[beagleboard] Why optimizing openCV for DSP while TI has an image processing lib?

GoogleGroups
1

I've just found out that there is a ratther complete image processing
library for TI c64x DSPs in TI website (http://focus.ti.com/docs/
toolsw/folders/print/sprc264.html). At the moment I am wondering why
someone might want to optimize openCV functions for DSP (as in one of
BeagleBoard contest projects)?! Aren't the functions in TI library
optimum enough?!

Thanks in advance,
--Nastaran

2

You get 3 guesses for what c6accel does :slight_smile:

3

As I see there are several DSP libraries on TI website. To be honest, by you mentioning c6Accel I feel just more confused.
Now I should ask what the difference between c6Accel and c6Run is! I guess c6Accel and c6run do the same job, but c6Accel itself contains some predefined DSP functions including image and video processing functions. I don’t know more.

Does this c6Accel cover all existing DSP libraries?

–Nastaran

4

As I see there are several DSP libraries on TI website. To be honest, by you mentioning c6Accel I feel just more confused.
Now I should ask what the difference between c6Accel and c6Run is! I guess c6Accel and c6run do the same job, but c6Accel itself contains some predefined DSP functions including image and video processing functions. I don’t know more.

C6Accel exposes a bunch of DSP functions to the ARM, ie. utilize. C6Run lets you create and run code on the DSP easily, ie. create.

Does this c6Accel cover all existing DSP libraries?

I doubt it. You should probably compare the functions in the image library or whatever other DSP library you might be looking at to the functions available in C6Accel.

5

Because you may not want to limit your application to TI platforms, or
port your existing openCV-based app to use the TI libs. Consider that
you have an existing solution that uses openCV to do face detection
and you want to run this on the Beagle.

Yaman