As I said before, the test results from the Laptop showed exactly no
change except from KDiskFree which started up much faster. I figured
you wouldn't be happy if I cluttered your list with all that data.
The desktop system is an AMD XP 1700+, JetWay J876, 512MB DDR 2100,
nVidia GF2MX200 PCI, Philips Harmonic Edge (runs so badly in Windows
that palettes of these things are being given away...they run perfectly
in Linux. Hah!), some Tulip PCI network card, 2 80GB IDE hard drives at
7200 RPM with 8MB cache, ATI TV Wonder, 3.5" and 5.25" (yay!) floppy, HP
DeskJet 842c (USB) and LaserJet 4 and a Plustek LPT scanner, and a
52x24x52x recorder. I'm running Red Hat 8.0 ("everything" installation)
with all updates applied. The SRPMS were from the install CDs with the
updated RPMs so that we're comparing apples to apples.
To perform these tests, I first booted clean twice, then performed
exactly the same set of operations each time (except for my XScreenSaver
mistake)
first start -
time to login screen:1:05 (From GRUB to "Enter username")
time to full desktop:0:15 (From "Enter username" to seeing System
Monitor)
time to start gedit:0:01
memory in use:36% (according to system monitor, although this isn't
totally accurate)
time to start OO.o:0:23
time to start Galeon:0:06
time to start xscreensaver:0:01
gears (normal) speed:50 fps
time to start Konqueror:0:06
time to start KDiskFree:0:02
time to start Abiword:0:01
memory in use:63%
second start - (Because of the xscreensaver mistake, this is a more
accurate run for comparison)
time to login screen:1:01
time to full desktop:0:15
time to start gedit:0:01
memory in use:40%
time to start OO.o:0:12
time to start Galeon:0:10
time to start xscreensaver:0:10
gears (normal) speed:50 fps
time to start Konqueror:0:05
time to start KDiskFree:0:02
time to start Abiword:0:01
memory in use:72%
(I cheated a little...I had XScreenSaver set to "disable" the first
time, and I turned it on to test gears. I forgot and left it on. I
will leave it on from here on in.)
first start after installing rebuilt RPMs-
time to login screen:1:07
time to full desktop:0:25
time to start gedit:0:01
memory in use:54%
time to start OO.o:0:25
time to start Galeon:0:05
time to start xscreensaver:0
gears (normal) speed:50 fps
time to start Konqueror:0:08
time to start KDiskFree:0:02
time to start Abiword:0:01
memory in use:86%
(woah, that's not good! everything bad, very bad!)
second start after installing rebuilt RPMs-
time to login screen:0:58
time to full desktop:0:16
time to start gedit:0:01
memory in use:31%
time to start OO.o:0:22
time to start Galeon:0:05
time to start xscreensaver:0
gears (normal) speed:50
time to start Konqueror:0:06
time to start KDiskFree:0
time to start Abiword:0
memory in use:63%
(err, is this an exceptional case?)
third start after installing rebuilt RPMs-
time to login screen:0:50
time to full desktop:0:16
time to start gedit:0:01
memory in use:31%
time to start OO.o:0:14
time to start Galeon:0:10
time to start xscreensaver:0
gears (normal) speed:50
time to start Konqueror:0:06
time to start KDiskFree:0
time to start Abiword:0
memory in use:63%
(I see...)
So, comparing the second stock start and the third start after
installing rebuilt RPMs, these are the differences.
time to login screen: - 0:11
time to full desktop: + 0.01
time to start gedit: no change
memory in use: - 9%
time to start OO.o: + 0:02 (grr)
time to start Galeon: no change
time to start xscreensaver: - 0:10
gears (normal) speed: no change
time to start Konqueror: + 0:01
time to start KDiskFree: - 0:02
time to start Abiword: - 0:01
memory in use: - 9%
Total time difference - 0:20 seconds
Assuming that I generally start each program once per day, I've saved
myself 20 seconds per day. Over a year, that comes to two hours saved.
Considering that the server was not slowed down or incapacitated during
recompiling, and also assuming that the binaries are tighter and more
reliable, it was worth it. In actuality, I start each program many many
times per day, and I run much more than that, so I'd say I've saved many
minutes from each day. Since several computers here have the same
configuration as mine, the time saving has been multiplied.
It's still not as fast as Vector or a sources-based installation, but I
still get my lovely Red Hat. ^_^
Next, I'll do the tests with prelinking. By saving all the rebuilt
RPMs, I can easily jump between optimized and not optimized in case
anyone wants me to test anything else.
