Friday, November 09, 2007
stupid LFSR!!!
i can't believe it! hubert (my tutor) and i spent ages scouring the web for a table with taps for maximum length LFSRs (linear feedback shift registers, if they're maximum length they cycle through every sequence of 0s and 1s except for the all-zeroes sequence, which is very useful for generating random test vectors for an exhaustive test). we found this incredible looking table, gives a solution for LFSRs of width up to 786. and when i saw the date (26th oct 2007, i.e. two weeks ago) i thought, wow lucky for me they published this just before i needed it :)
but now the very first LFSR i try fails! 13 bit, seriously! crap! it's supposed to have a period of 8192, guess what i get... seven!!! seven different vectors!!! i could have made seven different vectors on my own, thank you very much!!! gah!
now the question is: have i made a really stupid mistake (really stupid cause from the document it should be so easy to implement that even i should have got it right first time) or have these geniuses in new zealand managed to get the 13th vector of 786 completely wrong?! craaaaaaaaaaaaap!
maybe they weren't very good at counting: 1, 2, many, lots...
[edit@19:16: oopsy, i've found my devastatingly stupid error: i originally just wrote an LFSR with random taps so that my code would compile before i worked out where to get some sensible maximal length taps. when i wrote the proper LFSR, i forgot to remove the other one, so everything was being done properly and then overwritten... i'm a moron, i know...]
but now the very first LFSR i try fails! 13 bit, seriously! crap! it's supposed to have a period of 8192, guess what i get... seven!!! seven different vectors!!! i could have made seven different vectors on my own, thank you very much!!! gah!
now the question is: have i made a really stupid mistake (really stupid cause from the document it should be so easy to implement that even i should have got it right first time) or have these geniuses in new zealand managed to get the 13th vector of 786 completely wrong?! craaaaaaaaaaaaap!
maybe they weren't very good at counting: 1, 2, many, lots...
[edit@19:16: oopsy, i've found my devastatingly stupid error: i originally just wrote an LFSR with random taps so that my code would compile before i worked out where to get some sensible maximal length taps. when i wrote the proper LFSR, i forgot to remove the other one, so everything was being done properly and then overwritten... i'm a moron, i know...]
Labels: arrgghh, i'm special, pratchetty, uni
» Post a CommentI moderate comments, so they won't appear straight away. PLEASE don't leave ten versions of the same comment... ;-)
© 2006 I am NOT SWISS and I DON'T SOUND SWISS either!
Blogger Templates by GeckoandFly.