Dot Product of two vectors (of unsigned 8-bit data)

Walter Snafu

I'm looking for a speed-optimized software routine that produces the dot-product of two arrays of unsigned 8-bit data. (For the C6748) I don't know where to begin looking.

Thanks for your help.

over 15 years ago

0 desouza over 15 years ago

TI__Guru**** 171309 points

Walter,

Although it takes 16-bit parameters, the DSP_dotprod() function in C64x+ DSPLIB is a good start:

http://focus.ti.com/docs/toolsw/folders/print/sprc265.html

To further explore the capabilities of parallel multiplication in C64x+/C674x architecture, check the function IMG_mad_8x8() of the C64x+ IMGLIB code. Its source code implementation contains a loop that performs 8x8 minimum absolute difference calculation that can be easily adapted to a dot product (check the API reference guide).

http://focus.ti.com/docs/toolsw/folders/print/sprc094.html

At last, check the C674x DSPLIB if you are looking for some floating point implementations.

http://processors.wiki.ti.com/index.php/C674x_DSPLIB

Hope this helps,

Rafael

Code Composer Studio™︎

Code Composer Studio forum

Dot Product of two vectors (of unsigned 8-bit data)