CUDA vs OpenCL

This cluster centers on discussions comparing CUDA's dominance in GPU programming to alternatives like OpenCL, SYCL, Vulkan compute, and efforts for cross-platform GPU compute solutions that enable running standard C/C++ code on GPUs.

➡️ Stable 0.9x Hardware
4,241
Comments
19
Years Active
5
Top Authors
#7127
Topic ID

Activity Over Time

2008
12
2009
51
2010
39
2011
91
2012
90
2013
141
2014
139
2015
156
2016
245
2017
256
2018
206
2019
234
2020
394
2021
315
2022
211
2023
465
2024
594
2025
584
2026
18

Keywords

PTX CPU SPIR IMHO HIP GPU MIC CUDA AMD theincredibleholk.org opencl cuda gpu gpus compute amd nvidia hardware code rust

Sample Comments

kelsey98765431 Dec 14, 2024 View on HN

Running normal c directly on gpu has been the dream for a long time. this looks excellent

amelius Dec 21, 2022 View on HN

Shouldn't we have an API that can speak to both CUDA and opencl? Or is opencl sufficiently capable?

tgb Feb 4, 2017 View on HN

You're not aware of OpenCL, GLSL, or CUDA then!

idle_zealot Sep 11, 2017 View on HN

Something like https://en.wikipedia.org/wiki/OpenCL ?

mogoh Feb 12, 2024 View on HN

Why is CUDA so prevalent oppose to its alternatives?

jvlake Jun 13, 2024 View on HN

Cool story. How supported is OpenCL compared to CUDA again?

DrNosferatu Aug 7, 2023 View on HN

No mention of OpenCL / VK compute?

amelius Dec 11, 2024 View on HN

Do they have the equivalent of CUDA, and what is it called?

Fuck CUDA, OpenMP supports GPUs

orbifold Dec 22, 2017 View on HN

No there isn't, Google developed a CUDA frontend for clang/llvm which targets PTX (nvidias virtual machine instruction set), in principle someone could come along and implement a different backend for this frontend targeting for example AMDs gpu instruction architecture or something different. The other ingredient missing is a reimplementation of CUDAs runtime libraries.