Today, Steven Wagner wrote forth saying... > > BIG difference. i hope that none of you are killing web developers. ;) > > Too late, I already fed the pigs.
:P > Anyway, I understand. Those keys are so close together. hehe... very funny... careful. my pigs are hungry too. ;) > Although I'm not averse to writing documentation, I don't feel like > installing a jillion software packages to enable me to do it. Besides, I > haven't really added any features, just ported a buncha stuff. docbook is your friend... the learning/installation curve is well worth the payoff. i was hoping your great sense of humor would get sprinkled into the docs to make them more fun to read. > Speaking of which, I've got about 20% of the metrics done for Tru64. I > think. You never know 'till it's in production and starts acting funny. > Which also reminds me, after this I'm going to try very hard to figure out > why the heck multi-proc IRIX boxes don't report CPU percentages. > > I might move the CPU percentage calculation function into libganglia, since > it's the same everywhere, and put just the machine-specific cycle-gathering > code in the /gmond/machines/* file. From what I've found so far on > different platforms, pretty much everybody has sys, user, idle, and a nice > or WIO counter... and not only would it make porting easier, it would also > ensure that the percentage stats collected on each platform would be > determined using the same algo. Such as it is. :) > > Hmmmm. i really like this idea. we could do more elaborate running average style algorithms like the one in the old top program. the function would need to be threadsafe and reentrant but it's very doable especially for you. -matt
