Commit graph

113 commits

Author SHA1 Message Date
Andree Buschmann
3c63cb29e9 Resync mpc to r476 of svn.musepack.org. Improves error handling after seek.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@31093 a1c6a512-1295-4272-9138-f99709370657
2011-11-29 19:48:57 +00:00
Andree Buschmann
fabbeba59a Solve FS#12396 through rolling back r26592. This fixes distortions while mpc playback on Coldfire targets with GCC 4.5.2.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@31054 a1c6a512-1295-4272-9138-f99709370657
2011-11-25 19:42:26 +00:00
Andree Buschmann
f163b405c0 Fully remove floating point code from libmusepack. Introduce two small const arrays for precalculated scalefactors, correct the integer type of scalefactors to unsigned, migrate some metadata calculations to fixed point. No impact to decoder output.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@30497 a1c6a512-1295-4272-9138-f99709370657
2011-09-10 20:35:04 +00:00
Nils Wallménius
b086458e4e FS#12231
Centralize the selection of codec compilation flags to codec.make
to avoid having to duplicate the per cpu selection logic and the
build rule in the individual codec lib make files and make tweaking
easier.

The two special cases, libmad and libspeex were left alone.

I plan to introduce a var for arm generation so that appropriate flags
can be selected per generation as benchmarking results have shown that
different arm arches might have different optimal flags.



git-svn-id: svn://svn.rockbox.org/rockbox/trunk@30330 a1c6a512-1295-4272-9138-f99709370657
2011-08-18 16:33:41 +00:00
Andree Buschmann
7155be6ca4 Sync rockbox to r475 of musepack's svn.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@30240 a1c6a512-1295-4272-9138-f99709370657
2011-08-02 19:08:29 +00:00
Nils Wallménius
7c6056b352 FS#12140 by Sean Bartell, Make various codec stuff static.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@29942 a1c6a512-1295-4272-9138-f99709370657
2011-06-01 10:28:26 +00:00
Andree Buschmann
22c9b6ef8e Use more IRAM for mpc on S5L870x. Speed up decoding by ~2%.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@29891 a1c6a512-1295-4272-9138-f99709370657
2011-05-16 21:31:59 +00:00
Andree Buschmann
fa65362a42 Fix compiler warnings ('variable set ut not used') with GCC 4.6.0.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@29792 a1c6a512-1295-4272-9138-f99709370657
2011-04-27 17:59:49 +00:00
Nils Wallménius
5aaabbfd12 FS#12014 by Huan Zhang fixing some link errors with a recent mips toolchain.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@29615 a1c6a512-1295-4272-9138-f99709370657
2011-03-19 10:14:12 +00:00
Nils Wallménius
c8535f27d1 Switch coldfire builds over to new toolchain using gcc 4.5.2 and binutils 2.20.1
Retune codec compiler optimizations with new compiler. Overall speedup with aac and flac getting big speedups.


