GCC 4.9.4 was already used for MIPS and all hosted targets; this enables
it across the board for everything (ie m68k and arm native)
Other changes:
* Use '-Os' as the default optiomization for all targets
(was only disabled for arm native)
* Enable -funit-at-a-time and -Wextra
* Drop all obsolete toolchain patches
* Update ARM multilib/exception patch
* Bump toolchain libs (gmp, mpfr, mpc) to recommended versions, and
add 'isl' to enable better optimization & vectorization opportunities.
(Will revisit optimization for the codecs and plugins at a later date)
Confirmed working:
* armv4t (ipodmini2g and many other PP502x targets)
* arm >= v5 (sansaclipplus, ipod6g, ipodnano2g, sansafuzeplus)
* m68k (ihp100)
Change-Id: If9ed405ae0f289d9adea46d4cf46bfefc2f4250d
This codec requires floating point.
Original author: Peter Sovietov
Ported to Rockbox: Roman Skylarov
Further integration and bugfixes: Solomon Peachy
Change-Id: I781ecd3592dfcdbbc694063334350342534f1d6c
'swcodec' is now always set (and recording_swcodec for recording-capable
units) in feature.txt so the manual and language strings don't need to
all be fixed up.
Change-Id: Ib2c9d5d157af8d33653e2d4b4a12881b9aa6ddb0
By moving three structures to the heap. None are in the hot decode
loop, instead having to do with file sync / header state.
Has neglible impact on performance (within measurement noise) on Clip+,
Rocker, and Xduoo X3.
On PP5022 (ipodmini2g) performance drops from 138.66% to 138.22% realtime.
(0.3%)
Unknown effect on Coldfire which lacks D$.
Stack savings are pretty significant especially on lowmem devices.
Change-Id: Ic8a1e93062ff5a46230e824134032053c4e1986d
apparently we should be doing this anyway
mark4o> The packets overlap and may reuse state set by other recent packets,
so if you seek to a different position,
resetting the state helps to ensure that the subsequent
packets won't use the state set by the unrelated packets
that were processed before the seek.
remove stack bump WORKAROUND_FS13060
Change-Id: I1c14e23b1721a360b91e3e55202c1557aef0fcc6
opus requires the comment header to be a valid file our codec attemps to skip the comment data
in order to reduce the ram allocated originally it caused files with large album art to skip
the beginning of tracks my first attempt at fixing this then caused files with low bitrates
to do the same while fixing files with large album art
This patch should fix both although the initial start might be a bit slower but
this shouldn't cause too much of an issue
Change-Id: Ia1c3561347894cc45f24bb2659436914f8f03b43
knocks off about .5 second from decode time not a big change but might help a bit on
devices that barely achieve realtime
Change-Id: If6e822b7273613c9449c102ce7dd3543bf975d37
ogg_sync_reset() causes issues on the partial page boundary
due to the next page (already in buffer) being discarded
instead seek next page boundary past complete page
Change-Id: Ic05f188f489b015693d663f131e09cd92ad37ff7
Files with extension "aac" in ADTS or ADIF format are now playable.
Full credit goes to Igor Poretsky.
Change-Id: I413b34e15e5242fea60d3461966ae0984080f530
* More tolerance to the file format variations.
* AC3 coded files in realaudio format are now playable
Full credit to Igor Poretsky
Change-Id: Id24e94bc00623e89fb8c80403efa92f69ab1e5d7
In particular, this solves seeking glitches seen in ~6 hr mp3 files.
(Patch taken from Igor Poretsky's tree)
Change-Id: Id65b6726146b6d2d1a223e90b88e401d1b2d597a
On Classic, IRAM1 (second 128Kb of a total of 256KB available IRAM) is
slower than DRAM. Codecs that actually are using regions of IRAM1 runs
faster when DRAM is used, so IRAM1 is disabled and only IRAM0 remains
enabled: 48KB for core and 80KB for codecs/plugins.
The next test_codec results shows how decode time is decreased:
file boosted unboosted
*.ra ~1.5% ~0.5%
*.mpc ~21% ~4.5%
*.ogg ~0.5% ~0%
nero_he*.m4a ~8% ~1%
nero*.m4a ~25% ~7%
wmapro*.wma ~4.5% ~0%
wma*.wma ~25% ~7%
In addition there is a small power save when IRAM1 HW is disabled.
Change-Id: I102adee11458e82037f23076d5d5956e23235de8
Just use long so the compiler potentially doesn't complain about
use of other values not in the enum. It's also the type used
around the system for event ids.
Increase min codec API version.
No functional changes.
Change-Id: If4419b42912f5e4ef673adcdeb69313e503f94cc
On Windows 64-bit, the size of long is 32-bit, thus any pointer to long cast is
not valid. In any case, one should use intptr_t and ptrdiff_t when casting
to integers. This commit attempts to fix all instances reported by GCC.
