Hello Everyone,
I'm new to OpenCL, and I'm trying to figure out a good laptop that is supported with the OpenCL SDK. I've found no laptops from the System Supported List. Specifically, I was trying to figure out if the following Acer system works for OpenCL development. I plan to replace the Windows 8 with a compatible Linux version, but it's not clear if the Processor & Graphics chip will support the OpenCL dev.
thanks in advance for your replies.
Hi this will support OpenCL 1.2
Please check the below link
http://www.amd.com/us/products/notebook/graphics/oem-solutions/Pages/8700-8800.aspx#2
Thanks for your reply. Since the GPU supports OpenCL, does that mean that
the AMD SDK for OpenCL will also be supported? Secondly, does this laptop
support Linux (Redhat/CentOS)? I couldn't find the drivers for Linux on the
Acer website.
-Nagesh
Yes SDK is supported. You download drivers for Linux directly from AMD site.
Thanks for the reply.
Can you also please let me know if I would be able to run the SDK through a
VM using VirtualBox on the same AMD machine? Is it required to install Linux
using dual boot, or will the VM work?
Thanks,
Nagesh
In VM only CPU device will work.
GPU virtualization is not yet enabled. so, please do not use any VM software...if you want to use the GPU.
Okay. Thanks. I guessed so, and have installed OpenSuse Linux on the laptop.
I was able to download and install the drivers as well as the SDK. Now, when
I compile the examples & run, it all works fine. However, I'm unable to say
if it is running on the GPU or the CPU. Are there any tools for this?
How many GPU devices are on the Radeon 8750M? What should the devices flag
be set to when running the samples?
Thanks again. Your responses are very useful, and helpful!
-Nagesh
when you run examples from SDK they will run on GPU by default. in SDK there is clinfo program which print out information about OpenCL devices. you can use it to check if is GPU correctly recgnized as GPU device.
Thanks.
Specifically, I want to know this when I write a kernel, how can I tell if
it is executing on the GPU? Do you have some tools for this?
-Nagesh
If you execute kernel on GPU depends only from that what command queue you use and what device is associated with it. Look at context and command queue creation.
Hello,
Thanks for your replies.
While I understand this part, my question was different. Is there a real
time tool (like the task manager for the CPU) that shows how much the GPU is
loaded? Something I can run & then look to see how the device that I'm
running the kernels is getting loaded as I run it.
Thanks,
Nagesh
In catalyst control center there should be overdrive tab which should have GPU load indicator.
oh Linux. then you must run amdconfig --odgc and amdconfig --odgt in console.
Nagesh,
You can try CodeXL. I believe it is available for both Windows and Linux.
It is like a GUI based Profiler and Debugger for your OpenCl code.
It will even give a proper timeline of your various OpenCL API calls, which command Q and device it executed etc...
It might be very interesting to you,
Thanks,
Best Regards,
Bruhaspati
Nagesh,
To execute a Kernel, You first need to create a Context.
A Context is created over a set of devices (usually a device).
So, When you list the devices in a platform, you can always look at the device-type (CL_DEVICE_GPU) and then create a context over it.
Then, you can compile and execute kernels on that device.
Thats all.
These are the most basic steps in OpenCL initialization.....
I thnk you should start reading the spec...
Best Regards,
Bruhaspati