Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
12 years ago

Read Write to DRAM (main memory)

I am just switching from another fpga to altera.

Could you please tell me is this the write way to access main memory.

main_memory = (char *) DDR2_SDRAM_BASEADDR; // base address of main memory can be changed in system.h file

main_memory1 = (char *) DDR2_SDRAM_BASEADDR+1024; // base address of main memory can be changed in system.h file

for ( x = 0 ; x < 262122 ; x = x +1)

*(main_memory+x) = *(main_memory1+x);

regards

UCERD

3 Replies

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    This isn't the right way if the "main memory" is also used for your software. The beginning of the memory is generally used for the exception vectors, and if you overwrite those you will crash the system.

    It is a lot safer to allocate a block of memory to use, either by using malloc() to allocate at run time, or by using a static or global table to allocate at compile time.
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    --- Quote Start ---

    This isn't the right way if the "main memory" is also used for your software. The beginning of the memory is generally used for the exception vectors, and if you overwrite those you will crash the system.

    It is a lot safer to allocate a block of memory to use, either by using malloc() to allocate at run time, or by using a static or global table to allocate at compile time.

    --- Quote End ---

    Yes you are right but the reason for using directly baseaddress is, I have two DRAM controller on my FPGA board.

    The memory controller (We designed) can access read/write operation in parallel, that is why for comparison we are designing a baseline system (Nios system based) that can access both memories in parallel.
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    Well if you are sure that this memory section is never used by other parts of the software, then you can do it this way. If the memory buffer is shared with some hardware (a DMA for example) then remember to flush the cache before the hardware accesses the memory, or use uncached access through alt_remap_uncached()