Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
7 years ago

Vector_Add example with 4 Compute Units

Hi,

I am trying to understand the use of "num_compute_units(N)" attribute by using "Vector_Add" example . I have not a physical board so I am using the emulator. I attach an image of the OpenCL code with the above attribute. I have several doubts regards its functioning because of the kernel execution time without the attribute is better than the code with "num_compute_units(4)" attribute (as in the attached image). I expect that the code with four CUs reduces the execution time. Do I have to do some changes in the OpenCL code?

Thanks for your help

Marco Montini

1 Reply

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    7 years ago

    Run time in the emulator is *not* a representative of run time on the hardware; in fact, run time in the emulator *does not mean anything whatsoever*. Altera's emulator is not timing-accurate and hence, should not be used for anything other than *functional verification*.

    Furthermore, I already explained here as to why there is no point in using multiple compute units for the vector add example:

    https://www.alteraforum.com/forum/showthread.php?t=57921&p=235688#post235688