Jump to content

Talk:CUDA

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

We need to include how the architecture works

[edit]

Warps, hardware threads, grid, mesh, topologies; SIMT. All are relevant. --Daviddwd (talk) 16:33, 7 October 2018 (UTC)[reply]

There should be a basic code sample

[edit]

To have some basic understanding of CUDA programming experience there should be some source code example.

There are some source code examples in https://en.wikipedia.org/wiki/CUDA#Example but not for basic CUDA C++ usage.

--Ronan Keryell (talk) 19:32, 24 February 2021 (UTC)[reply]

CUDA Toolkit vs Visual Studio matrix

[edit]

I've added a table that shows which versions of Visual Studio are supported by each CUDA Toolkit SDK. This is super-useful so one doesn't keep trying to install different VS versions for a given card. Sorry I didn't have time to put into wikipedia's table format yet. Would be nice to create this matrix for Linux and Mac compilers also.
-- Ramakarl (talk) 21:50, 28 February 2024 (UTC)[reply]

GPUOpen

[edit]

The article mentions that GPUOpen facilitates running CUDA on AMD GPUs. However, the entry seems to be several years old, as only compatibility up to CUDA 4.0 is mentioned. As can be seen on AMD's GPUOpen Githubs, the GPUOpen libraries are updated every few days at least. So, how far has GPUOpen come by now when it comes to supporting newer versions of CUDA? --2003:EF:170D:4588:892A:CA15:5B84:A6C0 (talk) 08:38, 11 August 2022 (UTC)[reply]