Part Number: TMDSEVM6678
Hello,
I am facing problems while booting OpenMP based Image Processing Demo shipped as part of MCSDK for EVM6678 platform.
- Kindly note that everything works perfectly when I follow the steps mentioned in BiosMulticoreSDK_2.1_UserGuide (Page 137-138, Booting from NOR). In this example, the bin file (c:\ti\mcsdk_2_01_01_04\demos\image_processing\utils\mad\evmc6678l\images\mcip-plbp-c6678-le.bin) that is pre-generated and shipped as part of the MCSDK is written to NOR flash and finally, DSK is booted from NOR flash. This example works like a charm for me. However, note that the above-mentioned bin file was generated from IPC based Image Processing Demo.
- Within the same folder (C:\ti\mcsdk_2_01_01_04\demos\image_processing), there is an "openmp" folder, which provides an implementation of the same demo based on OpenMP. OUT file of this demo is shipped as part of the MCSDK distribution. I simply use the batch script located at (C:\ti\mcsdk_2_01_01_04\demos\image_processing\utils\mad\evmc6678l\build_mad_image_prelink_bypass_omp.bat) to generate bin file. The bin file is generated in the "images" folder in the same directory.
- Next, I start CCS 7.3, launch debug configuration, connect to Core 0, initialize global defaults from Scripts menu of CCS and finally load the BIN file at address 0x9E000000. When the bin file is successfully loaded to DDR, I set the PC to 0x9E001040 and run core 0. I successfully see the desired output of the program at hyper-terminal. This implies everything seems fine as long as the bin file is concerned.
- Next, I burn this bin file to NOR flash using the same steps I followed in (1). When NOR flash is successfully written, I reboot the DSK. Before rebooting the DSK, I make sure the switches are in the correct configuration as required for booting from NOR. The hyper-terminal shows that IBL loaded successfully, and 'Booting from NOR', but nothing happens next.
- When I connect Core 0 through CCS, I see PC (program counter) is stuck at 0x9E008384. This happens every time no matter which version of MCSDK I use. I tried mcsdk_2_01_01_04, mcsdk_2_01_02_05 and mcsdk_2_01_02_06 with same results.
- I also compared the contents of DDR (after I connect with Core0) and the bin file I generated in step 2 above and there are significant differences.
I will really appreciate if you guys could point me to some relevant help. BTW, there is only one relevant thread that I could find on e2e. Incidentally, this gentleman was facing an almost same problem and his Core0 was also stuck at the same address 0x9E008384. Unfortunately, the details of his dealing with this problem are not clear in the thread. I downloaded and tried the zip file (7230.mad-debug.zip) that apparently solved the problem for this gentleman. However, for me, the problem persisted! i.e. Core0 still halts at 0x9E008384.
Best Wishes,