Part Number: 66AK2H14
Tool/software: Code Composer Studio
Hello,
Currently I am working on an OpenMP-based application on EVM 66AK2H board. I need to use the DSP cores as accelerators (target devices) using the model explained in TI OpenMP Accelerator Model.
I am able to compile and run the application with clacc, following the methodology of OpenMP-Acc’s examples included in Processor SDK ( /usr/share/ti/examples/openmpacc). But, I would like to debug the application in order to fix some runtime bugs.
I have used CCS, as indicated here, in order to compile and debug the programs that are executed in the ARM cores. I have used gdbserver for debugging. I have enabled OpenMP adding the -fopenmp flag in the “Cross G++ Compiler” and “Cross G++ Linker” option. But, the problem that I have found is that the code in the “#pragma omp target”, which should be executed in the DSP cores, is actually executed only in the ARM that executes this pragma. Thus, the DSP cores do not execute code.
Could you give me some ideas about how to use a CCS debugging-project for an application that uses OpenMP on my board? This application should send code to the cores DSP using the “#pragma omp target” device construct. For example, how should I use the vecadd example included in Processor SDK using CCS?
Best Regards,
Javier
