TDA4VM: Memory allocation issue for the program

Seetharama Raju

Part Number: TDA4VM

Hello team,

I am working on an application, which requires a custom kernel. The kernel has four input and output tensors. The kernel requires 8 header files, where 8 LUT arrays are defined of the size 1280x944 of unsigned int data.

While building the application the following error log is appeared. The target core is DSP2. Can you please guide me how to clear the memory allocation error.

Fullscreen 7128.Ti_error_log.txt Download

"/home/raju/TDA4x_8.02/ti-processor-sdk-rtos-j721e-evm-08_02_00_05/vision_apps/platform/j721e/rtos/c66x_1/j721e_linker_freertos.cmd", line 81: error: 
   program will not fit into available memory.  placement with alignment fails
   for section ".const" size 0x2510876 .  Available memory ranges:
   DDR_C66x_1   size: 0xdffc00     unused: 0xdffc00     max hole: 0xdffc00  
error: errors encountered during linking;
   "/home/raju/TDA4x_8.02/ti-processor-sdk-rtos-j721e-evm-08_02_00_05/vision_ap
   ps/out/J7/C66/FREERTOS/release/vx_app_rtos_linux_c6x_1.out" not built
make[1]: [concerto/finale.mak:213: /home/raju/TDA4x_8.02/ti-processor-sdk-rtos-j721e-evm-08_02_00_05/vision_apps/out/J7/C66/FREERTOS/release/vx_app_rtos_linux_c6x_1.out] Error 1 (ignored)
Linking /home/raju/TDA4x_8.02/ti-processor-sdk-rtos-j721e-evm-08_02_00_05/vision_apps/out/J7/C66/FREERTOS/release/vx_app_rtos_linux_c6x_2.out
"/home/raju/TDA4x_8.02/ti-processor-sdk-rtos-j721e-evm-08_02_00_05/vision_apps/platform/j721e/rtos/c66x_2/j721e_linker_freertos.cmd", line 81: error: 
   program will not fit into available memory.  placement with alignment fails
   for section ".const" size 0x2510876 .  Available memory ranges:
   DDR_C66x_2   size: 0xdffc00     unused: 0xdffc00     max hole: 0xdffc00  
error: errors encountered during linking;
   "/home/raju/TDA4x_8.02/ti-processor-sdk-rtos-j721e-evm-08_02_00_05/vision_ap
   ps/out/J7/C66/FREERTOS/release/vx_app_rtos_linux_c6x_2.out" not built
make[1]: [concerto/finale.mak:213: /home/raju/TDA4x_8.02/ti-processor-sdk-rtos-j721e-evm-08_02_00_05/vision_apps/out/J7/C66/FREERTOS/release/vx_app_rtos_linux_c6x_2.out] Error 1 (ignored)
Linking /home/raju/TDA4x_8.02/ti-processor-sdk-rtos-j721e-evm-08_02_00_05/vision_apps/out/J7/C71/FREERTOS/release/vx_app_rtos_linux_c7x_1.out

Thanks,

Seetharama Raju.

over 2 years ago

0 Nikhil Dasan over 2 years ago

TI__Guru* 88401 points

Hi,

Have you modified the memory map in the SDK or is it the default vision_apps memory map?

Regards,

Nikhil

0 Seetharama Raju over 2 years ago in reply to Nikhil Dasan

Prodigy 170 points

Hi Nikhil,

It is default vision_apps memory map not the modified.

Thanks,

Seetharama Raju.

0 Nikhil Dasan over 2 years ago in reply to Seetharama Raju

TI__Guru* 88401 points

Hi,

The code region is 14 MB for C66x. But it seems that introducing 8 LUTs and header files increased this memory.

You would either have to do one of the below

1. Optimize your codes to fit in 14 MB.

2. Increase the code region. This would require modification of the memory map. You could refer the below for the same.

8.9. Understanding and updating SDK memory map for J721E — Processor SDK RTOS J721E

3. Move the LUT to the local heap region of C66x.

Regards,

Nikhil

0 Seetharama Raju over 2 years ago in reply to Nikhil Dasan

Prodigy 170 points

Hi,

I would try the possible options mentioned. Meanwhile, I have moved the kernel to A72 core and working further.There is one query regarding the same application.

The pipeline of our multi camera application is capture->LDC->scaler->preproc->TIDL->custom kernel

TIDL segmentation model gives segmented mask, which has four tensors corresponding to the four camera views. Till tidl all modules are replicated but the custom kernel node is not. In the kernel target.c file, I have added a print statement.

I have built the application and ran. No build and run time errors. But the application is run till the print statement appearing for four times and stopped.

How can I make the app be running continuously? Please guide me on this.

Thanks,

Seetharama Raju.

0 Seetharama Raju over 2 years ago in reply to Seetharama Raju

Prodigy 170 points

Hi Nikhil,

The issue mentioned in the above reply is solved. I have increased the buffer depth of TIDL from 4 to 8. Increase in the number of buffers before the custom node made the entire pipeline processed continuously. I want to keep this thread active and raise the questions here for further help, if needed.

Thanks,

Seetharama Raju.

0 Nikhil Dasan over 2 years ago in reply to Seetharama Raju

TI__Guru* 88401 points

Hi,

Seetharama Raju said:
The issue mentioned in the above reply is solved

That is great.

Seetharama Raju said:
I want to keep this thread active and raise the questions here for further help, if needed.

You could respond on this thread to reopen the same. However, no response for 30 days would lock the thread.

Regards,

Nikhil

Processors

Processors forum

TDA4VM: Memory allocation issue for the program