Thread: ANNOUNCE: Diablo-Turbo first beta available
View Single Post
maacruz is offline maacruz
2011-03-17 , 16:31
Posts: 1,101 | Thanked: 1,185 times | Joined on Aug 2008 @ Spain
#240
Originally Posted by jstokes View Post
You need the A2DP support package in the A2DP thread. That configures MPlayer to connect to your headset in lieu of using the speakers. You may also find the updated BlueZ packages to be of use or, alternatively, lardman's dsp-sbc (I don't know which is better: ARM opts or utilization of the DSP?)
Here you'll find dsp-sbc: https://garage.maemo.org/projects/dsp-sbc/
I'll provide in a couple of days arm optimized BlueZ packages to be part of DT.
I think using the DSP would be better, the CPU will be free to do other things and, by choosing the right op_dsp state, you can get significant battery life savings. Of course, only proper testing can provide the answer.

Originally Posted by Straycat View Post
Well, after some tests with dt packages plus new kernel there are my results:

1 - I don't see my device to be faster than before in any scenario.
LOL
I'll prove you wrong. There is one scenario where the device is faster by any measure or definition.
Open osso-mediaplayer
Start playing a mp3
Code:
time sh -c 'i=0; while [ $i -lt 5000 ]; do i=$((i+1)); done'

real    0m0.852s
user    0m0.762s
sys     0m0.008s
Stop playing the mp3
Code:
echo 0 >/sys/power/op_dsp
Start playing the mp3
Code:
time sh -c 'i=0; while [ $i -lt 5000 ]; do i=$((i+1)); done'

real    0m0.678s
user    0m0.621s
sys     0m0.008s
There, you have it, 20% faster in a pure cpu bound task
I can also provide a scenario with 40% longer battery life, just play music after 'echo 3 >/sys/power/op_dsp'
Now, except in this scenario, I never promised a faster device (since the cpu is the same and its maximum frequency is 400 MHZ it can't go any faster), but a more responsive one. Unfortunately it is much more difficult to provide an objective measurement for responsiveness.
2 - I can't to shutdown properly.
Yes, it's an old issue. I'll finally provide updated mini-rc packages to workaround it, since I can't reproduce it in qemu.
 

The Following 5 Users Say Thank You to maacruz For This Useful Post: