Hi Ti folks,
We are currently developing an application for the c6678 DSP and we need to boot fast. We are using the eight cores which are running the same image file. We would like to have a large heap that is shared between all the cores to store and buffer data. We are using HeapMem to create the heap which is defined in the .cfg file.
We have two problems, where the second one is critical:
1. The largest heap possible to create is 0x1FFFFFFFh large and we found a forum here on e2e that explained that there is no software support for larger heaps. Are we going to see a solution to this? Now we are creating some "dummy heaps" that are linked to be placed as one large section in our memory.
2. The boot time increases with growing heap size. When creating a heap of 1.2 GB we spend 8 seconds just to start the slave cores. I have not really understood what is happening between the "wake-up interrupt" and branch to c_int00. Why does the initiation of the heap take so long time? It seems like all values of the heap are set to 0, by all cores. Is there a way to get around this problem?
Regards,
Niklas