/******************************************************************************* * Copyright © 2017 TRINAMIC Motion Control GmbH & Co. KG * (now owned by Analog Devices Inc.), * * Copyright © 2023 Analog Devices Inc. All Rights Reserved. This software is * proprietary & confidential to Analog Devices, Inc. and its licensors. *******************************************************************************/ /* * This is a generic implementation for a CRC8 generator supporting * both compile-time (1) and run-time initialized Lookup tables for efficient CRC8 calculation. * You can store multiple tables for different polynomials and (non-)reflected CRCs. * The different tables are referenced by an index, with an upper limit set at compile time (CRC_TABLE_COUNT). * * To generate CRCs you must first generate the Lookup-table by calling fillCRCTable() * with any index. CRCs can then be generated from any data buffer by calling CRC() * with the same index previously given to fillCRCTable(). * * The table generation has been optimized for speed so that the runtime * table generation can even be done during normal operation if required. * However, as long as the required polynomials are known on initialization, * the table generation should be done at that time. * On the Landungsbruecke the initialization of a CRC table takes ~250µs. (2) * Should your application still have problems with the table calculation time, * this algorithm could probably be speed up by preparing a 2- or 4-bit lookup table * to speed up the actual table generation. * * (1): For compile-time CRC tables, just fill the table(s) by initializing CRCTables[] to the proper values. * (2): Tested by toggling a GPIO pin, generating a table in-between and measuring the GPIO pulse width. */ #include "CRC.h" typedef struct { uint8_t table[256]; uint8_t polynomial; bool isReflected; } CRCTypeDef; CRCTypeDef CRCTables[CRC_TABLE_COUNT] = { 0 }; static uint8_t flipByte(uint8_t value); static uint32_t flipBitsInBytes(uint32_t value); /* This function generates the Lookup table used for CRC calculations. * * Arguments: * uint8_t polynomial: The CRC polynomial for which the table will be generated. * bool isReflected: Indicator whether the CRC table will be reflected or not. * uint8_t index: The index of the table to be filled. * * How it works: * A CRC calculation of a byte can be done by taking the byte to be CRC'd, * shifting it left by one (appending a 0) and - if a 1 has been shifted out - * XOR-ing in the CRC polynomial. After 8 iterations the result will be the * CRC of the Byte. * * The function below does this in a compact way, by using all 4 bytes of a * uint32_t to do 4 separate CRC bytes at once. * For this to work without the Byte shifting interfering with adjacent bytes, * the polynomial has the 8th bit (0x100) set. That way, if the shifted-out bit * is 1, the following XOR-ing with the CRC polynomial will set that 1 to a 0, * resulting in the shifted-in 0 for the adjacent byte. * This process will go from the the lowest to the highest byte, resulting in * fully independent byte-wise CRC calculations. For the highest byte, the value * of the shifted-out byte needs to be stored before shifting the bytes (isMSBSet). * * The for-loop that iterates over all uint8_t values starts out with the * uint8_t values 3 to 0 stored in one uint32_t: 0x03020100 * for each iteration each uint8_t value will increase by 4.. * 0 -> 4 -> 8 -> C -> ... * 1 -> 5 -> 9 -> D -> ... * 2 -> 6 -> A -> E -> ... * 3 -> 7 -> B -> F -> ... * ..resulting in an increase of the uint32_t by 0x04040404: * 0x03020100 -> 0x07060504 -> 0x0B0A0908 -> 0x0F0E0D0C -> ... * The loop ends as soon as we have iterated over all uint8_t values. * We detect that by looking for the