How to use GPU(Cuda/OpenCL) for mathematical operations

How to use GPU(Cuda/OpenCL) for mathematical operations - Printable Version

+- Sinisterly (https://sinister.ly)
+-- Forum: Coding (https://sinister.ly/Forum-Coding)
+--- Forum: Visual Basic & .NET Framework (https://sinister.ly/Forum-Visual-Basic-NET-Framework)
+--- Thread: How to use GPU(Cuda/OpenCL) for mathematical operations (/Thread-How-to-use-GPU-Cuda-OpenCL-for-mathematical-operations)

How to use GPU(Cuda/OpenCL) for mathematical operations - noXxio - 07-04-2013

Hey,

so today I'll try to show you how to calculate mathematical operations on your GPU using Cuda(on Nvidia), or OpenCL(on Ati/Intel). We're using "Cudafy.Net" to achive this. I have made a simple Example on which you can see how it works.

Download

Screenshot:

Spoiler:

Code explanation:

Our "whole" Code:

PHP Code:
using Cudafy;
using Cudafy.Host;
using Cudafy.Translator; 

PHP Code:
private void btn_gpu_en_Click(object sender, EventArgs e)
        {
            CudafyModes.Target = eGPUType.OpenCL;
            CudafyModes.DeviceId = 0;
            CudafyTranslator.Language = eLanguage.OpenCL;
            try
            {
                GPGPU gpu = CudafyHost.GetDevice(CudafyModes.Target, CudafyModes.DeviceId);
                CudafyModule km = CudafyTranslator.Cudafy();
                gpu.LoadModule(km);
                Stopwatch sw = new Stopwatch();
                int c;
                int[] dev_c = gpu.Allocate<int>();
                sw.Start();
                gpu.Launch().encrypt(Convert.ToInt32(numericUpDown1.Value), dev_c);
                gpu.CopyFromDevice(dev_c, out c);
                sw.Stop();
                txt_gpu.Text = c.ToString();
                lbl_gpu_status.Text = sw.ElapsedMilliseconds.ToString() + " msec";
            }
            catch { }
        } 

and:

PHP Code:
[Cudafy]
        public static void encrypt(int max, int[] c)
        {
            int a = 8919259;
            int b = 8915159;
            int i = 0;
            int y = 0;
            int i_ = max;
            while (i != 99999999)
            {
                while (i_ != 0)
                {
                    y = ((a * b) / 8) * 555;
                    i_--;
                }
                i++;
            }
            c[0] = y;
        } 

So the only thing I'll explain is:

PHP Code:
private void btn_gpu_en_Click(object sender, EventArgs e)
        {
            CudafyModes.Target = eGPUType.OpenCL;
            CudafyModes.DeviceId = 0;
            CudafyTranslator.Language = eLanguage.OpenCL;
            try
            {
                GPGPU gpu = CudafyHost.GetDevice(CudafyModes.Target, CudafyModes.DeviceId);
                CudafyModule km = CudafyTranslator.Cudafy();
                gpu.LoadModule(km);
                Stopwatch sw = new Stopwatch();
                int c;
                int[] dev_c = gpu.Allocate<int>();
                sw.Start();
                gpu.Launch().encrypt(Convert.ToInt32(numericUpDown1.Value), dev_c);
                gpu.CopyFromDevice(dev_c, out c);
                sw.Stop();
                txt_gpu.Text = c.ToString();
                lbl_gpu_status.Text = sw.ElapsedMilliseconds.ToString() + " msec";
            }
            catch { }
        } 

The Rest should be selfexplaining.

PHP Code:
CudafyModes.Target = eGPUType.OpenCL;
CudafyModes.DeviceId = 0;
CudafyTranslator.Language = eLanguage.OpenCL; 

With this Code you Define the Target to OpenCL, not to CUDA or emulator, if you are just using a Nvidia Card change both OpenCL to CUDA.
You map the DeviceId 0, the first device.
And define the Language your C# Code will be translated to OpenCL, if you are using Cuda, it should be of course translated to Cuda Wink

PHP Code:
GPGPU gpu = CudafyHost.GetDevice(CudafyModes.Target, CudafyModes.DeviceId);
CudafyModule km = CudafyTranslator.Cudafy();
gpu.LoadModule(km); 

1. Defines a new Gpu, the mapped target 0.
2. Defines a new Module, in this case the Translator.
3. The in Step 1. defined Gpu Loads the in Step 2. defined Module.

PHP Code:
gpu.Launch().encrypt(Convert.ToInt32(numericUpDown1.Value), dev_c);
gpu.CopyFromDevice(dev_c, out c); 

1. Launches our Void.
2. Extracts the Output of our Void to "c".

The rest should be self explaining, if you have any Questions feel free to ask Smile

~noXxio

RE: How to use GPU(Cuda/OpenCL) for mathematical operations - Shebang - 07-04-2013

This is pretty interesting, I'd like to see a practical implementation of this in newer releases.

RE: How to use GPU(Cuda/OpenCL) for mathematical operations - Wet - 07-14-2013

C# is a dear language for me. And this is a great tutorial. I knew about this, and i know it's useful, so what should i say? GJ and keep HQ work up!

RE: How to use GPU(Cuda/OpenCL) for mathematical operations - noXxio - 07-23-2013

Thank you very much for replying Smile

)

RE: How to use GPU(Cuda/OpenCL) for mathematical operations - Platinum - 07-23-2013

Great, I will come back to this when I do some more stuff with this language.