On Wednesday 20 April 2011 11:25:27 Joost Roeleveld wrote: > On Wednesday 20 April 2011 03:42:13 Dale wrote: > > Joost Roeleveld wrote: > > > On Tuesday 19 April 2011 17:54:02 Dale wrote: > > >> Kfir Lavi wrote: > > >>> I do a lot of compiling on my laptop using Catalyst for > > >>> embedded. > > >>> I upgraded my RAM to 8GB (2x4GB = 90$ ebay). > > >>> I mount /var/tmp as tmpfs. > > >>> Thats how I keep my SSD ;) > > >>> > > >>> Kfir > > >>> > > >>> -- > > >>> Poison [BLX] > > >>> Joshua M. Murphy > > >> > > >> I got 16Gbs in my rig and I mounted portages work directory on > > >> tmpfs, > > >> it > > >> was actually slower. That is likely a good idea to keep from > > >> wearing > > >> out the SSD but it doesn't seem to make anything compile faster. > > > > > > Strange, it actually got faster on mine when doing that. > > > Did you mount with "noatime"? :) > > > > > > -- > > > Joost > > > > Yep. It actually took a few seconds longer compiling in memory. I > > thought that was weird to tho. It doesn't make much sense to me. > > > > Dale > > > > :-) :-) > > Hmm... > Will be doing some timing-tests then.... > Openoffice is a good one for that ;)
Ok, just done the tests. using tmpfs for /var/tmp/portage is quicker, but not by much. Without TMPFS: # time emerge -v openoffice real 32m44.742s user 20m18.320s sys 5m38.000s With TMPFS: # mount -onoatime,size=7G -ttmpfs none /var/tmp/portage # time emerge -v openoffice real 31m30.835s user 20m3.510s sys 5m38.030s Specification of this machine: 12GB RAM Quad Core Xeon W3565 @ 3.2Ghz with HT enabled There are 2 drives in stripe-mode (software RAID-0) which does speed up the I/O a lot. -- Joost