byte-wise overflow into the next byte: * 0xFFFEFDFC <- last uint32_t value to be calculated * 0xFF, 0xFE, 0xFD, 0xFC <- the corresponding uint8_t values * 0x103, 0x102, 0x101, 0x100 <- incremented uint8_t values (overflow into the next byte!) * 0x04030200 <- uint32_t value with the overflowed bytes * * We have the lower uint8_t values at the lower bytes of the uint32_t. * This allows us to simply store the lowest byte of the uint32_t, * right-shift the uint32_t by 8 and increment the table pointer. * After 4 iterations of that all 4 bytes of the uint32_t are stored in the table. */ uint8_t tmc_fillCRC8Table(uint8_t polynomial, bool isReflected, uint8_t index) { uint32_t CRCdata; // Helper pointer for traversing the result table uint8_t *table; if(index >= CRC_TABLE_COUNT) return 0; CRCTables[index].polynomial = polynomial; CRCTables[index].isReflected = isReflected; table = &CRCTables[index].table[0]; // Extend the polynomial to correct byte MSBs shifting into next bytes uint32_t poly = (uint32_t) polynomial | 0x0100; // Iterate over all 256 possible uint8_t values, compressed into a uint32_t (see detailed explanation above) uint32_t i; for(i = 0x03020100; i != 0x04030200; i+=0x04040404) { // For reflected table: Flip the bits of each input byte CRCdata = (isReflected)? flipBitsInBytes(i) : i; // Iterate over 8 Bits int j; for(j = 0; j < 8; j++) { // Store value of soon-to-be shifted out byte uint8_t isMSBSet = (CRCdata & 0x80000000)? 1:0; // CRC Shift CRCdata <<= 1; // XOR the bytes when required, lowest to highest CRCdata ^= (CRCdata & 0x00000100)? (poly ) : 0; CRCdata ^= (CRCdata & 0x00010000)? (poly << 8 ) : 0; CRCdata ^= (CRCdata & 0x01000000)? (poly << 16) : 0; CRCdata ^= (isMSBSet)? (poly << 24) : 0; } // For reflected table: Flip the bits of each output byte CRCdata = (isReflected)? flipBitsInBytes(CRCdata) : CRCdata; // Store the CRC result bytes in the table array *table++ = (uint8_t) CRCdata; CRCdata >>= 8; *table++ = (uint8_t) CRCdata; CRCdata >>= 8; *table++ = (uint8_t) CRCdata; CRCdata >>= 8; *table++ = (uint8_t) CRCdata; } return 1; } /* This function calculates the CRC from a data buffer * * Arguments: * uint8_t *data: A pointer to the data that will be CRC'd. * uint32_t bytes: The length of the data buffer. * uint8_t index: The index of the CRC table to be used. */ uint8_t tmc_CRC8(uint8_t *data, uint32_t bytes, uint8_t index) { uint8_t result = 0; uint8_t *table; if(index >= CRC_TABLE_COUNT) return 0; table = &CRCTables[index].table[0]; while(bytes--) result = table[result ^ *data++]; return (CRCTables[index].isReflected)? flipByte(result) : result; } uint8_t tmc_tableGetPolynomial(uint8_t index) { if(index >= CRC_TABLE_COUNT) return 0; return CRCTables[index].polynomial; } bool tmc_tableIsReflected(uint8_t index) { if(index >= CRC_TABLE_COUNT) return false; return CRCTables[index].isReflected; } // Helper functions static uint8_t flipByte(uint8_t value) { // swap odd and even bits value = ((value >> 1) & 0x55) | ((value & 0x55) << 1); // swap consecutive pairs value = ((value >> 2) & 0x33) | ((value & 0x33) << 2); // swap nibbles ... value = ((value >> 4) & 0x0F) | ((value & 0x0F) << 4); return value; } /* This helper function switches all bits within each byte. * The byte order remains the same: * [b31 b30 b29 b28 b27 b26 b25 b24 .. b7 b6 b5 b4 b3 b2 b1 b0] * || * \||/ * \/ * [b24 b25 b26 b27 b28 b29 b30 b31 .. b0 b1 b2 b3 b4 b5 b6 b7] */ static uint32_t flipBitsInBytes(uint32_t value) { // swap odd and even bits value = ((value >> 1) & 0x55555555) | ((value & 0x55555555) << 1); // swap consecutive pairs value = ((value >> 2) & 0x33333333) | ((value & 0x33333333) << 2); // swap nibbles ... value = ((value >> 4) & 0x0F0F0F0F) | ((value & 0x0F0F0F0F) << 4); return value; }