DSP_dotprod

Implementation Notes

- Bank Conflicts: No bank conflicts occur if the input arrays x[ ] and y[ ] are

 

offset by 4 half-words (8 bytes).

 

- Interruptibility: The code is fully interruptible.

 

- The code is unrolled 4 times to enable full memory and multiplier

 

bandwidth to be utilized.

 

- Interrupts are masked by branch delay slots only.

 

- Prolog collapsing has been performed to reduce codesize.

Benchmarks

Cycles

nx / 4 + 14

 

Codesize

64 bytes

C64x+ DSPLIB Reference

4-61

Page 89
Image 89
Texas Instruments TMS320C64X manual Cycles Nx / 4 + Codesize Bytes