The greatest challenges lie in accommodating arbitrary control flow among threads within a cooperative thread array. NVIDIA GPUs are SIMD multiprocessors, but they include a thread activity stack that enables serialization of threads when they reach diverging branches. Without hardware support, this kind of thing becomes difficult on SIMD processors which is why Ocelot doesn't include support for SSE yet. It is also one of the obstacles for supporting AMD/ATI IL at the moment, though solutions are in order.
Translation from PTX to LLVM to multicore x86 does not necessarily throw away information concerning the PTX thread hierarchy initially. The first step is to express a PTX kernel using LLVM instructions and intrinsic function calls. This phase is [theoretically] invertible and no information concerning correctness or parallelism is lost.
To get to multicore from here, a second phase of transformations insert loops around blocks of code within the kernel to implement fine-grain multithreading. This is the part that isn't necessarily invertible or easy to translate back to GPU architectures and is what is referenced in the note you are citing.
Disclosure: I'm one of the core contributors to the Ocelot project.
Havok wasn't obligated to do this. It is a kind (and perhaps savvy) gesture. They weren't obligated to do it, but let's be honest: they were somewhat forced to. NVIDIA bought PhysX not too long ago and announced they were implementing it with CUDA so GPUs could provide physics acceleration.
The NVIDIA PhysX binary-only SDK has been available for a while now.
"If PC gamers are left with these options to save them from consoles, do they even have a chance?" Save PCs from consoles? As others have said, no console currently on the market can come close to the performance exhibited by even the ATI (or AMD, however you want to spell it) Crossfire solution, let alone the GeForce 9800 GX2. On a console, you're stuck with 1280x1080 at best with lower levels of detail in the geometry and textures. A $180.00 GeForce 9600 GT could provide equivalent or superior graphics at comparable framerates to current consoles. The article concludes the bottlenecks for the GF9800GX2 are 512 MB of RAM *per GPU*, each with a 256-bit memory bus. For some perspective, the XBox 360 has 512 MB *total* shared by the GPU and the CPU. The PS3 has 256 MB for the GPU and 256 MB for the CPU.
Also, both NVIDIA and ATI/AMD developed the graphics technology that went into today's consoles. It's not like console technology will somehow overtake what's available for PCs; it's the same technology, only the product cycle for PCs is a lot shorter.
Consoles simply cannot defeat the PC as a gaming platform on the basis of somehow having better hardware. Sure, Wii games are fun. Gameplay is always important, blah blah. That's no reason to assume consoles are, have, or will be crushing PCs anytime soon. As long as compilers are available, small timers will be making games for PCs. And successful small timers occasionally become bigger timers. But what about profit? We all played WoW on our PCs.
Here is why you shouldn't care too much about the results listed in the article. The GF9800 GX2 isn't just for 3D graphics. The reason why NVIDIA rushed the GF9800 GX2 to market now is to support the brave new world of high-performance computing they are envisioning. NVIDIA recently announced that a CUDA implementation of PhysX would be released; you'll probably want two GPUs for that. Additionally, CUDA 2.0 is due real soon now, and this will certainly have enhanced support for multi-GPU application development. To buy one of these just for gaming right now is, well, not economical. To buy a system capable of this degree of performance in this form factor (*eight* GPUs fit on one Extended ATX Intel Skulltrail) intended for research, scientific, or industrial computing is, well, a steal. Hats off to gamers for making this kind of technology affordable.
And now, the sensationalist closing: could this be the year of the Slashdot article summary that concludes without baseless rhetorical questions?
This was hardly notable. The Reaction Research Society routinely surpasses this using amateur rocket motors. Their 475 lb 14,000 lb thrust solid fuel rocket launched a dart reaching an altitude of 280,000 feet. Their project site outlines the launch pretty well. To my knowledge, no amateur group has attempted or reached this before or since.
I would imagine the resulting increase in entertainment spending may help offset the reduced work output by the crazy Star Wars fans who don't seem to notice that "Attack of the Clones" is also the name of a kid's cartoon.
Who wouldn't be absolutely thrilled to know their music doesn't just help humans fall in love? It sure makes me want to chase women around rooms biting at their backs and lower regions.
Slashdot Mirror
The greatest challenges lie in accommodating arbitrary control flow among threads within a cooperative thread array. NVIDIA GPUs are SIMD multiprocessors, but they include a thread activity stack that enables serialization of threads when they reach diverging branches. Without hardware support, this kind of thing becomes difficult on SIMD processors which is why Ocelot doesn't include support for SSE yet. It is also one of the obstacles for supporting AMD/ATI IL at the moment, though solutions are in order.
Translation from PTX to LLVM to multicore x86 does not necessarily throw away information concerning the PTX thread hierarchy initially. The first step is to express a PTX kernel using LLVM instructions and intrinsic function calls. This phase is [theoretically] invertible and no information concerning correctness or parallelism is lost.
To get to multicore from here, a second phase of transformations insert loops around blocks of code within the kernel to implement fine-grain multithreading. This is the part that isn't necessarily invertible or easy to translate back to GPU architectures and is what is referenced in the note you are citing.
Disclosure: I'm one of the core contributors to the Ocelot project.
The NVIDIA PhysX binary-only SDK has been available for a while now.
http://developer.nvidia.com/object/physx.htm
Kindness? Neigh: competition.
Also, both NVIDIA and ATI/AMD developed the graphics technology that went into today's consoles. It's not like console technology will somehow overtake what's available for PCs; it's the same technology, only the product cycle for PCs is a lot shorter.
Consoles simply cannot defeat the PC as a gaming platform on the basis of somehow having better hardware. Sure, Wii games are fun. Gameplay is always important, blah blah. That's no reason to assume consoles are, have, or will be crushing PCs anytime soon. As long as compilers are available, small timers will be making games for PCs. And successful small timers occasionally become bigger timers. But what about profit? We all played WoW on our PCs.
Here is why you shouldn't care too much about the results listed in the article. The GF9800 GX2 isn't just for 3D graphics. The reason why NVIDIA rushed the GF9800 GX2 to market now is to support the brave new world of high-performance computing they are envisioning. NVIDIA recently announced that a CUDA implementation of PhysX would be released; you'll probably want two GPUs for that. Additionally, CUDA 2.0 is due real soon now, and this will certainly have enhanced support for multi-GPU application development. To buy one of these just for gaming right now is, well, not economical. To buy a system capable of this degree of performance in this form factor (*eight* GPUs fit on one Extended ATX Intel Skulltrail) intended for research, scientific, or industrial computing is, well, a steal. Hats off to gamers for making this kind of technology affordable.
And now, the sensationalist closing: could this be the year of the Slashdot article summary that concludes without baseless rhetorical questions?
This was hardly notable. The Reaction Research Society routinely surpasses this using amateur rocket motors. Their 475 lb 14,000 lb thrust solid fuel rocket launched a dart reaching an altitude of 280,000 feet. Their project site outlines the launch pretty well. To my knowledge, no amateur group has attempted or reached this before or since.
I would imagine the resulting increase in entertainment spending may help offset the reduced work output by the crazy Star Wars fans who don't seem to notice that "Attack of the Clones" is also the name of a kid's cartoon.
Who wouldn't be absolutely thrilled to know their music doesn't just help humans fall in love? It sure makes me want to chase women around rooms biting at their backs and lower regions.