TMS320VC5509A外部DRAM操作问题

这个64K的访问限制跟C55x Compiler有关。请看下面的FAQ。

In C55x, why does a long pointer fail when incremented over a 64k boundary?

Problem:
When a long pointer is incremented and crosses over a 64k boundary (ie, an address which is a multiple of 0x0001_0000) why isn't the upper 7 bits of the address modified? It appears the address just wraps around the same 64k block of memory.

Solution:

Even when the large memory, which supports 23-bit pointers, is used the
compiler does not calculate the offset from the base address correctly
when crossing a 64k page boundary.

For example, if you have a pointer which starts at 0x2f000 and increment
the pointer to pass address 0x2ffff, instead of becoming 0x30000 the address
will wrap around and become 0x20000. The same is true for arrays, which
should not span a 64k page boundary.

A workaround is to manually check for a boundary change and keep
track of what the next page should be.

#pragma DATA_SECTION(gMp3, "ERAM")
int gMp3[200000];
int *p;
unsigned long int nextPage;
unsigned long i;
nextPage = (((unsigned long int)gMp3 & 0x7F0000) >> 16) + 1;
for ( i = 0; i < 200000; i++){
*p++ = 0x0000;
/* IMPORTANT: need this to manage page boundries manually. */
if (((unsigned long int)p & 0xFFFF) == 0)
{ p = (int *)(nextPage++ << 16);
}
}