The game is a Far Cry (pun intended) from DirectX 10. The demo was also on a Windows machine. Let's see the same demo on Linux.
The original poster is absolutely correct.
C++ written for KDE is going to run on Windows now. Well, having worked at Adobe when I was just out of college, I can say that they have VERY bright people there.
Bright enough to get C++ code written for Linux to run on Windows? in a VM? Which itself would be in a browser?
I have to think that VMWare or Parallels would have done that already if it was doable. Or, if it is doable, and just hard, VMWare and Parallels will do it better than Adobe in any case. Be assured, with the value inherent in such a product they WOULD release their own versions.
I can hear the Flash VM talking now:
Scenario 1:
Oh C++ code . . . you want me to run some Intel vectorized math routines? No problem . . . those Microsoft templates you got there are REALLY cool . . . uhh . . . well let me just get this Objective C library here on this shiny new MacBook Pro and . . . Oh Wait? . . . Well . . . Maybe I can try this GCC library and . . . SLASH!!!
<Leopard kills Flash VM process, and drags carcass home to feed cubs.>
Moral of the story: Predatory cats always target the slow ones.
Scenario 2:
Ahh C++ code . . . I get it now . . . you want to use this really cool GeForce 9xxx series card to do instancing. That is AWESOME!!! Let me get started on that for you . . . uh . . . ahh . . . OK so I'm an OpenGL context and you want me to make a DirectX call . . . Oh OK no problem . . . ahh . . . . umm . . . OK so I'm on Linux, can I ask a quick question? Your developer didn't supply you with a software DirectX 10 implementation did he?
<C++ code makes sad realization that both developer AND Flash VM are idiots>
Moral of the story: Even intelligent things, C++, can be put in situations where it looks really stupid.
Scenario 3:
Oh mapped memory? SURE, I'll just go over here to this pool of unused hard drive space and . . . THWOCK!!! THACK!!! WOCK!!! WHACK!!!
<Flash VM realizes that VISTA is actually a Dominatrix, and he is now being beaten about the head, neck and chest with the interrupt controller he asked for.>
Moral of the story: Something like unused hard drive space, or cross compilation, may sound REALLY good . . . but make sure you know who has a claim on it . . . because she might be a REAL bitch. Or worse . . . Microsoft.
Scenario 4:
Oh . . . multicore . . . umm . . . why don't I go ahead and kill myself now so you can get to the blue screen faster.
<Single gunshot is heard from the other room, system hangs without getting to the bluescreen, because Flash VM had problems even managing to commit suicide. The bullet went through its brain, but its vital organs were untouched.>
Moral of the story: Actually test claims before you start making plans, a lot of the potential technologies you want to use, or potential spouses you will meet, will make claims of intelligence, but there may not be too much between the ears.
I don't think you fully understand what Adobe is doing -- they are compiling down to AVM2 bytecode using LLVM. And DirectX is abstracted out of Actionscript, so you gain the benefits of DirectX when it's available, but the AVM will resort to other means when it's not.
That Quake II demo is a completely self-contained flash animation, and will run anywhere Flash runs (including linux).
Edit: It sounds like some people think Adobe is promising to run ANY C++/Java/Whatever application in Flash. They're not. Some C applications however, like Quake II, can be coerced to run in Flash's single-threaded platform-agnostic environment. The developer who gave that presentation at Max described the work he had to do to emulate threading in order for Quake II to run. It wasn't easy, but I'm sure it was far easier than porting the whole codebase.
There is no SUITABLE replacement for DirectX instancing under OpenGL that would be available on other platforms. To say that you will get the benefits of DirectX where it is not available is to be disingenuous. Microsoft goes to great lengths to make sure that, even with a VM doing the work for you, game developers will find it difficult to
"gain the benefits of DirectX when it's available, but the AVM will resort to other means when it's not" - as you put it.
Over the years 100's of millions of dollars have been spent to ensure that you CAN'T do that. If you talk to Parallels or VMWare they will tell you that they can get DirectX 8 to work at times, Direct X 9 seldomly, and then they will have a good belly laugh while pointing at you when you ask them about virtualizing Direct X 10. This is the way things are for now, until the requisite extensions have been developed for OpenGL.
This is not just me spouting off at the mouth. The entire gaming industry has grudgingly accepted this for quite some time. Steve Jobs has been looking for ways around it forever. If Adobe has people who can do this, they won't have them for long, because Steve is sending over a limo I guarantee it.
And what about proprietary templates? Adobe is going to translate all of those too.
Multicore programming? Gonna translate between the Intel multicore stuff and . . . say . . . the Cell processor stuff?
Taking it further, we all know that CUDA code will work beautifully if the machine you're on happens to have an NVidia card, but what if you have an ATI card?
Don't even get me started on some of these templates for mapped memory management . . .
RESPONSE TO PARENT EDIT:
Thanx for clarifying the assertions you were making.
I assume you mean it's pointless because C++ sacrifices ease-of-coding for performance, but when C++ is compiled to bytecode it won't be appreciably faster than other languages.
While it would be odd for someone to write new C++ code solely to target bytecode, I think that a C++ bytecode compiler would still be useful in situations where, for example, you have an existing C++ codebase that you would like to reuse in a web client.
No, I mean it's pointless because I think there are very few use cases for it. You're pretty much limited to libraries which do not rely on any OS-specific calls or which only do very generic OS-calls like file I/O.
FWIW, I'm the author of GCC-CIL, which provides a backend to GCC to emit .NET bytecode, and I still don't see the point. I mean, what are you going to recompile for flash or .NET? BLAS and LINPACK? :P At least with .NET you have access to libc and other native code libraries via P/Invoke, but you don't even get that from flash...
Ah, I wasn't sure what you meant. I agree that there aren't too many use cases. Personally, I think some of the existing computational geometry libraries could be useful, but I am thinking of a niche set of applications. More useful, perhaps, might be existing libraries to read/write file formats. But I'd be surprised if there was anyone looking to port their huge C++ application wholesale to Flash.
