Olivier Giroux's CppCon 2019 talk recounts the roughly ten-year effort to bring std::atomic and the C++ memory model to CUDA GPUs, through ten stories from that work.
Why Hacker News points people here: Cited as a reference when discussing NVIDIA's C++ memory model, libcu++, and GPU C++ resources.
“NVidia is designing their GPGPUs to run C++ code”
Cited as evidence NVidia designs GPUs to run C++.
Full comment
Right now, NVidia is designing their GPGPUs to run C++ code. linkthis video With Intel just releasing oneAPI (aka Data Parallel C++) last month, and Khronos pushing SYCL as alternative to CUDA C++. It took 30 years for C++ to reach this stage, and in some domains (like embedded) it is still fighting for relevance. That is something to keep in mind whenever to advocate Rust as replacement for Domain XYZ.
Cites itdavvidSep 2020
Related CppCon talk on std::atomic in CUDA.
Full comment
Here's a somewhat related talk from CppCon '19: "The One-Decade Task: Putting std::atomic in CUDA" this video
Cites itpjmlpApr 2021
Listed among CUDA C++ resources.
Full comment
CppCon 2019: Olivier Giroux “The One-Decade Task: Putting std::atomic in CUDA.” this video Future of ISO and CUDA C++ link Thrust, the CUDA C++ template library link Optix, single-ray shader programming model using C++, and ray Tracing acceleration using RT Cores. link Phew, so many wrappers yet to list.
Cites itpjmlpAug 2021
Listed among GPU C++ resources.
Full comment
Sure, "CppCon 2016: “Bringing Clang and C++ to GPUs: An Open-Source, CUDA-Compatible GPU C++ Compiler" link "CppCon 2019: Olivier Giroux “The One-Decade Task: Putting std::atomic in CUDA.”" this video "Octane render. The Future of GPU Rendering: Real-Time Raytracing" (with CUDA) link NVIDIA HPC SDK link
Cites itmoonchildOct 2022
Cited to show Nvidia has a defined memory model.
Full comment
> I really wish that everybody would finally admit that the "fixed function pipeline" is dead There is a great deal of space between 'fixed function pipeline' (which no one is using) and 'all compute all the time' (which pretty much no one is using for real-time graphics either). > GPUs don't yet have a well-defined memory model Nvidia has a memory model. (I have heard--from an nvidia employee--that amd plays a bit more fast and loose.) this video > GPUs don't have the equivalent of an MMU--so they can't be partitioned or protected. etc. You do not need an MMU to partition or to protect. (Not to say that GPU security isn't a shitshow.) > The problem is that switching to a "compute" model and making that work well would commodify the GPU--and that's the last thing that AMD and NVIDIA want. link look at where nvidia's revenue is coming from, and tell me again that they're not already there.
