Generic GPU Kernels in Julia • Hacker News Board

Re: Generic GPU Kernels in Julia

#11

Post by maleadt » Fri, Aug 25, 2017, 6:00 AM UTC

arcanus wrote: post #2

Generic GPU kernels... for just NVIDIA devices.

As the author of the underlying framework: the dependency on CUDA is unfortunate indeed, but it was the only viable option at the time. OpenCL tooling was (is) way more fragmented, there's no unified compilation target or back-end, differences between vendors, etc. However, much of this work has dealt with figuring out how to compile Julia for accelerators while integrating with the existing compiler, all of which is transferable to future back-ends. Once the SPIR-V back-end is better integrated with upstream LLVM, we will try to target that toolkit instead.

Re: Generic GPU Kernels in Julia

#12

Post by dotdi » Fri, Aug 25, 2017, 7:18 AM UTC

I cannot wait to throw away the C++11/CUDA mess I am dealing with right now.

However, this doesn't really seem ready for prime time, i.e. undocumented stuff that is "in flux". Going to keep an eye on this.

Re: Generic GPU Kernels in Julia

#13

Post by nicwilson » Fri, Aug 25, 2017, 7:54 AM UTC

maleadt wrote: post #11

arcanus wrote: post #2

Generic GPU kernels... for just NVIDIA devices.

As the author of the underlying framework: the dependency on CUDA is unfortunate indeed, but it was the only viable option at the time. OpenCL tooling was (is) way more fragmented, there's no unified compilation target or back-end, differences between vendors, etc. However, much of this work has dealt with figuring out how to compile Julia for accelerators while integrating with the existing compiler, all of which is…

I maintain https://github.com/thewilsonator/llvm/tree/compute and https://github.com/thewilsonator/llvm-target-spirv which has (unused) tablegen definitions for SPIR-V, for LDC and DCompute. I'm _hoping_ to get some consensus on backend integration into LLVM trunk at IWOCL in May. More users of the backend are useful for convincing the LLVM folks of the usefulness of having a single backend in trunk.

Re: Generic GPU Kernels in Julia

#14

Post by autopoesis » Fri, Aug 25, 2017, 1:32 PM UTC

dotdi wrote: post #12

I cannot wait to throw away the C++11/CUDA mess I am dealing with right now. However, this doesn't really seem ready for prime time, i.e. undocumented stuff that is "in flux". Going to keep an eye on this.

Have you had a look at SYCL? I found a nice intro at [1]. Hopefully at some point soon there's an implementation of it for nVidia GPUs, too; right now I've only come across support for Intel and AMD OpenCL using Codeplay's compiler [2]. But it seems quite promising!

[1] https://blog.tartanllama.xyz/sycl/ [2] https://www.codeplay.com/products/computesuite/computecpp

Re: Generic GPU Kernels in Julia

#15

Post by dotdi » Sat, Aug 26, 2017, 9:11 PM UTC

autopoesis wrote: post #14

dotdi wrote: post #12

I cannot wait to throw away the C++11/CUDA mess I am dealing with right now. However, this doesn't really seem ready for prime time, i.e. undocumented stuff that is "in flux". Going to keep an eye on this.

Have you had a look at SYCL? I found a nice intro at [1]. Hopefully at some point soon there's an implementation of it for nVidia GPUs, too; right now I've only come across support for Intel and AMD OpenCL using Codeplay's compiler [2]. But it seems quite promising! [1] https://blog.tartanllama.xyz/sycl/ [2] https://www.codeplay.com/products/computesuite/computecpp

I haven't seen SYCL until now, but it looks quite promising. Hopefully Codeplay will keep their interest and investment into that piece of software.

Unfortunately, I don't have the time to wait for NVidia support. I have a window of opportunity to rewrite some core things right now, but that window will soon close.