri-Ace hack, which isn't needed anymore on the interpreter.

AmyRoxwell · 2024-11-13T20:39:32Z

I remember seeing on the public dev channel that stuntman not longer had AI issues with this and not longer the car AI failed? it's been a while

@AmyRoxwell Can you provide a reference to this? There's no indication of it being fixed in #2990, and I assume the devs would have closed it if it were. In any event it involves pathing in Driver 3 as well.

I meant like, while using this PR, not that is has been fixed. Sorry if it was misunderstood. But if it's not mention on the PR maybe the thing it needs it's not here by this initial implementation.

GitHubProUser67 · 2024-11-13T20:42:35Z

Driv3r seemed fine when it was tested, Stuntman NTSC is a lot better but still can "slightly" deviate.

I suspect it is once again, the interpreter rounding/clamping values somewhere.

Shoegzer · 2024-11-13T21:34:36Z

@AmyRoxwell Ah, I understand you now. That's great news.

@GitHubProUser67 Thanks, nice to see Driv3r is looking better. Would it make sense to list these games in your OP?

Goatman13 · 2024-11-24T09:45:34Z

I would say the main issue came from the fact it was widely assumed the PS2 and PS3 shared the same "float" calculation behaviours, while in actual fact, they do not.

That was the plan as far as I know. Probably someone underestimated how important 1:1 implementation was to keep compatibility. PS3 CELL implemented special Altivec/VMX mode to make PPE core vector floats compatible with SPU, but also with PS2 emulation in mind. Later it turned out that they really miscalculated importance of accurate floating point math in PS2 emulation. To the point that first fixes in ~2008-2010 were strictly per game, like "on this pc, add 0x01 to this reg, and get back execution to recompiler". Soft floats, including accurate DIV, came later, when new people were hired. :) In the end only VU1 is using real SPE, everything else use PPE VMX compatibility mode.

Current Div/Sqrt system is exactly like a PS3 emulating the PS2 in therms of results, which means way closer (and fixes the VAST majority of games), but still one bit off in some edge cases.

Add/Sub/Mul are PS2 perfect.

And that's indeed a great outcome. Nice work. :)

GitHubProUser67 · 2024-12-07T21:42:52Z

Thank you Goatman13! The PCSX2 team also gets massive credits here for their help and support, of course the goal is to be PS2 perfect on this one. There is this one game, Mortal Combat Shaolin Monks which doesn't work fully yet on the Soft-Float div due to the game doing loads of Log2 on top of existing Log2 results, leading to it relying on rounding in booth directions...

…eed-up simple calculations. This greatly improves performance while using Soft-Floats.

TellowKrinkle · 2024-12-10T02:21:46Z

BTW the reason macOS builds are failing is they have PCH off, which means you don't get a bunch of headers auto-included. In particular, you need an #include "common/Pcsx2Defs.h" at the top of PS2float.c to get the u32, etc defines.

…s to their respectives place. I don't like the SIMD way of doing it, it can be slower and less practical to use (expensive casting).

GitHubProUser67 · 2024-12-19T16:40:55Z

Now the only code style remaining is the Global VU TMP.

TellowKrinkle · 2024-12-20T06:20:50Z

common/BitUtils.h

+		for (s32 i = 31; i >= 0; i--)
+		{
+			if (((value >> i) & 1) != 0)
+				return i;
+		}
+		return -1;


Suggested change

for (s32 i = 31; i >= 0; i--)

{

if (((value >> i) & 1) != 0)

return i;

}

return -1;

#ifdef _MSC_VER

unsigned long bit;

return _BitScanReverse(&bit, value) ? bit : -1;

#else

return value ? __builtin_clz(value) ^ 31 : -1;

#endif

And you can drop the rest of the changes and make BitScanReverse8 and clz just call this (or pick one and stop using the rest).
As a side note, not handling the value == 0 case will be a bit cheaper, and in my experience in pretty much all cases, the caller has already checked and is only calling this if they know value is nonzero, so the separate check here is just wasted cycles.

Ok, I will do that when possible and push, I admit I lacked experience with the BitScan part.

GitHubProUser67 · 2024-12-27T13:35:36Z

I declare Final Fantasy X fixed on PCSX2!

GT4/Tourist Trophy are an edge case that we can hopefully sort out soon.

