[LAD] MidiSport vs. UA25

[Arnout Engelen]

On Sat, Sep 26, 2009 at 04:28:42AM +0200, Jens M Andreasen wrote:
> On Fri, 2009-09-25 at 21:50 +0200, Arnout Engelen wrote:
> > Attached are the results on an untuned Debian Testing machine (2.6.30-1-686)
> > started with 'chrt -f 50 ./src/alsa-midi-latency-test -s 40'
> 
> What do you mean by "untuned"? Is this without RT-patches?

This was without RT-patches. I now installed the 2.6.31-rt11-686-multimedia 
from 64studio. Without any changes, this actually seemed to perform somewhat 
poorer compared to the stock Debian one.

I did some tweaks:
- increased the rtprio for the USB IRQ handler
- decreased the 'latency_timer' value for my wireless network card and cardbus
  slot
- tried to increase 'latency_timer' for the USB controller, but it remained 
  at 0 - i guess it doesn't support this option

After this, the performance is comparable.

> > The Edirol sure outperformed the MidiSport - the difference didn't seem huge
> > though.
> 
> What I find most disturbing is the distribution, ranging from better
> than perfect to on the edge of unusable. How much of this can be
> attributed to the USB-stack?

Anything I could try to isolate that?


Arnout
-------------- next part --------------
> calibrating high precision timer loop.. done.
time diff =                           0.00
loops/sec =            8096691308977433.00

> sampling 10000 midi latency values - please wait..
> press Ctrl+C to abort test.
> skipping first 40 latency samples.

sample; latency_ms; latency_ms_worst
    41;       2.31;       2.31     
    42;       3.04;       3.04     
    57;       3.15;       3.15     
    58;       3.95;       3.95     
   148;       4.00;       4.00     
   311;       4.01;       4.01     
   434;       4.03;       4.03     
   624;       4.06;       4.06     
   710;       4.06;       4.06     
 10000;       2.40;       4.06     
> done.

> latency distribution:
...
  1.1 -  1.2 ms:        8 #
...
  1.4 -  1.5 ms:        1 #
...
  1.7 -  1.8 ms:       37 #
  1.8 -  1.9 ms:       60 #
  1.9 -  2.0 ms:       20 #
  2.0 -  2.1 ms:        8 #
  2.1 -  2.2 ms:       10 #
  2.2 -  2.3 ms:       91 #
  2.3 -  2.4 ms:     6444 ###################################################
  2.4 -  2.5 ms:     1087 #########
  2.5 -  2.6 ms:      515 #####
  2.6 -  2.7 ms:      241 ##
  2.7 -  2.8 ms:      115 #
  2.8 -  2.9 ms:       94 #
  2.9 -  3.0 ms:      110 #
  3.0 -  3.1 ms:       74 #
  3.1 -  3.2 ms:       43 #
  3.2 -  3.3 ms:       32 #
  3.3 -  3.4 ms:       38 #
  3.4 -  3.5 ms:       38 #
  3.5 -  3.6 ms:       37 #
  3.6 -  3.7 ms:       26 #
  3.7 -  3.8 ms:       24 #
  3.8 -  3.9 ms:       15 #
  3.9 -  4.0 ms:      719 ######
  4.0 -  4.1 ms:      111 #

-------------- next part --------------
> latency distribution:
...
  2.2 -  2.3 ms:        6 #
  2.3 -  2.4 ms:        4 #
...
  2.5 -  2.6 ms:        1 #
  2.6 -  2.7 ms:        1 #
  2.7 -  2.8 ms:        1 #
  2.8 -  2.9 ms:     3345 ###################################
  2.9 -  3.0 ms:     4900 ##################################################
  3.0 -  3.1 ms:      357 ####
  3.1 -  3.2 ms:      362 ####
  3.2 -  3.3 ms:      232 ###
  3.3 -  3.4 ms:       98 #
  3.4 -  3.5 ms:       55 #
  3.5 -  3.6 ms:       27 #
  3.6 -  3.7 ms:       16 #
  3.7 -  3.8 ms:       17 #
  3.8 -  3.9 ms:       18 #
  3.9 -  4.0 ms:       12 #
  4.0 -  4.1 ms:       16 #
  4.1 -  4.2 ms:        7 #
  4.2 -  4.3 ms:       14 #
  4.3 -  4.4 ms:       16 #
  4.4 -  4.5 ms:        9 #
  4.5 -  4.6 ms:        6 #
  4.6 -  4.7 ms:       13 #
  4.7 -  4.8 ms:        4 #
  4.8 -  4.9 ms:        9 #
  4.9 -  5.0 ms:      369 ####
  5.0 -  5.1 ms:       83 #

> SUCCESS