Re: [gentoo-user] lazy gcc switching

[Alan McKinnon]

On Thursday 22 July 2010 02:03:15 Dale wrote:
> Alan McKinnon wrote:
> > On Thursday 22 July 2010 00:18:05 Bill Longman wrote:
> >> On 07/21/2010 12:39 PM, Alan McKinnon wrote:
> >>> On Wednesday 21 July 2010 17:49:46 Bill Longman wrote:
> >>>> And to play devil's advocate, I'll chime in with my experience. The
> >>>> 4.4 GCC, at least on AMD CPUs, creates noticeably faster code. I
> >>>> recompiled all my packages after I upgraded to 4.4 and it was a
> >>>> noticeable difference.
> >>>>
> >>>> But, to make perfectly clear what Alan and Dale have stated
> >>>> previously, it is not a requirement to recompile anything. The
> >>>> binaries that are created still call the same system calls as they
> >>>> did before. The kernel still publishes them in the same locations.
> >>>> And to prove to yourself this is true, grab a statically linked
> >>>> binary, compiled for a stock standard i686, and run it on your
> >>>> machine.
> >>>
> >>> I'd love to be able to experience the speedups of gcc-4.4 and by rights
> >>> I should be able to - my last "rip gentoo apart and put it back
> >>> together again" stunt needed an emerge -e world to fix it all.
> >>>
> >>> But, and this is the bit that makes me cry, the slowdown from KDE-4.4.5
> >>> has obliterated all that advantage several times over.....
> >>>
> >>> raster *really* needs to hurrry up now and release e17
> >>
> >> Might I suggest a small hardware upgrade:
> >>   http://www.supermicro.com/Aplus/motherboard/Opteron6100/SR56x0/H8QGi-F
> >>   .cfm
> >
> > Might I submit that that will be a tad difficult to squeez into this:
> >
> > # dmidecode | grep -B3 "Product Name"
> > Handle 0x0100, DMI type 1, 27 bytes
> > System Information
> >
> >          Manufacturer: Dell Inc.
> >          Product Name: XPS M1530
> > :
> > :-)
>
> Heck, the mobo most likely cost more than your whole laptop.  Froogle
> reports over $700.00 for that thing.  O_O   I wouldn't want the light
> bill for that thing tho.  I would like to see foldingathome running on
> it.  LOL  Gkrellm would be fun to watch.

Looks like a quad cpu, each one dual core. I've got one of those in the Data 
Centre next door and each core is running that new fancy hyper-threading that 
actually works:

$ cat /proc/cpuinfo 
...
processor       : 15
vendor_id       : GenuineIntel
cpu family      : 6
model           : 26
model name      : Intel(R) Xeon(R) CPU           X5570  @ 2.93GHz
stepping        : 5
cpu MHz         : 1596.000
cache size      : 8192 KB

$ free
             total       used       free     shared    buffers     cached
Mem:      98996716   95962284    3034432          0    1855976   32633760
-/+ buffers/cache:   61472548   37524168
Swap:      4192956          0    4192956

$ top
top - 10:07:17 up 9 days, 10:01,  1 user,  load average: 130.27, 134.99, 
122.32
Tasks: 246 total,   1 running, 245 sleeping,   0 stopped,   0 zombie
Cpu(s): 18.9%us,  0.5%sy,  0.0%ni, 29.2%id, 51.2%wa,  0.0%hi,  0.2%si,  0.0%st
Mem:  98996716k total, 96184800k used,  2811916k free,  1856248k buffers
Swap:  4192956k total,        0k used,  4192956k free, 32848132k cached


The grunt this thing has is unbelievable. Check the load - and the box is 
still completely responsive. It runs a database of traffic through all our 
routers so customers can check their traffic graphs going back 45 days.

On the old hardware we used to have to pamper the bloody thing and do a 
juggling act with all the insert scripts. It was always running two hours 
behind (on a good day). With this new baby, we just let it rip and ram data in 
as fast as we can get it. It now runs 90 seconds behind :-0

Sometimes gigantic amounts of grunt are just the thing you need.





-- 
alan dot mckinnon at gmail dot com