Garrett Cooper wrote: > Joao Barros wrote: >> On 6/4/07, Alexander Kabaev <kabaev_at_gmail.com> wrote: >>> On Sun, 3 Jun 2007 17:30:15 -0700 (PDT) >>> youshi10_at_u.washington.edu wrote: >>> >>> > I just wanted to thank everyone who helped make the gcc 4.2 merge >>> > possible -- I can see a noticeable performance increase (1.5x ~ 2.0x >>> > load time decrease) in my server (P4 HTT with SMP). >>> > >>> > This will be one of the great changes that will put FreeBSD back on >>> > the map closer to many Linux distros. >>> > >>> > Once gcc 4.2.1 gets in the tree, things will be much better, I'm >>> > sure :). >>> > >>> > Thank you very much, and let me know if you need testing with gcc >>> > 4.2.1 when it rolls around ;). >>> > >>> > -Garrett >>> > >>> Thanks, but I think kudos are addressed to wrong people. I very much >>> doubt that GCC can be responsible for any, let alone dramatic, >>> performance differences. Most likely you see effects of hard work done >>> by SMP scalability guys. >>> >>> -- >>> Alexander Kabaev >>> >>> >> >> I emailed Garrett but forgot to cc current_at_ >> >> My question was: "Can you specify the workload that you verify that >> much improvement?" >> Like you I had my doubts gcc was responsible for such an improvement. >> >> > The differences I saw (ok, limited), were a decrease in bootup time > (about 15-30 seconds faster), login via ssh (almost instantaneous when > the sshd thread seemingly hasn't gone to sleep), restarting the samba > daemon (3 seconds to ~1 second), and running make buildworld and > installworld. > > The only thing that I've noticed so far is that acpi is broken > again... I was shocked when I hit my power button and my machine > instantly shut off 0_0... > > Synced to HEAD yesterday and rebuilt. > > ASUS P4800DX motherboard with a P4 3.0GHz Northwood chip. > > -Garrett 3 other things I forgot to mention: 1. I don't use WITNESS on my server. 2. I use the 4BSD scheduler, not the ULE scheduler. 3. The last time I synced and built everything was around early May if I remember correctly. -GarrettReceived on Mon Jun 04 2007 - 13:30:13 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:11 UTC