STM32F412 CRC32 uint8_t array as input.

[NKorn.12]

Hi,

I have (for example) a byte array with the size of 6. Now I want to calculate its crc32 value. The first 4 bytes are no problem - I put them in as word, but what about the last 2 bytes?

Thanks for the Help in advance.

BR

Nico

[Tesla DeLorean]

The HW CRC on the part operates on 32-bit at a time. To handle odd bytes you must either use consistent padding bytes or work the smaller amounts​ in SW.

[waclawek.jan]

uint32_t CrcCCITTBytes(const uint8_t * data, uint32_t size);
    uint32_t CrcCCITTBytes(const uint8_t * data, uint32_t size) {
      uint32_t i;
      CRC->CR = CRC_CR_RESET;
      while(CRC->CR & CRC_CR_RESET);  // avoiding the not-reset-fast-enough bug
      i = size % 4;
      switch(i) {
        case 0:
          break;
        case 1:
          CRC->DR = 0xFFFFFFB9;
          CRC->DR = 0xAF644900
            | (__RBIT(*(uint32_t*)(uintptr_t)&data[0]) >> 24)
          ;
          break;
        case 2:
          CRC->DR = 0xFFFFB950;
          CRC->DR = 0x64490000
            | (__RBIT(*(uint32_t*)(uintptr_t)&data[0]) >> 16)
          ;
          break;
        case 3:
          CRC->DR = 0xFFB9509B;
          CRC->DR = 0x49000000
            | (__RBIT(*(uint32_t*)(uintptr_t)&data[0]) >> 8)
          ;
          break;
      }
 
      for (; i < size; i += 4) {
        CRC->DR = __RBIT(*(uint32_t*)(uintptr_t)&data[i]);
      }
      return __RBIT(CRC->DR) ^ 0xFFFFFFFF;
   }

This is to demonstrate how to

> use consistent padding bytes

I started with one, two or three extra 0xFF, calculated CRC from just them, and then added the "data to change CRC to whatever I want" (and I wanted 0xFFFFFFFF) as per Clive's recipe (last post in that thread). That's how those constants above came to be. The rest is input and output bit reversal as per CCITT-32, and the final xor.

JW