Hi,
We are conducting verification to confirm the effect size of IIR accelerator and FIR accelerator.
I am evaluating on the ADSP-21569
For verification, we use "BiquadNCascade" and "FIRNChan" respectively.
According to the help for each module, up to 32 channels are processed by the accelerator, and the remaining channels are processed by the core.
We record the DSP load displayed on the AWE server while changing the number of channels from 1 to 64, and calculate the "rate of change in processing load" expressed by the ratio of the number of channels of the filter module to the processing load of the DSP. By focusing on this, we verified the operation of the accelerator. According to the help for each module, the rate of change in processing load should change around 32 channels.
As a result, BiquadNCascade's "rate of change in processing load" did not change depending on the number of channels (numStage=1).
Also, for FIRNChan, there was a change in the "rate of change in processing load" after channel 48 (numTaps=64).
Is accelerator enabled in BiquadNCascade? If so, please let me know the effective verification method.
In addition, why does FIRNChan's "rate of change in processing load" not change after 32 channels? If so, please let me know the effective verification method.
Best regards
9:53am
Hello Makotosudou,
Thanks for your comment. I will email you on the side to discuss this, as I will have to ask about the specific AWE Core library you are using. My first thought is that perhaps your SHARC+ AWE Core library may be out of date.
Thanks,
Kevin