Audio: volume: Optimize source and sink buffers usage

[andrula-song]

This patch replaces the audio stream read/write frag based
access to source and sink by block processing based on
audio_stream_bytes_without_wrap() bytes count.

Test all platform first.

Signed-off-by: Andrula Song xiaoyuan.song@intel.com

[lgirdwood]

@andrula-song do you have a before and after MCPS value to show as why this should be merged ?

[andrula-song]

hi @lgirdwood , the specific platforms's statistic have not been measured yet. but by the xtensa simulation testbench Seppo sent me, the cycles the processing function spent would reduce about 25% than now. In this way, there would only be one circular buffer, we only set the begin and end address once in processing function rather than 3 times in every for loop in processing function. I want to make sure there are no other potential bugs before giving specific measurements.

[andrula-song]

from the xtensa test bench, after this optimization, each format reduce many the cycles the scale_vol function cost.
the optimized result is:

the original result is:

for S16 format, opt version cost 4281000 cycles, the original version cost 5751000 cycles, reduce about 25% cycles
for S24 format, opt version cost 5899500 cycles, the original version cost 8985000 cycles, reduce about 34% cycles
for S32 format, opt version cost 6760500 cycles, the original version cost 9777000 cycles, reduce about 30% cycles

[lgirdwood]

@singalsu good for you ?

[singalsu]

Great that this idea worked. Good work!

[lgirdwood]

@andrula-song any reason why this is still draft ?

[andrula-song]

@andrula-song any reason why this is still draft ?
should I change the commit message to include those optimization info?

[andrula-song]

SOFCI TEST

[lgirdwood]

@plbossart @bardliao CI showing a SDW timeout and a rt711 jack timeout on PM test.
https://sof-ci.01.org/sofpr/PR5452/build12228/devicetest/?model=TGLU_RVP_SDW&testcase=check-suspend-resume-with-playback-5
Should be unrelated to this PR, any ideas ?

[bardliao]

@plbossart @bardliao CI showing a SDW timeout and a rt711 jack timeout on PM test. https://sof-ci.01.org/sofpr/PR5452/build12228/devicetest/?model=TGLU_RVP_SDW&testcase=check-suspend-resume-with-playback-5 Should be unrelated to this PR, any ideas ?

@lgirdwood It is unrelated to this PR. See thesofproject/linux#3459

[lgirdwood]

@kkarask @wszypelt looks like CI may be blocked ?

[miRoox]

SOFCI TEST

[andrula-song]

the failure items is unrelated with this PR, because daily test exist, too.