Hi All,

Appears that bitstream packing is also one of the most CPU hungry
operations in SBC encoder, which did not get proper attention yet.
Could somebody review this patch and provide feedback/comments?

Benchmarks were done by running the following script:

./sbcenc -s 8 big_buck_bunny_480p_mono.au > /dev/null
./sbcenc -s 8 big_buck_bunny_480p_stereo.au > /dev/null
./sbcenc -s 8 -j big_buck_bunny_480p_stereo.au > /dev/null
./sbcenc -s 4 big_buck_bunny_480p_mono.au > /dev/null
./sbcenc -s 4 big_buck_bunny_480p_stereo.au > /dev/null
./sbcenc -s 4 -j big_buck_bunny_480p_stereo.au > /dev/null
./sbcenc -b 128 -s 8 big_buck_bunny_480p_mono.au > /dev/null
./sbcenc -b 128 -s 8 big_buck_bunny_480p_stereo.au > /dev/null
./sbcenc -b 128 -s 8 -j big_buck_bunny_480p_stereo.au > /dev/null
./sbcenc -b 64 -s 4 big_buck_bunny_480p_mono.au > /dev/null
./sbcenc -b 64 -s 4 big_buck_bunny_480p_stereo.au > /dev/null
./sbcenc -b 64 -s 4 -j big_buck_bunny_480p_stereo.au > /dev/null

Times from ARM device (ARM11):

before patch:
real    9m 42.98s
user    8m 51.06s
sys     0m 39.50s

after patch:
real    6m 35.55s
user    5m 44.09s
sys     0m 40.43s

Times from x86 PC (Intel Core2):

before patch:
real    0m49.437s
user    0m45.267s
sys     0m3.708s

after patch:
real    0m27.606s
user    0m23.869s
sys     0m3.656s

 
Best regards,
Siarhei Siamashka