Maurice and Jerry,
Not one, but two completely general CRC packages! Thank you, thank you, thank you. The only thing for me to add is the ability to compute CRCs over bit fields whose width is not an integer multiple of 8 (suggestions welcome). My approach is going to be to compute the CRC over as many bytes as possible, then add in the effect of the final bits.
Again, thank you both very much.
Doug.
=> 12/22/99 1:00:59 PM EST, djm@san.rr.com => << Does anybody out there have code for generic CRC generation? The implementation does not have to be efficient. I am aware of the 16-bit algorithm in the current image, but need to be able to generate CRCs with non-standard polynomials for arbitrary bit length messages. Thanks in advance. >>
Doug,
Attached is CRC package I did a while ago. I hacked a change to remove BitMasks Class Variable from class Integer; it wouldn't load on 2.6 otherwise. 2.7a fixes bug. Much slower table creation without that Class Variable. It cached BitMasks (1-bit in a width of 0s). This is based rather strictly on the paper by Ross N. Williams. I built the table driven version and then went back and added the reference version from the paper to fully document the program. I later added an 'incremental' version <processes crc's a character at a time, but uses tables for calculating intermediate results>. There are various test methods to execute internal tests; they are probably the best way to figure out how to add your own needs to the package..
I've taken a quick look at Maurice's package. His algorithms are based on Peterson & Weldon's 'Error Correction Codes' so the packages look different when setting up your own crc processor. His goal of using 30 bits for intermediate results is definitely the right way to go for Squeak (with 2 extra bits over SmallInteger in a 32-bit word) so I'll have to figure out how he does the algorithm.