Andree Buschmann
|
2dbb424eb9
|
Further performance optimization of the atrac3 decoder. Rework the internal sample representation and usage of dsp routines. For now a quick and dirty solution is used to add a fract part of 2 bits. Through this several buffers and functions as well as copy loops could be removed. Furthermore add some ASM for coldfire and place some additional data in IRAM on PP5022/24 and X5/M5. Speedup on ARM: +3%, speedup on Coldfire: +639%. Both ARM and Coldfire can decode in realtime now.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@22561 a1c6a512-1295-4272-9138-f99709370657
|
2009-08-30 14:14:22 +00:00 |
|
Andree Buschmann
|
f4774bf5bf
|
Submit interim version of FS#10565. Performance optimization of atrac3 decoder for ARM. Introduce ASM routines for multiplications and two synthesis loops, refactured parts of synthesis and windowing. Speeds up decoding by a factor of 2.4 on PP502x.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@22548 a1c6a512-1295-4272-9138-f99709370657
|
2009-08-29 19:44:27 +00:00 |
|
Jens Arnold
|
3d6f86eb2f
|
Make those functions actually inline. Around 20% speedup on coldfire, 10% speedup on arm.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@22344 a1c6a512-1295-4272-9138-f99709370657
|
2009-08-16 12:41:04 +00:00 |
|
Mohamed Tarek
|
4f26112b1b
|
Revert the asm fixmul stuff for now.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@22310 a1c6a512-1295-4272-9138-f99709370657
|
2009-08-14 17:31:08 +00:00 |
|
Mohamed Tarek
|
a0092593c9
|
Call the asm definitions of fixmul functions when possible.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@22303 a1c6a512-1295-4272-9138-f99709370657
|
2009-08-13 21:18:30 +00:00 |
|
Mohamed Tarek
|
432e2ecc13
|
Modify libatrac to use fixed-point arithmetic.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@22298 a1c6a512-1295-4272-9138-f99709370657
|
2009-08-13 20:38:59 +00:00 |
|