HAL on Nokia 770

Rob Taylor rob.taylor at collabora.co.uk
Fri Oct 20 11:33:58 PDT 2006


I've been doing some testing of HAL on the Nokia 770 hardware, here's
some initial profiling results:

With Hal 0.5.8.1 with some fixes and --exit-after-probing backported

Memory usage (derived from smaps)
---------------------------------

hald: heap: 992KB, stack: 268KB, codespace 24KB
hald-runner: 40KB     stack: 12KB, codespace: 12 KB
hald-addon-keyboard: heap: 24KB, stack: 8KB, codespace 4 KB

and shared:
/usr/lib/libhal.so.1.0.0 codespace 4KB

Timing
------
time sh -c '/usr/sbin/hald --daemon=no --exit-after-probing > /dev/null
2>&1'
real    0m 5.35s
user    0m 3.79s
sys     0m 0.76s


With git HEAD:

Occasionally hald seems to hang after printing "Runner started - allowed
paths are '/usr/lib/hal:/usr/lib/hal/scripts:/usr/bin'", which obviously
needs some looking into, but when it runs...

Memory usage (derived from smaps)
---------------------------------

hald: heap: 276KB, stack: 112Kb, codespace: 12 KB
hald-runner: heap: 44KB, stack: 12KB, codespace 12 KB
hald-addon-keyboard:  24KB, stack: 8KB, codespace 4 KB

and shared:
/usr/lib/libhal.so.1.0.0: codespace: 4 KB

Timing
------

time sh -c '/usr/sbin/hald --daemon=no --exit-after-probing > /dev/null
2>&1'
real    0m 1.84s
user    0m 0.82s
sys     0m 0.25s


The results are so dramatically different for head, I'm slightly
suspicious I may have introduced a problem, so I'll look more deeply
into this, along with the hang - but the number of devices reported is
the same between 0.5.8.1 and HEAD, so it may just be that big
congratulations are due to David and his optimisation spree!


Thanks,
Rob Taylor
Collabora Ltd


More information about the hal mailing list