SigmaStudio+ is used in dual-core mode

Hi,

Could you please share your schematic for analyzing this issue.

Thanks

Hi Chiru & Sakthivel,

Any updates for this issue?

Our schematics were so complex that one core could not run well enough. We urgently needed to use 21593's dual core. Please give us useful help, thank you.

Hi Chiru & Sakthivel,

Any updates for this issue?

Our schematics were so complex that one core could not run well enough. We urgently needed to use 21593's dual core. Please give us useful help, thank you.

Hi Jonny,
We will check and get back to you, since your schematic having 48 Input/40 Output channels we need to make some modifications to the schematic to test on evaluation boards.

whether your simple schematic worked with 48 Input/40 Output channels configuration? 
Is my understanding correct that you are facing issue only when using complex schematic with more modules?
Any idea what your schematic MIPS is?
Since your schematic block size is set to 16 there could be a possibility for MIPS overshoot when running with 48 Input/40 Output channels.     

Hi Sakthivel,

1. The schematic of attachment can work normally

2, I checked the MIPS and it didn't overshoot 

3, Our project has 48 input  & 40 output channel. In my test on EV board only 16 input /16 output channel is actually used. The attachment has the related configuration of Sport port, and the unused channel is not connected to the data input/output port through SRU

4. Because there are too many input and output channels, if the block size is set to 64, it will consume more memory, so I try to set it to 32 and 16

5, The audio process is as follows:

Core1 Sport In (48CH) -> Core1 SS+ -> Out to Core2 (40CH) -> Core2 In(40CH) - >  Core2 SS+ -> Out to Core1(40CH) -> Core1 Sport Out(40CH)

Attached is my test schematic and part of the source code for your reference.

6354.Test.zip

VID20240521161431.zip

In your video the first schematic is working fine and the second schematic having some target exception when enabled some audio stream from Adobe Audition tool. Is my understating correct?
Could you please explain what audio stream is enabled in Adobe Audition tool and how that stream linked with target application?

Are you using the SigmaStudio+ Schematic generated source file in 21593 target application? If yes, please regenerate the source file for the schematic which you are facing issue.

Also, could you please try increasing the stack memory allocated for target application in app.ldf file?

The first schematic is simple enough to work even when playing multiple channels; The second schematic diagram is more complicated, playing one channel can work properly, playing multiple channels can not work properly;
I used normal audio files, attached is my test audio file; It is connected to the sound card via USB and then transferred to the DSP via A2B;
I used SigmaStudio+ to compile the schematics and download them to 21593. I also tried to set both heap and stack to 20KBytes, but the program still crash

Test Audio.zip

Few more question, did you integrate the A2B stack into the target application or using in PC mode A2B schematic?
Is this issue reproduced with ADSP-21593 SOM Eval board?

The ADSP-21593 receives data from corresponding SPORT through DAI pins we don't think there will be a problem for single or multiple channels data because all SPORT's will run once enabled even though no audio data. We will try to simulate this issue and get back to you. 

1, I used a USB to A2B sound card, the following A2B project and sound card A2B configuration is exactly the same, for reference

2. " The ADSP-21593 receives data from corresponding SPORT through DAI pins we don't think there will be a problem for single or multiple channels data because all SPORT's will run once enabled even though no audio data. "

This is also what I think is very strange. Even if there is no audio data, ADSP21593 still does the same processing, why will it crash when there are too many audio channels input? 

Also note: it is normal to work with single-core DSP only. So I believe the audio input/output, A2B configuration, SPORT configuration should be no problem.  SS+ framework support for dual core is not perfect, has it been extensively verified?

adi_a2b.zip

The Lib integration example what we provided is for demonstration of communication and target library integration for SigmaStudio+ schematic processing. You can use demo example application as standard dual core processing application and see whether same issue is there.

Few more question, did you measure the schematic MIPS? can you reduce the MIPS by reducing the number of filter stages? To confirm whether there is any run time MIPS overshoot. Filters and compressors will take more MIPS based on the input signals, so I am doubting there may be MIPS overshoot, since the processing block size also less.

I checked the MIPS and it didn't overshoot. It takes up about 60% of the CPU power.

Your demonstration can't download schematics online. I've shared my project with you. What experiment did you do?  Can you provide some meaningful response ??