Scarface: The World is Yours is an arcade game (TPP) released in 2006 by Sierra Entertainment. The game was released for PC (Windows), PlayStation, Wii and XBox. A game based on the events of the film The Man with a Scar from 1983. Here, you play as Tony Montana just before the attack of.

TL;DR - if you are not interested in an in-depth overview of what was wrong with the game and how it was fixed,scroll down to Download section for a link anda Patreon announcement.

Today’s subject is Scarface: The World is Yours. This tie-in game to a 1983 movie Scarface wasdeveloped by Radical Entertainment and released in late 2006.

The Micro SMG is a weapon available in Scarface: The World Is Yoursit is unlocked at the first reputation level up and Tony can take it from his weapon locker as soon as he accesses it, it is also a common sight in the hands of gang members early in the game. Arms dealers will sell it for $600, ammo for $150 each clip and it can be moddified with a double clip, a stock and a surpressor. Scarface: The World Is Yours (Video Game 2006) cast and crew credits, including actors, actresses, directors, writers and more.

DEFINITIVE EDITION Scarface The World Is Yours Dec 29 2020 Released Dec 30, 2020 Third Person Shooter DEFINITIVE EDITION is a mod for the game Scarface The World Is Yours. It aims to iprove the graphical quality of the game and its gameplay. HIP-HOP: Afrika Bambaataa: Don't Stop.Planet Rock True Crime: New York City, Thrasher: Skate and Destroy, NBA ShootOut 2004 (Instrumental), NBA 2K15, DJ Hero 2; Paul Oakenfold remix in Dance Dance Revolution Extreme: Afrika Bambaataa: Looking For The Perfect Beat Grand Theft Auto: Vice City.

Naturally, by now you likely know where it’s going – what is wrong this time?As always, PCGamingWiki page providesan answer – at the time of writing this post, you can see the issue listed as one of the key points:

Graphics are corrupted on modern versions of Windows

This doesn’t tell much about how bad is it, so what does it look like?

It is… very bad. While results vary across different PC’s greatly, nearly every modern PC displays thisbroken mess, rendering the game unplayable. Of course, community came up with workarounds, but they sadly have their drawbacks:

dgVoodoo can be used with the game, and it seems to fix the issue. However, dgVoodoo wraps game’s Direct3D 9 to Direct3D 11,which might bump up system requirements considerably. Considering how complex DX9 is, it’s also hard to prove that absolutelyeverything renders flawlessly.
Curiously, this issue also goes away when the game runs through… PIX, a graphics debugger! However, similar issues apply– given PIX is used to collect data, it might introduce additional overhead. Also, having to always run the game throughit is a bit clunky, isn’t it?

While these might suffice for casual play, neither of these workarounds address the root cause of the issue.However, we want to know what exactly went wrong, so those aren’t enough – armed with a debugger anda virtual machine, I can proceed to figuring this issue out.

Step one to figuring out such issues is of course attaching a debugger. This time, since the issue seems to bespecific to D3D, I aided myself with DirectX Wrappersfrom Elisha Riedlinger. Having a minimal d3d9.dll wrapper is excellent for prototyping,since it allows me to instantly and reliably tap into game’s rendering code without any game specific hacking.

That said, this issue could really be anything. That’s where experience helps, so I was able to come up with theoriesby just observing how the bug looks visually:

When in game, moving around makes those shapes move, possibly corresponding player character’s animations.This means that at least part of this garbage is in fact Tony’s model.
Results vary every session, and on some machines it might even look very different – for example, I have seen variations where most of the environment was not corrupted, but people were T-posing.

I theoretized that this issue relates to vertex buffers not updating properly. At this point it’s worthto mention that people thought this issue shows up only on multicore CPU’s. If this is true, then it could meanthat the issue is a classic race condition – which also would explain why it looks different every time.Conceptually, it also makes sense – game could be loading models on multiple threads, setting up D3D resourcesconcurrently and I imagine that failing to do so in a thread safe manner could result in artifacts looking like that.

That theory is trivial to confirm or debunk via a DX9 wrapper. Set up additional code to verifythat all calls to IDirect3DVertexBuffer9::Lock and IDirect3DVertexBuffer9::Unlock are done from the same thread.If they are done from multiple threads, it could be a possible culprit.

I set up verification, and… nothing ¯_(ツ)_/¯ Theory debunked.

Let’s take another look at vertex buffer locks though. I spotted that some (not all) locks are done with aD3DLOCK_DISCARD flag. Looking at MSDN docs, it’s defined as (emphasis added by me):

The application discards all memory within the locked region.For vertex and index buffers, the entire buffer will be discarded.This option is only valid when the resource is created with dynamic usage.

I don’t know about you, but I see possible room for error here. It is possible to lock only a part of the buffer,so I can imagine somebody not realizing that locking a part of the buffer with discard flag would throw it awayentirely.

However, game always locks entire buffers:

Analyzing the code further though, I was not fully convinced that after discarding game always fills the entire buffer.So what if we assume the flag is added there wrongly and remove it?

Scarface The World Is Yours Mac

It works! It doesn’t seem to be a fluke either – the game was consistently fixed for me and several other peoplewho tested.

A cautious reader might stop right here and ask a few questions:

Clearly, game worked as-is at some point, so how do you know this is not just another workaround andit’ll break again?
Why does using PIX and dgVoodoo fix it?

With a DX9 wrapper and knowledge on when D3DLOCK_DISCARD is supposed to be used, I could come up with proofthat this flag was used wrongly. In principle, you’d use this flag when locking a buffer if you did not careabout the buffer’s past contents and intended to fill it with new data. If the game used this flag, but also expectedthe buffer to retain its old contents, then the flag was misused.

To test whether the game really cares about previous buffer contents, I deliberately filled it with garbageif the game locked it with D3DLOCK_DISCARD:

Much to my relief, the result was more or less what I expected:

Graphics were once again broken, which proves that game locks buffers with a D3DLOCK_DISCARD flag andexpects the contents not to be thrown away! For me that’s a satisfactory enough proof of API misuse,and a proof that removing this flag is not a workaround, but a correct fix.

What about PIX and dgVoodoo? The latter is a wrapper, so I can imagine it not emulating the behaviour ofthe discard flag. PIX on the other hand is a graphics debugger, and in order to be able to capture the frameit likely preserves past and current buffer’s contents. While I can’t prove it for certain, I wouldn’t be surprisedif PIX just ignored this flag overall.

Are we done? Well…

I could technically finish here, as the most important issue has been fixed, and the game is playable.However, something was still off…

I have a i7-6700K and GTX 1070, mind you.

My PC is nowhere close to “bad”, yet I was unable to maintain stable 60 FPS – in the area presented on a screenshot,I in fact got consistent 40-45 FPS. To say it’s “terrible” would be an understatement.

However, looking up process affinity in Task Manager reveals something… interesting:

It is like this on every game lau