…ftfloat on some extra obscure VU ops and fixes a denormals check on the checkDivideByZero method in the FPU. Fixes: - Final Fantasy X (fully playable) - Klonoa 2 Partially Fixes: - Mortal combat Shaloin Monks - Gran Turismo 4 (game patch will be neceassary to skip Licence Test CRC check). - Tourist Trophy (game patch will be neceassary to skip Licence Test CRC check). The stop gap div measure is not yet enough to fully fix GT4/TouristTrophy (they will need different rounding modes per licences). Currently this is bridged on the Div Rounding Mode setting.

AmyRoxwell · 2024-12-28T07:36:43Z

Hopefully This could someday get implemented into the recompiler. Having this many games fixes (And some outright going from broken to playable) is really nice to see! Keep up the great work you doing!

LoStraniero91 · 2024-12-30T13:57:59Z

#5963

Superman Returns
Needs VU1 - MUL/DIV to fix SPS
Note: the SPS is fixed with the "VU Overflow hack", but this causes buildings to fade in and out near the screen edges when flying at high altitude, the soft-float setting completely solves both issues.

#11207

Colin McRae Rally 3
AI on super special stages will desync from its course even on YOLO Mode (EE/VU0/VU1 Interpreters and all soft-floats enabled)

user18081972 · 2024-12-30T19:25:40Z

so this PR does not fix the bugs in Colin McRae Rally 3, correct?

LoStraniero91 · 2024-12-30T19:55:39Z

so this PR does not fix the bugs in Colin McRae Rally 3, correct?

Correct. Even tried with EE Cache and full floats, no dice.

seta-san · 2024-12-30T21:41:44Z

so this PR does not fix the bugs in Colin McRae Rally 3, correct?

Correct. Even tried with EE Cache and full floats, no dice.

either floating point needs to be perfect or it’s a timing issue

vcx33 · 2024-12-30T21:56:13Z

I declare Final Fantasy X fixed on PCSX2!

GT4/Tourist Trophy are an edge case that we can hopefully sort out soon.

For FFX, the boss turning invisible bug doesn't happen when using the interpreter, but it doesn't seem to be affected by soft floats, and seems to be a regression, this will have to be re tested on the recompiler.

…lags + uses built-in clz for Add/Sub. Fixes a TON of games. The flags are not yet in use in the Interpreters, this will ideally be commited next (requires VU code changes). The Div/Sqrt method is unoptimized for now, the team is working on a faster equivalent.

GitHubProUser67 · 2025-01-04T20:25:18Z

And there we have our New-Year present! Fully accurate FPU in PCSX2!

Aside a few things remaining to adjust (flags and newer Div/Sqrt method), we now have a solid basis that replicate the PS2 operations.

The last commit fixes:

[BUG]: FPU rounding issues in Shadow of the Colossus (SCUS-97472) #11528
[BUG]: Tourist Trophy: License Tests are broken and not completeable, player is placed out of bounds #11507
[BUG]: Disney's UP TLB Miss #4760
[BUG]: Need for Speed Undercover not loading save games properly unless they’re created within PCSX2 #9831
Gran Turismo 4 license tests
Mortal Combat Shaolin Monks
Tokyo Xtreme Racer Zero (Accurate Mul/Div for the EE FPU)
Any other games using the NegDiv hack or any other FPU rounding mode.

seta-san · 2025-01-04T20:33:30Z

Jesus. This is amazing. many of the oldest issues are suddenly just fixed. This is truly God's work you and Tellow and everyone else has done here.

LoStraniero91 · 2025-01-04T23:57:01Z

And there we have our New-Year present! Fully accurate FPU in PCSX2!

Aside a few things remaining to adjust (flags and newer Div/Sqrt method), we now have a solid basis that replicate the PS2 operations.

The last commit fixes:

[BUG]: FPU rounding issues in Shadow of the Colossus (SCUS-97472) #11528

[BUG]: Tourist Trophy: License Tests are broken and not completeable, player is placed out of bounds #11507

[BUG]: Disney's UP TLB Miss #4760

[BUG]: Need for Speed Undercover not loading save games properly unless they’re created within PCSX2 #9831

Gran Turismo 4 license tests

Mortal Combat Shaolin Monks

Tokyo Xtreme Racer Zero (Accurate Mul/Div for the EE FPU)

Any other games using the NegDiv hack or any other FPU rounding mode.

