How AMD Is Taking Standard C/C++ Code To Run Directly On GPUs

[schmidtbag]

Kinda makes me wonder why this wasn't done nearly 20 years ago? Even though there are many good reasons for the CPU to do the work, having the option for the GPU to do "raw" C code would still make sense and perhaps would have seen CUDA with a lot less adoption.

[Nille]

Probably because you need a certain feature set to run a program independent from all else on the gpu.

[carewolf]

GPUs are turing complete these days, so "slow" rather than "not possible". I think at the moment it's mainly "this code can be run on the GPU" rather than automatic offloading - which, while it would be nice for the lazy programmer, would probably not be possible on a standard GPU architecture. That's mainly because the GPU and CPU do not share the same memory, so offloading a loop from CPU to GPU has a high fixed cost and I really don't see a good way of being able to automatically determine when it's worth doing.

Using the GPU integrated into the CPU that use the same memory?

[Guest]

Bad idea,

why do you make a negative assessment of what many are waiting for with great joy? For many, this is a step into the new. You can stay, no one is calling you to go ahead. we are tired of CPUs, with their inability to run simultaneous code execution. Now, when Google announced the possibility of releasing quantum processors, in 2024, thanks to people like you, we still cannot use the GPU to run user functions on them.

[sophisticles]

This is a great idea, here's what I would like to see happen:

They create a simple, Python style language, that has easy to learn and read syntax and is easily expandable via libraries.

Then add a function decorator that allows a programmer to specify which parts of the code should be run where, for instance:

@_CPU
def function(ONE)

@_GPU
def function(TWO)


@_CPU_128_SIMD
def function(Three)

I guess what I am saying is write a compiler for Python that allows a programmer to specify on what hardware he wants that code portion to be executed.

[Limit]

I guess what I am saying is write a compiler for Python that allows a programmer to specify on what hardware he wants that code portion to be executed.

Your description sounds like what Mojo is trying to achieve.

[ojsl1]

Kinda makes me wonder why this wasn't done nearly 20 years ago? Even though there are many good reasons for the CPU to do the work, having the option for the GPU to do "raw" C code would still make sense and perhaps would have seen CUDA with a lot less adoption.

Twenty years ago AMD wasnt there pushing for open-source, that's why. Stereotypical corpo thinks keeping their hardware inaccessible makes it look more magical than it actually is.

[habilain]

Using the GPU integrated into the CPU that use the same memory?

Probably not possible? ROCm doesn't officially support APUs, so I'm pretty sure ROCm doesn't support zero copy. Or if it does, that's weird.

Even if it is possible, I'm not sure there'd be that much benefit. APUs aren't that powerful - and it would likely still need special code to get decent performance. For example, double precision floating points tend to be quite slow on a GPU.

[marlock]

is AMD finally thinkering and trying to use their iGPUs instead of CPU wide vector extensions like AVX-whatever-size-is-biggest-now?

[lem79]

Probably not possible? ROCm doesn't officially support APUs, so I'm pretty sure ROCm doesn't support zero copy. Or if it does, that's weird.

Even if it is possible, I'm not sure there'd be that much benefit. APUs aren't that powerful - and it would likely still need special code to get decent performance. For example, double precision floating points tend to be quite slow on a GPU.

Maybe not weird if there has been work already to support Strix Halo (we could hope that's the case).