I think I should clarify the DSP situation a bit.
Above I said that the 410 processor doesn’t have an “Audio DSP”, this statement is correct (I didn’t say there is no DSP in the 410). The audio interface hardware (Low Power Audio Sub System, or LPASS) has all of the registers to access the external codecs (I2S, Digital Mics, ADC/DAC in the PMIC, etc.), The LPASS registers in the 410 can only be read or written by the main A53 processors. Hence every audio sample must be transferred to/from the LPASS by the main processor.
There is a DSP in the 410. BUT this DSP does not have access to the LPASS registers, hence if you use this DSP to perform tasks like MP3 decompress you need to transfer the compressed stream to the DSP, and then transfer the decompressed stream from the DSP to the LPASS. All this shuffling around of the data may cause a significant delay, and it may be faster to just run the algos on the A53 and save the transfers.
Access to the 410 DSP is not currently available but is in the works. When it is ready you will be able to write DSP code, load it into the DSP and run it. The Hexagon DSP compiler is available, but the infrastructure to load a compiled binary into the DSP and transfer data to/from the DSP is not in place yet.
The other chips sets (600 and 820) do have “Audio DSPs” and the LPASS registers can only be accessed by the DSP, The main processors do not have access to the LPASS registers, in these chip sets all audio must pass through the Audio DSP.
Full Disclosure: I am an employee of Qualcomm Canada, any opinions I may have expressed in this or any other post may not reflect the opinions of my employer.