LLVM -> SPIR-V


Larry Gritz
 

Thought this thread was promising. If this ends up happening, it could pave a much easier path for us to eventually make OSL work on non-Cuda GPU platforms.

https://lists.llvm.org/pipermail/llvm-dev/2021-March/148905.html



--
Larry Gritz
lg@...


Roman Zulak
 

Count me in...gonna dust off that OSL clang front-end.

There's a clang-based c++ interpreter at CERN that's pretty neat, someone focused on CUDA support is giving an update tomorrow.
Maybe I can nerd out on OSL fragments running at the LHC drawing ray-traced or Vulkan tachyons (well lit of course) soon.

>> Our next meeting will be on 04 March 2021 at 17:00 Geneva (CH) Time/8:00PST.
>> We are happy to have Simeon Ehrig presenting his work on C++ and CUDA -- "Cling’s CUDA Backend: Interactive GPU development with CUDA C++".



On Tue, Mar 2, 2021 at 7:31 PM Larry Gritz <lg@...> wrote:
Thought this thread was promising. If this ends up happening, it could pave a much easier path for us to eventually make OSL work on non-Cuda GPU platforms.

https://lists.llvm.org/pipermail/llvm-dev/2021-March/148905.html



--
Larry Gritz
lg@...










boberfly@...
 

Hey Larry,

I directed Sean Baxter to the mailing list awhile back about his project Circle-lang as another way to approach this, emitting SPIR-V directly from C++ code here:
https://lists.aswf.io/g/osl-dev/message/4399?p=,,,20,0,0,0::relevance,,sean,20,2,0,77939956

But I guess if Intel are donating an official SPIR-V backend that supports Vulkan-style and not just OpenCL, then that's great too!