OpenCL: Dot Product Sample Code

Accelerated Computing CUDA CUDA Programming and Performance
1

I have run the Dot Product program from Nvidia’s sample codes. I also write a Dot Product program to compute the similar amount of data (five millions) for a single CPU using Visual Studio without OpenCL. The performance of the later is ten time faster.

Is there any suggestion for me to make the performance speedup in OpenCL?

Thank you,
Thuy

2

I’d like to add more information about my system:

It’s an Intel Core2 Duo.

64-bit Windows Vista

Nvidia graphics card GeForce 9200M GS

Nvidia driver version 8.16.11.9089

Please let me know what I should do to improve OpenCL’s performance? Your helps are really appreciated.

Thank you,

3

To my understanding, a GeForce 9200M GS is only half of a 9400M, so it has only 1 MP (total 8 “cores”). That means it probably has no more computation power than your Core 2 Duo. With the extra overhead of calling a OpenCL kernel, copying memory, etc. it’s normal that you don’t see much performance out of it.