Posted on May 17, 2011 at 13:17I spent a *large* amount of time recently determining this. The trouble is that the web is full of ''CRC programs'' and many produce different but valid results. The issue is that CRC's are bit based but programmers typically operate on bytes or words.
My advice, to all, is to read (and understand) this paper on the subject
http://www.ross.net/crc/download/crc_v3.txt
This contains C source code that parameterizes all the variant ways of calculating CRC's (unlike most stuff you'll find with Google).
I wrote my own program, using crcmodel.c & crcmodel.h from the above file, to calculate CRC's the same (undocumented) way that the STM32 CRC unit does.
Here it is for the benefit of all
🙂 Code:
<BR>////////////////////////////////////////////////////////////////////////// <BR>// <BR>// FUNCTION NAME : <BR>// CRC_CalcBlockCRC <BR>// <BR>// <BR>// FUNCTIONAL DESCRIPTION : <BR>// Calculate a CRC the same way as the STM32F10x hardware generator. <BR>// <BR>// <BR>// FORMAL INPUT PARAMETERS : <BR>// buffer - pointer to series of 32-bit words (NOT bytes). <BR>// words - count of 32-bit words to do. <BR>// <BR>// <BR>// FORMAL OUTPUT PARAMETERS : <BR>// None <BR>// <BR>// <BR>// RETURN VALUE : <BR>// 32-bit CRC value as per the STM32F10x CRC generator. <BR>// <BR>// <BR>// IMPLICIT INPUTS AND OUTPUTS : <BR>// None <BR>// <BR>// <BR>// IMPORTED FUNCTIONS AND MACROS : <BR>// cm_ini(), cm_nxt(), cm_crc() <BR>// <BR>// <BR>////////////////////////////////////////////////////////////////////////// <BR> <BR>static crc32_t CRC_CalcBlockCRC(ulword *buffer, ulword words) <BR>{ <BR>cm_t crc_model; <BR>ulword word_to_do; <BR>ubyte byte_to_do; <BR>int i; <BR> <BR> // Values for the STM32F generator. <BR> <BR> crc_model.cm_width = 32; // 32-bit CRC <BR> crc_model.cm_poly = 0x04C11DB7; // CRC-32 polynomial <BR> crc_model.cm_init = 0xFFFFFFFF; // CRC initialized to 1's <BR> crc_model.cm_refin = FALSE; // CRC calculated MSB first <BR> crc_model.cm_refot = FALSE; // Final result is not bit-reversed <BR> crc_model.cm_xorot = 0x00000000; // Final result XOR'ed with this <BR> <BR> cm_ini(&crc_model); <BR> <BR> while (words--) <BR> { <BR> // The STM32F10x hardware does 32-bit words at a time!!! <BR> <BR> word_to_do = *buffer++; <BR> <BR> // Do all bytes in the 32-bit word. <BR> <BR> for (i = 0; i < sizeof(word_to_do); i++) <BR> { <BR> // We calculate a *byte* at a time. If the CRC is MSB first we <BR> // do the next MS byte and vica-versa. <BR> <BR> if (crc_model.cm_refin == FALSE) <BR> { <BR> // MSB first. Do the next MS byte. <BR> <BR> byte_to_do = (ubyte) ((word_to_do & 0xFF000000) >> 24); <BR> word_to_do <<= 8; <BR> } <BR> else <BR> { <BR> // LSB first. Do the next LS byte. <BR> <BR> byte_to_do = (ubyte) (word_to_do & 0x000000FF); <BR> word_to_do >>= 8; <BR> } <BR> <BR> cm_nxt(&crc_model, byte_to_do); <BR> } <BR> } <BR> <BR> // Return the final result. <BR> <BR> return (cm_crc(&crc_model)); <BR>} <BR>
Edit: BTW the above was written to be endian independent--that is it should work on big or little endian machines.
[ This message was edited by: swhite on 16-07-2009 20:35 ]
