Re: [fedora-arm] Final Build Times Comparison Table?
Hi Jon,
On Sun, Aug 12, 2012 at 1:31 AM, Jon Chiappetta
<Jon.Chiappetta(a)senecacollege.ca> wrote:
Ok I made 2 different pages:
http://142.204.133.82/jon/builds/timesa.html -- This is based on all
buildarch times averaged together for each task and then summed together &
compared
http://142.204.133.82/jon/builds/.pa.f18.log.all.log.a -- The CSV data
file containing: pa pkg , pa task , pa time , sa pkg , sa task , sa time
^^^ we might be ~15 days slower possibly
http://142.204.133.82/jon/builds/timesb.html -- This takes the single
fastest buildarch time for each pa task and compares it to the fastest hsv
buildarch time for sa
http://142.204.133.82/jon/builds/.pa.f18.log.all.log.b -- CSV: pa pkg , pa
task , builderA=timeA ; builderB=timeB , sa pkg , sa task , builderA=timeA ;
builderB=timeB
^^^ we might be ~7 days slower possibly
Let me know if you have any complaints, suggestions, or anything. I think
ive spent way too much time on this heh.
This looks great! Thanks for the work on it. It's great to see some
comparisions to mainline to see the actual build differences easily in
front of you and the type of builder it was built on. It's heartening
to see that the quad core devices are very similar to that of the new
mainline VM builders.
For the actual build times I don't think we can explicitly say how
long it will be. Also a rebuild isn't serial so a lot of it will
depend on the amount of builders available to see how parallelised it
can be. For promotion to PA we'll be moving to all quad core 4Gb RAM
builders whree a lot of what we have at the moment is dual core 1Gb of
RAM. Also a lot of the smaller packages are I/O bound.
Thanks again.
Peter