git-svn-id: svn://svn.rockbox.org/rockbox/trunk@29042 a1c6a512-1295-4272-9138-f99709370657
2011-01-12 22:28:43 +00:00
Andree Buschmann
0dbc4a6cae Use MEM_ALIGN_ATTR in mpc codec. Speed up by 22-24% was measured on ARM11.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28627 a1c6a512-1295-4272-9138-f99709370657
2010-11-21 14:03:55 +00:00
Andree Buschmann
bb56bc4ba0 Minor speed up of mpc requantization.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28563 a1c6a512-1295-4272-9138-f99709370657
2010-11-12 07:09:09 +00:00
Andree Buschmann
ee610d47c7 Re-submit ARM asm optimizations in mpc synthesis as the performance regressions on S5L870x have been solved with r28561.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28562 a1c6a512-1295-4272-9138-f99709370657
2010-11-12 07:07:57 +00:00
Andree Buschmann
3f4e0cf25b Set alignment for major data arrays in the mpc codec. This avoids performance regressions on S5L870x CPUs when changing code and allows further improvements.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28561 a1c6a512-1295-4272-9138-f99709370657
2010-11-12 07:04:07 +00:00
Andree Buschmann
92c38a728e Roll back r28488 as it slows down mpc by 2% on S5L8701.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28545 a1c6a512-1295-4272-9138-f99709370657
2010-11-09 22:19:07 +00:00
Nils Wallménius
f174c3513b libmusepack: use postindexing on the pointer when storing to increment it for free, saves another instruction in the loop in mpc_decoder_windowing_D speed difference is in the noise though.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28492 a1c6a512-1295-4272-9138-f99709370657
2010-11-05 16:21:25 +00:00
Nils Wallménius
9d98583a18 libmusepack: tiny optimization of the ARMv4 mpc_decoder_windowing_D implementations, using register indexed addressing to store data, saving one instruction in the loop and deleting one instruction adter the loop which isn't needed.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28488 a1c6a512-1295-4272-9138-f99709370657
2010-11-05 11:49:30 +00:00
Nils Wallménius
dbdc0a8a8c libmusepack: ARMv6 assembler for mpc_decoder_windowing_D, speeds up decoding of 128kbps sample file 2MHz, or 8%, on gigabeat S. The output difference to the c implementation and the other ARM implementation is +/-1 in less than 0.1% of the output samples.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28487 a1c6a512-1295-4272-9138-f99709370657
2010-11-05 11:20:50 +00:00
Andree Buschmann
c355122cce Minor optimization of mpc codec via moving dewindowing asm to icode section for S5L870x.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28454 a1c6a512-1295-4272-9138-f99709370657
2010-11-02 21:13:39 +00:00
Andree Buschmann
560d89130f Using icode attribute for S5L870x is faster on some codecs. Speed up for mpc is ~3%, for aac ~1%.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28453 a1c6a512-1295-4272-9138-f99709370657
2010-11-02 21:06:20 +00:00
Andree Buschmann
48bb3d00db Optimize mpc's synthesis filter by reducing stalls for arm9 and above. Speed up ranges from 4% (nano 2g) to 11% (beast).
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28452 a1c6a512-1295-4272-9138-f99709370657
2010-11-02 21:00:34 +00:00
Andree Buschmann
5245c9254a Fix FS#11710. On Coldfire CPU's mpc uses a speed optimization which was not correctly re-initialized when rebuffering.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28437 a1c6a512-1295-4272-9138-f99709370657
2010-11-01 20:43:17 +00:00
Andree Buschmann
27122e4cc9 Fix FS#11650. Correct sliencing of subband in mpc decoder. Thanks to Leo Witt for reporting and supporting.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28200 a1c6a512-1295-4272-9138-f99709370657
2010-10-02 23:02:36 +00:00
Andree Buschmann
2ddcd7fd94 Optimization of mpc's buffering routines through using the maximum possible buffer size. This reduces read access calls.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28198 a1c6a512-1295-4272-9138-f99709370657
2010-10-02 17:52:12 +00:00
Andree Buschmann
32838e96d0 Finally fix FS#11476 (data abort when playing some sv8 mpc files on lowmem targets like Clip+/Fuzev1/Fuzev2). This fix disables loading of seek tables from the sv8 bitstream. Those table are located at the end of sv8-files. When playing sv8 files this resulted in seek to filebegin, then fileend, then back to filebegin. Such seeking behaviour results in heavy rebuffering when playing files on targets with low RAM.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@28197 a1c6a512-1295-4272-9138-f99709370657
2010-10-02 17:43:36 +00:00
Thomas Martitz
87409a2619 Move memset6() declaration to string-extra.h, kills a warning compiling for android since it ships a memory.h.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@27789 a1c6a512-1295-4272-9138-f99709370657
2010-08-12 13:55:01 +00:00
Nils Wallménius
4f5b390a6d Convert inline coldfire assembler to a 'real' assembler function, with tweaks by Buschel. Speeds up mpc decoding by ~1% on h300.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@27504 a1c6a512-1295-4272-9138-f99709370657
2010-07-20 23:35:07 +00:00
Andree Buschmann
d7ee0fe076 Housekeeping. Comment musepack code seqments unused in rockbox.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@27111 a1c6a512-1295-4272-9138-f99709370657
2010-06-24 18:48:04 +00:00
Andree Buschmann
a3541c07de Disable unneeded parts of mpc's file-I/O interface.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@27024 a1c6a512-1295-4272-9138-f99709370657
2010-06-21 18:45:34 +00:00
Andree Buschmann
dc8af759ce Fix FS#11416. Resuming mpc was broken since r26032 due to false assumption of amount of buffered data.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@26990 a1c6a512-1295-4272-9138-f99709370657
2010-06-20 13:02:31 +00:00
Rafaël Carré
45c7498f59 FS#11335 by me: make ARM assembly functions thumb-friendly
We can't pop into pc on ARMv4t when using thumb: the T bit won't be
modified if we are returning to a thumb function
Code running on ARMv4t should use the new ldrpc / ldmpc macros instead
of ldr pc, [sp], #4 and ldm(cond) sp!, {regs, pc}
No modification on pure ARM builds and ARMv5+

