Compiling program in C

[carlesureta]

Hi, so I am trying again to write a very simple code, that accesses memory, and I thought of accessing a vector of 128 integers. The thing is that when I compile the code, the asm file, and thus the application_image, is wrong. You can see no instructions in the main section of the file main.asm.
main.zip

[stnolting]

The program from your ZIP file (main.c):

int main() {

  int vector[128];
  for (int i = 0; i < 128; i++) {
    vector[i] = i +1;
  }

  return 0;
}

The "problem" here is that the compiler removes all of the code due to optimization. You need some kind of input/output or your have to make the compiler keep your code for example by using volatile variables.

Btw, do you want to simulate your program (and look at waveforms) or do you want to run in on real hardware?

[carlesureta]

Do I need to do anything in the configuration of the processor for that? Like how do I assign the external memory space to that adress?

[stnolting]

You need to enable the external memory interface and you should use the timeout feature. The mapping of the external memory depends on your setup...

I recommend to do a simulation using the default (simple) testbench. This testbench enables all modules of the processor including the external memory interface. Furthermore, the testbench already provides an external memory at address 0xF0000000 (but it is "only" 64 bytes large) so you can use the program I have just posted:

neorv32/sim/simple/neorv32_tb.simple.vhd

Lines 82 to 83 in 3e6a098

	constant ext_mem_c_base_addr_c : std_ulogic_vector(31 downto 0) := x"F0000000"; -- wishbone memory base address (default begin of EXTERNAL IO area)
	constant ext_mem_c_size_c : natural := 64; -- wishbone memory size in bytes

But be aware that this memory has a very high latency. However, I think this is perfectly suited to have a look at the memory request in simulation.

[carlesureta]

I executed this code you sent me but instead of starting at 0xF0000000 i am starting at 0x80000000, and using the internal memory, but nothing is written in the memory positions 0x80000000, and forward. It is just writting different numbers at 0x80001FF4.
This is the code below just to clarify.

int main() {

// initialize the neorv32 runtime environment
// this will take care of handling all CPU traps (interrupts and exceptions)
neorv32_rte_setup();

// setup UART0 at default baud rate, no parity bits, no hw flow control
neorv32_uart0_setup(19200, PARITY_NONE, FLOW_CONTROL_NONE);

int* vector = (int*)0x80000000UL; // map beginning of array to 0xF0000000
for (int i = 0; i < 4; i++) {
vector[i] = i +1;
}

for (int i = 0; i < 4; i++) {
neorv32_uart0_printf("vector[%i] @ 0x%x = %i\n", i, (uint32_t)(&vector[i]), vector[i]);
}

return 0;
}

[carlesureta]

just trying to make it work without external memory first

[stnolting]

Making manual mapping like int* vector = (int*)0x80000000UL; in the DMEM is no good idea as this might override stack, heap or dynamic data.

In case you just want to check out memory access inside the DMEM better let the compiler do the "placement" of the array (no explicit address mapping). Here, I am making vector[] a global variable, so it won't reside in the stack but somewhere in the data section.

volatile int vector[4]; // global variable
 
int main() {

  // initialize the neorv32 runtime environment
  // this will take care of handling all CPU traps (interrupts and exceptions)
  neorv32_rte_setup();

  // setup UART0 at default baud rate, no parity bits, no hw flow control
  neorv32_uart0_setup(19200, PARITY_NONE, FLOW_CONTROL_NONE);

  for (int i = 0; i < 4; i++) {
    vector[i] = i +1;
  }

  for (int i = 0; i < 4; i++) {
    neorv32_uart0_printf("vector[%i] @ 0x%x = %i\n", i, (uint32_t)(&vector[i]), vector[i]);
  }

  return 0;
}

Output:

vector[0] @ 0x80000000 = 1
vector[1] @ 0x80000004 = 2
vector[2] @ 0x80000008 = 3
vector[3] @ 0x8000000c = 4