Hey Maarten,

with your test tool svolume-test, i get some new test results shows
the patch has a positive performance improvement. please note i made
some little change in svolume-test to disable "testing correctness"
part, this save some time, the goal is to find out performance
difference with fixed volume.

i made three times test, one with patch, two without(make sure system
stable during the test).

i paste piece of data here, details are in attached log.

Here's the stastics for format s16, channel 1~4, fixed volume, with patch:
====================
checking s16le - 1 channels - fixed volume
        ref:  70.4 ms   (min =  70.4, max =  71.4, stddev =  0.2)
        asm:   2.9 ms   (min =   2.9, max =   3.0, stddev =  0.0)

checking s16le - 2 channels - fixed volume
        ref:  85.1 ms   (min =  85.1, max =  85.7, stddev =  0.1)
        asm:   5.9 ms   (min =   5.9, max =   6.2, stddev =  0.1)

checking s16le - 3 channels - fixed volume
        ref: 127.8 ms   (min = 127.7, max = 128.6, stddev =  0.2)
        asm:   8.9 ms   (min =   8.9, max =   8.9, stddev =  0.0)

checking s16le - 4 channels - fixed volume
        ref: 170.2 ms   (min = 170.1, max = 170.9, stddev =  0.2)
        asm:  11.8 ms   (min =  11.8, max =  11.9, stddev =  0.0)
=====================

while same format/channel/fixed volume without patch:
=====================
checking s16le - 1 channels - fixed volume
        ref:  70.4 ms   (min =  70.4, max =  70.8, stddev =  0.1)
        asm:   2.9 ms   (min =   2.9, max =   3.0, stddev =  0.0)

checking s16le - 2 channels - fixed volume
        ref: 141.0 ms   (min = 140.9, max = 142.4, stddev =  0.3)
        asm:   6.0 ms   (min =   6.0, max =   6.1, stddev =  0.0)

checking s16le - 3 channels - fixed volume
        ref: 213.7 ms   (min = 213.5, max = 214.5, stddev =  0.2)
        asm:   9.0 ms   (min =   8.9, max =   9.3, stddev =  0.1)

checking s16le - 4 channels - fixed volume
        ref: 286.5 ms   (min = 285.5, max = 301.0, stddev =  2.6)
        asm:  12.0 ms   (min =  12.0, max =  12.3, stddev =  0.1)
==========================

please help review the results.

--xingchao

Attachment: screenlog.1
Description: Binary data

_______________________________________________
pulseaudio-discuss mailing list
pulseaudio-discuss@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/pulseaudio-discuss

Reply via email to