branch trace compression
Ken Barr
this new framework? I asked the same question of Jared about the CBP1
scheme. They are both similar to value-predictor based compression,
and they are both really cool.
I ended up writing about similar techniques in this year's ISPASS [1]
and my version of what you guys were treating as one-off code done in
spare time actually became a chunk of my thesis :)
I just want to make sure I'm giving credit where credit is due and that
I am heading off any defense-day surprises like "Stark and Jimenez did
this ages ago!"
Thanks for any insight or references,
Ken
1. http://www.cag.lcs.mit.edu/scale/papers/bpred-ispass06.pdf
djimenez
No, I haven't published the compression scheme. Honestly I didn't even
think about it; it was just a kind of fun thing that I spent a few days
on. Maybe that should have been another track in CBP-2 -- see how well
you can compress the traces :-) In terms of publishing, I would have
figured that Martin Burtscher would have covered that area pretty well,
but I'll be interested to see your paper.
--
Daniel Jiménez