Memory TIMEOUT ERROR

[dboufeas]

Hi everyone,
I'm new to FPGAs, embedded systems, etc. and I want to implement PRESENT cipher in NEORV32. I first started to implement a software version and then I want to add a hardware module to do the crypto-operations.

The problem is that in my program I want to save the generated round keys so I used "neorv32_cpu_load_unsigned_word()" function but I get a "Load access fault [TIMEOUT_ERR] " when I try to save in specific addresses.

I'm simulating the processor with makefiles in the console. Is something wrong with my code? I attached my code and the output I get in the files below.

console_output.txt
main.txt

[stnolting]

Hey Jimmy!

I first started to implement a software version and then I want to add a hardware module to do the crypto-operations.

Interesting project! 👍
This seems a like a great application for the processor's Custom Functions Unit. 😄

The problem is that in my program I want to save the generated round keys so I used "neorv32_cpu_load_unsigned_word()" function but I get a "Load access fault [TIMEOUT_ERR] " when I try to save in specific addresses.

It is always "dangerous" to use hard-coded pointers in C (however, I also do that a lot). 😅 In the case of your application, you are manually accessing a memory region starting at 0xA0000000. But the default testbench does not provide any memory mapped to this address space. This is why the access times out - as there is nothing that could respond to the memory request.

📚 The available address space / memories provided by the default testbench are listed in Table 3 in UG: 14.1. Testbench.

---

I would recommend to use a simple array and let the compiler do the address mapping.

So, instead of accessing memory using a pointer

  // key shedule
  uint32_t addr, data;
  addr = 0xA0000000;

  // save first round key lsb
  data = (uint32_t) key.high64;
  neorv32_cpu_store_unsigned_word(addr, data);
  addr += 4;

...

    // save round key LSB
    data = (uint32_t) key.high64;
    neorv32_cpu_store_unsigned_word(addr, data);
    addr += 4;

...

  // print saved round keys (LSB)
  addr = 0xA0000000;
  for (uint8_t i = 0; i < 32; i++) {
    data = neorv32_cpu_load_unsigned_word(addr);
    addr += 4;
    neorv32_uart0_printf("%d: %x\n", i, data);
  }

just use a plain array with an index:

  // key shedule
  uint32_t data[32];

  // save first round key lsb
  data[0] = (uint32_t) key.high64;

...

    // save round key LSB
    data[i] = (uint32_t) key.high64;

...

  // print saved round keys (LSB)
  for (uint8_t i = 0; i < 32; i++) {
    neorv32_uart0_printf("%d: %x\n", i, data[i]);
  }

[dboufeas]

Thanks for the reply! It was very helpful.