hjabird · HackerTrans

hace 2 añosdiscuss

There are some great replies to my comment - my original comment was too reductive. However, I still think that entrenching CUDA as the de-facto language for heterogeneous computing is a mistake. We need an open ecosystem for AI and HPC, where vendors compete on producing the best hardware.

hjabird

hace 2 añosdiscuss

True. This is the big advantage of an open standard instead jumping from one vendors walled garden to another.

hjabird

hace 2 añosdiscuss

The problem with effectively supporting CUDA is that encourages CUDA adoption all the more strongly. Meanwhile, AMD will always be playing catch-up, forever having to patch issues, work around Nvidia/AMD differences, and accept the performance penalty that comes from having code optimised for another vendor's hardware. AMD needs to encourage developers to use their own ecosystem or an open standard.

hjabird

hace 3 añosdiscuss

Intel's modern compilers (icx, icpx) are clang-based. There is an open-source version [1], and the closed-source version is built atop of this with extra closed-source special sauce.

AOCC and ROCm are also based on LLVM/clang.

[1] https://github.com/intel/llvm

hjabird

hace 3 añosdiscuss

What kind of performance is achieveable with some of the features that Vcc enables (true function calls, function pointers, goto), and what are some of the limitations?

On GPU, function calls are much more expensive than on CPU. Usually it seems to be worth inlining as much as possible. Implementing a stack for recursion on GPUs is also likely to have performance implications. The whole point of using a GPU is to obtain good performance, but I can see the argument for having these features in order to port CPU code to GPU code incrementally.

For function pointers, how does that work? Multiple different implementations of the function are needed to support different devices and the host, which limits what a single pointer can do.

EDIT: To answer my second question, neither function nor data pointers are portable between host and device since Vulkan doesn't support unified addressing.

hjabird

hace 3 añosdiscuss

I wouldn't expect them to drop CUDA support, even if SYCL is a viable alternative: * The CUDA backend is mature, featureful, and significant effort has been invested into optimising it on Nvidia hardware. One does not simply throw away a performant, well validated HPC code! * Nvidia GPUs dominate the GPGPU market - unlike AMD's. * The SYCL backend is still is very new in comparison (they even state in the docs to pay extra attention to validation), and doesn't have Nvidia-specific optimisations yet. Why prioritise reimplementing what already exists?

hjabird

hace 3 añosdiscuss

A challenge with this is that all current AMD GPU support in SYCL compilers (DPC++ w/ Codeplay's oneAPI for AMD GPUs, and AdaptiveCpp), is built atop of ROCm / HIP.

If AMD were to move away from ROCm, they would have to adopt some other API for SYCL to use as a backend. Potentially, this could be the level zero spec from UXL foundation, but that is not the same as SYCL.

However, no changing between APIs etc. will help with troublesome drivers, or issues with OS compatibility, or documentation. Those issues are orthogonal.

hjabird

hace 3 añosdiscuss

> The performance of any application on SYCL is currently quite poor.

SYCL can get pretty much equivalent performance in Kernels to eg. CUDA. Try looking at SYCL performance papers on Arxiv. Eg. see [1].

That isn't to say that SYCL code is optimised on every platform without tweaking - you do still need to put effort into target specific optimizations to get the best performance, like you would in the CUDA or HIP.

> Why drop ROCm (used on the world's largest supercomputer?)

Some of the world's largest super-computers / HPC applications do use SYCL for AMD! The application I'm most aware of for this is GROMACS. As to why? - because having 3 version of the same code using different programming APIs is a big maintenance burden.

[1] https://arxiv.org/pdf/2309.09609.pdf

hjabird

hace 3 añosdiscuss

SYCL (closely related to oneAPI) isn't single-vendor-controlled. It's a Khronos open standard. If you take a look at the spec, you'll see contributions from various universities, Qualcomm, Huawei, Argonne, Altera, and AMD (Xilinx I think). Intel just adopted it (and bought Codeplay, the original contributor).

oneAPI is a set of SYCL related standards. Originally that was Intel, but now that's owned by the UXL foundation, which is part of the Linux foundation.

hjabird

hace 3 añosdiscuss

> Why would you want OpenCL?

OpenCL is stricter with the results of floating point operations, and makes different assumptions with respect to memory aliasing. Whether or not this is important the AI domain I don't know.

> Why would you want a unified address space?

A unified address space doesn't always imply that the memory can be accessed from anywhere (although that might also be supported with some memory allocation mechanisms), and you still may have to copy between host and device memory. But it makes it much easier to have pointers in your GPU kernels, instead of having to deal with objects like OpenCL buffers.

hjabird

hace 3 añosdiscuss

SYCL isn't based on OpenCL.

SYCL (SYCL-2020 spec) supports multiple backends, including Nvidia's CUDA, AMD's HIP, OpenCL, Intel's Level-zero, and also running on the host CPU. This can either be done with Intel's DPC++ w/ Codeplay's plugins, or using AdaptiveCpp (aka. hipSYCL, aka openSYCL). OpenCL is just another backend.

It is also a very long way from OpenCL C++. The code is a single C++ file, and you don't need to write any special kernel language. The vast majority of SYCL is just C++, so -if you avoid a couple of features- you can use SYCL in library-only form without even any special compiler! This is possible for instance with AdaptiveCpp.

hjabird

hace 3 añosdiscuss

SYCL is gaining traction, especially in the HPC community since it can target AMD, Nvidia and Intel hardware with one codebase. A fun fact is the GROMACS (a major application for molecular dynamics, and big consumer of HPC time) recommends SYCL for running on AMD hardware!

comments