hi all
consider the following pseudo code snippet for a GPP side application:
Uint16* DataBufPtr = 0;
// there are various types of structs / classes defined:
typedef struct myObj1{
int a, short b, int c;
} MyObj1;
typedef struct myObj2{
char x, float y[1000];
} MyObj2;
Class A;
// allocate a large buffer (which shall be used "heap-like")
POOL_alloc(0, DataBufPtr, 1MB);
// inside that buffer, place different kind of objects
MyObj1* obj1 = DataBufPtr[1000];
MyObj2* obj1 = DataBufPtr[2000];
A* a = DataBufPtr[3000];
// fill the objects' content
obj1->a = 1; obj1->b = 2; obj1->c = 3;
obj2->x = 'x'; obj2->y = { ... }
Question:
How would the DSP application be able to access such objects? Can you (resp. have to?) do a POOL_writeback(0, obj1, sizeof(MyObj1))? Or does POOL_writeback() only work on a complete POOL buffer? Or can you also write back only parts of an allocated buffer (maybe the parts must be DSP cache aligned)?
The same question applies to POOL_translateAddr: Does it work on any address inside a pool buffer or only on the start address of a buffer?
My general idea:
How would it be possible to make something like a "shared memory heap" where i can place objects (as described above) from the GPP side and then read / write these object from the DSP side too?
Any advices are very welcome! Thx.
