Geforce gts360M OpenCL Question

Hi !

I want to develop an OpenCL application ( matrix multiplication) by using Visual Studio and Nvidia CUDA Toolkit that i’ve already installed on my laptop. My laptop has a Geforce GTS360M. I don’t find this graphic card in the list of NVIDIA graphic cards compatible with OpenCL, but Visual Studio tells me that my graphic card (GTS360M) is a valid OpenCL device. In fact when i try to run the application it seems to work. Could i have some problem or malfunctioning ? I want also to use a profiler then to measure the performances. It’s ok although GTS360M is not in the list of graphic card compatible with OpenCL ?

The GTS 360M with an appropriate driver should support OpenCL 1.1 functionality.

I’m not sure which list you are referring to.

I’m reffering to this list :
I can’t find GTS360M, but my matrix multiplication program seems to work

Ok OpenCL application that i developed run on my GTS360M but now i have an other issue.
When I try to use Visual Profiler i have this error :

I’m using CUDA Toolkit 7.5 on Windows 10 and I’ve aldready installed latest driver for my graphic card.
Somebody can help me?

The only drivers that will allow your GTS360M to work are incompatible with anything from CUDA toolkit 7.5.

If you intend to use the 360M, you should uninstall CUDA 7.5 and install CUDA 6.5 or an earlier toolkit.

I’m also not sure what the behavior of the visual profiler is with an OpenCL application. Maybe it works, I don’t know. But the issue in your picture is due to the driver incompatibility with CUDA 7.5, just as indicated. Earlier toolkits may also have incompatibility problems with windows 10.

I installed CUDA toolkit 6.5. Now using Visual Profiler i haven’t any errors when it runs my OpenCL application to generate timeline , but at the end of execution nothing happens. I have no timeline and no results. Why?

Well I never said that the profilers support OpenCL. I don’t know if they do or they don’t. Maybe they don’t. Also, windows 10 is not an officially supported environment for CUDA 6.5, so it’s possible there may be an issue there.