C8051F93X Datasheet, PDF (169/330 Pages) Silicon Laboratories – Pipelined intstruction architecture executes 70 of instruction in 1 or 2 system clocks

C8051F93x-C8051F92x
15.2. 32-bit CRC Algorithm
The C8051F93x-C8051F92x CRC unit calculates the 32-bit CRC using a poly of 0x04C11DB7. The CRC-
32 algorithm is "reflected", meaning that all of the input bytes and the final 32-bit output are bit-reversed in
the processing engine. The following is a description of a simplified CRC algorithm that produces results
identical to the hardware:
Step 1. XOR the least-significant byte of the current CRC result with the input byte. If this is the
first iteration of the CRC unit, the current CRC result will be the set initial value
(0x00000000 or 0xFFFFFFFF).
Step 2. Right-shift the CRC result.
Step 3. If the LSB of the CRC result is set, XOR the CRC result with the reflected polynomial
(0xEDB88320).
Step 4. Repeat at Step 2 for the number of input bits (8).
For example, the 32-bit 'F93x/92x CRC algorithm can be described by the following code:
unsigned long UpdateCRC (unsigned long CRC_acc, unsigned char CRC_input)
{
unsigned char i; // loop counter
#define POLY 0xEDB88320 // bit-reversed version of the poly 0x04C11DB7
// Create the CRC "dividend" for polynomial arithmetic (binary arithmetic
// with no carries)
CRC_acc = CRC_acc ^ CRC_input;
// "Divide" the poly into the dividend using CRC XOR subtraction
// CRC_acc holds the "remainder" of each divide
//
// Only complete this division for 8 bits since input is 1 byte
for (i = 0; i < 8; i++)
{
// Check if the MSB is set (if MSB is 1, then the POLY can "divide"
// into the "dividend")
if ((CRC_acc & 0x00000001) == 0x00000001)
{
// if so, shift the CRC value, and XOR "subtract" the poly
CRC_acc = CRC_acc >> 1;
CRC_acc ^= POLY;
}
else
{
// if not, just shift the CRC value
CRC_acc = CRC_acc >> 1;
}
}
// Return the final remainder (CRC value)
return CRC_acc;
}
The following table lists several input values and the associated outputs using the 32-bit 'F93x/92x CRC
algorithm (an initial value of 0xFFFFFFFF is used):
Rev. 1.3
169