Is OpenCL slowly dying?

Started by
12 comments, last by JoeJ 7 years, 6 months ago

Maybe it was just my feeling, as I haven't seen pretty much anyone working with it in last year or so


There are official Khronos teleconferences of the OpenCL group just about every week and an upcoming face-to-face meeting next month, and various other meetings going on all the time.

OpenCL is certainly being updated. OpenCL 2.1 was a surprisingly major release (despite the minor version bump) and came out in 2015. OpenCL 2.2 came out earlier this very year and moved C++ kernel support into the core spec, among other things.

Compute just isn't really changing that much compared to the earlier exciting days of GPGPU research. Most of the updates on any of the frameworks these days is on tooling improvements, e.g. OpenCL gaining SPIR-V and C++ support.

Sean Middleditch – Game Systems Engineer – Join my team!

Advertisement

Thanks.

I see, as it pretty much covers the GPU architecture - which isn't really evolving that fast (I know there were quite good hardware changes like going down to 14nm in the last generation, but from the functionality point of view, there aren't any major steps). So pretty much they are just improving the actual language and support.

I will definitely google some info about OpenCL and C++ support in it (which kind of interests me), I didn't even realized they moved that far.

My current blog on programming, linux and stuff - http://gameprogrammerdiary.blogspot.com

Kind of random, but I came across this OpenCL example of flocking today... hope to try it out soon!

http://www.gameaipro.com/GameAIPro/GameAIPro_Chapter45_Introduction_to_GPGPU_for_AI.pdf

The more i get forward optimizing my stuff, the more changes my initial experience of Vulkan Compute being faster than OCL.

Actually everything i have is faster in OpenCL (shaders from the same source file - preprocessor adapts the rest)

So no - OpenCL for games does not deserve to die yet :)

This topic is closed to new replies.

Advertisement