Tested Colin McRae Rally 3 with this commit and the AI still goes out of sync and crashes into walls. What I should also add is that the AI keeps accelerating against the wall, then it "reminds" it could reverse as I can see it tries to back out from the wall, but it stays in place and goes nowhere. After almost 3 minutes, the AI "dies", as it probably thought it has finished the race.
Very temperamental little game.

Kaibayugi2002 · 2025-01-05T00:08:47Z

@LoStraniero91
I'm assuming that only testers can download this at the moment. So unless your a tester, we the general public don't have access yet.

Valtekken · 2025-01-05T00:10:55Z

@LoStraniero91 I'm assuming that only testers can download this at the moment. So unless your a tester, we the general public don't have access yet.

You can download it yourself from the artifacts right below the comments.

kamfretoz · 2025-01-05T00:12:50Z

I'm assuming that only testers can download this at the moment. So unless your a tester, we the general public don't have access yet.

You only need a GitHub account to download the CI artifacts.

And you can download it from here

EDIT:
Alternatively, you can go to the "Checks" tab near the top of the page and choose the OS you are using.

NIWDERED-07 · 2025-01-05T00:14:40Z

¡Y ya tenemos nuestro regalo de Año Nuevo! ¡Un FPU totalmente preciso en PCSX2!

Aparte de algunas cosas que quedan por ajustar (banderas y el nuevo método Div/Sqrt), ahora tenemos una base sólida que replica las operaciones de PS2.

La última confirmación corrige:

[ERROR]: Problemas de redondeo de FPU en Shadow of the Colossus (SCUS-97472) #11528

[ERROR]: Trofeo Turístico: Las pruebas de licencia están rotas y no se pueden completar, el jugador queda fuera de los límites #11507

[ERROR]: Disney's UP TLB Miss #4760

[ERROR]: Need for Speed Undercover no carga los juegos guardados correctamente a menos que se creen dentro de PCSX2 #9831

Pruebas de licencia de Gran Turismo 4

Monjes Shaolin de combate mortal

Tokyo Xtreme Racer Zero (multiplicación y división precisas para EE FPU)

¿Algún otro juego que utilice el hack NegDiv o cualquier otro modo de redondeo de FPU?

Wonderful :O, I look forward to them soon being able to repair driver 3, it makes me want to play it again :D

Tokman5 · 2025-01-05T13:19:40Z

Demo replays of Tokyo Xtreme Racer 0 are MUCH closer to the ones running on console than ever!
Unfortunately, they are still not equal (I don't know if this is solely due to FPU accuracy). It affects the behavior of non-player cars, so some replays are still broken. Even so, the last commit is still a huge step.

We use a faster checked method that achieve the same result.

GitHubProUser67 changed the title ~~[Soft-Float] - Initial Intepreter Implementation of Ps2's floating point uint specification~~ [Soft-Float] - Initial Intepreter Implementation of Ps2's floating point unit specification Nov 12, 2024

github-actions bot added Build | Project System Vector Units GUI/Qt First Time Contribution labels Nov 12, 2024

github-actions bot reviewed Nov 12, 2024

View reviewed changes

GitHubProUser67 changed the title ~~[Soft-Float] - Initial Intepreter Implementation of Ps2's floating point unit specification~~ [Soft-Float] - Initial Interpreter Implementation of Ps2's floating point unit specification Nov 12, 2024

GitHubProUser67 force-pushed the soft-float-int branch from 96414bd to de047ea Compare November 12, 2024 20:48

[FPU] - Uses Soft-Float comparison.

b09bfb0

More accurate approach to compare.

[VUops] - Implements accurate SQRT/RSQRT + removal of TriAce hack.

3bd88f7

Implements accurate SQRT options, also removes Tri-Ace hack, which isn't needed anymore on the interpreter.

[Soft-Float] - Improves the PS2Float class by using a raw float to sp…

745e474

…eed-up simple calculations. This greatly improves performance while using Soft-Floats.

GitHubProUser67 force-pushed the soft-float-int branch from dca7e3c to 745e474 Compare December 7, 2024 21:47

[Soft-Float] - Fixes MAC-OS compile error + moves the bitUtils method…

5b94f53

…s to their respectives place. I don't like the SIMD way of doing it, it can be slower and less practical to use (expensive casting).