Note: USE_THUMB is currently never defined, no targets can currently be
built with -mthumb, see FS#6734

git-svn-id: svn://svn.rockbox.org/rockbox/trunk@26756 a1c6a512-1295-4272-9138-f99709370657
2010-06-11 04:41:36 +00:00
Andree Buschmann
45ab395c2f Some asm code beautification in the mpc decoder.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@26592 a1c6a512-1295-4272-9138-f99709370657
2010-06-05 17:52:31 +00:00
Andree Buschmann
9ab57e510e Small changes to asm for better readability.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@26375 a1c6a512-1295-4272-9138-f99709370657
2010-05-29 14:56:25 +00:00
Andree Buschmann
db42781ca8 Fix FS#11261 (slow seek forward in mpc files). Since the introduction of the sv8 library mpc did not use efficient buffering when seeking forward. With this change the implementation will not flush the buffer on each frame but will keep the already buffered data as long as possible.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@26032 a1c6a512-1295-4272-9138-f99709370657
2010-05-14 22:06:28 +00:00
Nils Wallménius
abd6b147cc Restor setting of EMAC mode on coldfire lost in r25056 fixes whitenoise when transitioning from a codec using a different EMAC mode.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@26013 a1c6a512-1295-4272-9138-f99709370657
2010-05-14 10:48:26 +00:00
Andree Buschmann
452c5b2197 Unroll mpc's requantization loops for some (minor) speed up.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@25891 a1c6a512-1295-4272-9138-f99709370657
2010-05-07 21:36:55 +00:00
Andree Buschmann
f2b1f8d768 Refactor mpc's requantization to use identical macros for each subframe.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@25887 a1c6a512-1295-4272-9138-f99709370657
2010-05-07 19:58:34 +00:00
Bertrik Sikken
1cd1e66ed3 Clean up libmusepack: move declarations to internal.h and set svn properties.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@25676 a1c6a512-1295-4272-9138-f99709370657
2010-04-18 18:12:32 +00:00
Andree Buschmann
c82cac4c3d Restore musepack's IRAM configuration to pre r25127 for non PP5022/24.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@25133 a1c6a512-1295-4272-9138-f99709370657
2010-03-12 18:36:01 +00:00
Andree Buschmann
263e4d5cbf Minor optimization for musepack codec through moving functions to IRAM on PP processors. Speed ups codec by 0.25 MHzon PP5022.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@25127 a1c6a512-1295-4272-9138-f99709370657
2010-03-11 20:58:38 +00:00
Andree Buschmann
398ec52817 Refacturate IRAM configuration for musepack codec. No functional change.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@25126 a1c6a512-1295-4272-9138-f99709370657
2010-03-11 20:19:31 +00:00
Andree Buschmann
4ad89fdf6d Use more IRAM for musepack on S5L870X CPUs.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@25064 a1c6a512-1295-4272-9138-f99709370657
2010-03-07 21:38:42 +00:00
Andree Buschmann
b3d9578c27 Major change to musepack decoder: Import v1.3.0 (r458 from svn.musepack.net) to rockbox. Several adaptions in the musepack decoder were made to get the library work and perform fast under rockbox on several targets. With this change mpc sv8 is supported, including seek, replay gain and metadata support. The decoding speed is a 1-4% lower than the last implementation. Reason for this is main restructuring in the bitstream demuxer.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@25056 a1c6a512-1295-4272-9138-f99709370657
2010-03-07 19:34:44 +00:00
Andree Buschmann
cc572378fc Submit FS#10974: Speed up mpc-codec through changing the dct32-algorithm. +5% on PP5020, +12% on M5.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@24544 a1c6a512-1295-4272-9138-f99709370657
2010-02-07 14:04:46 +00:00
Andree Buschmann
db4d7a313e 1st part of fix for FS#10637. Correction of mpc header parsing.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@24408 a1c6a512-1295-4272-9138-f99709370657
2010-01-31 11:43:42 +00:00
Bertrik Sikken
6fef14bd16 Set svn:keywords property on .c and .h files that didn't already have it. Correct svn:executable property on some files.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@24165 a1c6a512-1295-4272-9138-f99709370657
2010-01-03 14:28:09 +00:00
Jens Arnold
02c031709c * ARM asm DSP and codec/plugin functions: Use r12 scratch register properly
* Fix saving another unused reg in dsp code
* Use less regs in the generic ARM mpegplayer adding idct pure DC case
* Fix ARMv6 mpegplayer adding idct using an unsaved register in pure DC case


git-svn-id: svn://svn.rockbox.org/rockbox/trunk@21803 a1c6a512-1295-4272-9138-f99709370657
2009-07-12 13:14:35 +00:00
Bertrik Sikken
4f87abf90a Cosmetic fix: remove duplicate semicolons
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@20198 a1c6a512-1295-4272-9138-f99709370657
2009-03-04 18:15:06 +00:00
Jens Arnold
6421f94c0d Silence warning from 'ar' if the archive had to be created.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@20151 a1c6a512-1295-4272-9138-f99709370657
2009-03-01 09:04:15 +00:00
Daniel Stenberg
2e6d604bb6 Stop hiding errors by redirecting stderr to /dev/null. If we really need to do re-introduce somewhere we should rather make it dependent on the V variable so that make V=1 would still show the error and only "normal" builds would hide it.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@20090 a1c6a512-1295-4272-9138-f99709370657
2009-02-23 08:45:16 +00:00