TellowKrinkle reviewed Dec 20, 2024

View reviewed changes

GitHubProUser67 mentioned this pull request Dec 27, 2024

[BUG]: Smoking to death in Deus Ex: The Conspiracy crashes PCSX2 #12112

Open

GitHubProUser67 force-pushed the soft-float-int branch from d1282ec to 5b94f53 Compare December 27, 2024 18:44

SecularSteve mentioned this pull request Dec 30, 2024

[BUG]: Need for Speed - Hot Pursuit 2 (NTSC): Intro video out of sync/Sound dropouts. Attract mode goes immediatly back to menu. #7877

Open

[Soft-Float] - Fixes Itof method on Clang compiler.

99ee5d5

We use a faster checked method that achieve the same result.

GitHubProUser67 force-pushed the soft-float-int branch from 41f561b to 99ee5d5 Compare January 6, 2025 20:34

[Soft-Float] - Initial Interpreter Implementation of Ps2's floating point unit specification #12001

Are you sure you want to change the base?

[Soft-Float] - Initial Interpreter Implementation of Ps2's floating point unit specification #12001

Conversation

GitHubProUser67 commented Nov 12, 2024 • edited Loading

github-actions bot left a comment

Choose a reason for hiding this comment

Thank you for submitting a contribution to PCSX2

seta-san commented Nov 12, 2024

refractionpcsx2 commented Nov 12, 2024

seta-san commented Nov 12, 2024

MrCK1 commented Nov 12, 2024

AmyRoxwell commented Nov 13, 2024 • edited Loading

Shoegzer commented Nov 13, 2024

AmyRoxwell commented Nov 13, 2024

Blackbird88 commented Nov 13, 2024 • edited Loading

Tokman5 commented Nov 13, 2024 • edited Loading

Goatman13 commented Nov 13, 2024 • edited Loading

GitHubProUser67 commented Nov 13, 2024 • edited Loading

Shoegzer commented Nov 13, 2024

weirdbeardgame commented Nov 13, 2024 • edited Loading

Goatman13 commented Nov 13, 2024

ghost commented Nov 13, 2024

AmyRoxwell commented Nov 13, 2024

GitHubProUser67 commented Nov 13, 2024

Shoegzer commented Nov 13, 2024

Goatman13 commented Nov 24, 2024

GitHubProUser67 commented Dec 7, 2024

TellowKrinkle commented Dec 10, 2024

GitHubProUser67 commented Dec 19, 2024

TellowKrinkle Dec 20, 2024 • edited Loading

Choose a reason for hiding this comment

GitHubProUser67 Dec 20, 2024

Choose a reason for hiding this comment

GitHubProUser67 commented Dec 27, 2024

AmyRoxwell commented Dec 28, 2024 • edited Loading

LoStraniero91 commented Dec 30, 2024 • edited Loading

user18081972 commented Dec 30, 2024

LoStraniero91 commented Dec 30, 2024

seta-san commented Dec 30, 2024

vcx33 commented Dec 30, 2024

GitHubProUser67 commented Jan 4, 2025

seta-san commented Jan 4, 2025

LoStraniero91 commented Jan 4, 2025

Kaibayugi2002 commented Jan 5, 2025 • edited Loading

Valtekken commented Jan 5, 2025

kamfretoz commented Jan 5, 2025 • edited Loading

NIWDERED-07 commented Jan 5, 2025

Tokman5 commented Jan 5, 2025 • edited Loading

GitHubProUser67 commented Nov 12, 2024 •

edited

Loading

AmyRoxwell commented Nov 13, 2024 •

edited

Loading

Blackbird88 commented Nov 13, 2024 •

edited

Loading

Tokman5 commented Nov 13, 2024 •

edited

Loading

Goatman13 commented Nov 13, 2024 •

edited

Loading

GitHubProUser67 commented Nov 13, 2024 •

edited

Loading

weirdbeardgame commented Nov 13, 2024 •

edited

Loading

TellowKrinkle Dec 20, 2024 •

edited

Loading

AmyRoxwell commented Dec 28, 2024 •

edited

Loading

LoStraniero91 commented Dec 30, 2024 •

edited

Loading

Kaibayugi2002 commented Jan 5, 2025 •

edited

Loading

kamfretoz commented Jan 5, 2025 •

edited

Loading

Tokman5 commented Jan 5, 2025 •

edited

Loading