When relevant, I replaced code by the macros PTR_ADD, ALIGN_UP from system.h
Change-Id: I2273b0e8465d3c4689824717ed5afa5ed238a2dc
The mingw linker uses strlen() in some cases, and codeclib.c redefines it, that
leads to mingw runtime init to call into our strlen() and then ci->strlen() which
of course crashes. Apply the same fix as for malloc and friends: rename the symbol.
The codeclib.h include is necessary for normal builds.
Change-Id: Ifa85901a3e4a31cc0e10b4b905df348a239d5c99
In DEBUG build, the codec API struct is consider with DEBUG flag in apps/
but without DEBUG flah in rbcodecs/, leading to unmatched structure and horrible
crashes in some cases (mostly encoders). I have no idea why the codecs Makefile
removes the DEBUG flag (maybe for performance reasons?) but it cannot be right.
Change-Id: Idb2c5f66741408ec2939624590fc39c4cf69fc2b
The codec wasn't calling ci->set_offset() while decoding; as a result,
the saved offset in ci.id3->offset was only updated at the start of the
file and when seeking.
To reproduce the problem in the simulator or on a real device:
- Start playing an Opus file.
- Let it play until 15s, then turn the player off.
- Turn back on and resume playback. This'll resume correctly from 15s
(using time-based resume, I think, as the offset was 0?).
- Let it play until 30s, then turn the player off again.
- Turn back on and resume playback. This'll resume from 15s, based on
the initial position from last time, when it should resume from 30s.
I believe this will also fix FS#12799 ("Resuming opus file from bookmark
is not working correctly").
Change-Id: Iba67368e0029c968ef802693767e0722719bc38b
ffmpeg_bitstream.c is included in libcodec, so there doesn't seem to
be any reason for individual codecs to also compile it (and clobber
any previous copy while they're at it, leading to broken builds)
Change-Id: I2bedc277ab109f44a6e8feb3d12ed01a720e00a6
Fixes a buffer overflow present when MP3 is encoded at 32000 Hz sample
rate, affected bitrates are 320 and 256 kbps.
Change-Id: I7634e70409be9d675d47be316a42630dd3147636
- Leave original ptr untouched if allocation fails
(bail out early)
- Behave like malloc() in case ptr is NULL
Change-Id: Ib854ca19bd0e069999b7780d2d9a533ece705add
This file revealed several problems with our ASF parser:
1) The packet count in the ASF was actually a 64 bit value,
leading to overflow in very long files.
2) Seeking blindly trusted the bitrate listed in the ASF header
rather than computing it from the packet size and number of packets.
Fix these problems and fix a few minor issues.
Change-Id: Ie0f68734e6423e837757528ddb155f3bdcc979f3
Forgot to (void) an unused parameter when priorityless.
usb-drv-rl27xx.c was using a compound init to initialize a semaphore
but the structure changed so that it is no longer correct. Use
designated initializers to avoid having to complete all fields.
Forgot to break compatibility on all plugins and codecs since the
kernel objects are now different. Take care of that too and do the
sort thing.
Change-Id: Ie2ab8da152d40be0c69dc573ced8d697d94b0674
Speeds up decoding of the 64 kbps test file by 2.59 MHz and the
128 kbps test file by 4.31 MHz on H300 (cf). Decoding the same
files on c200 is sped up by 0.33 MHz and 0.55 MHz respectively.
Change-Id: I0f9f9ef6a7293581cf45e3201b33c65504c95c81
The recent merge of upstream changed the fft to use C_MUL which
wasn't implemented in asm for coldfire.
Speeds up decoding 64 kbps test file by 2.68 MHz and 128 kbps
test file by 2.80 MHz on H300.
Change-Id: I8b61fc0f9568d6350431e311a12e44fe4f60f72e
Sync to commit bb4b6885a139644cf3ac14e7deda9f633ec2d93c
This brings in a bunch of optimizations to decode speed
and memory usage. Allocations are switched from using
the pseudostack to using the real stack. Enabled hacks
to reduce stack usage.
This should fix crashes on sansa clip, although some
files will not play due to failing allocations in the
codec buffer.
Speeds up decoding of the following test files:
H300 (cf) C200 (arm7tdmi) ipod classic (arm9e)
16 kbps (silk) 14.28 MHz 4.00 MHz 2.61 MHz
64 kbps (celt) 4.09 MHz 8.08 MHz 6.24 MHz
128 kbps (celt) 1.93 MHz 8.83 MHz 6.53 MHz
Change-Id: I851733a8a5824b61feb363a173091bc7e6629b58
Implicit promotion of integer literals to unsigned long introduced a subtle bug
on 64-bit systems due to weird sign extensions (leads to audible glitches in a
few files). The table is originally designed for unsigned 32bit integers, and
it works with those so use them. As a consequence the lookup table size is
halved as well.
Change-Id: I35d878d6df03300387f0e403e0f3c3bdc73eea00
This complements offset-based resume and playback start funcionality.
The implementation is global on both HWCODEC and SWCODEC.
Basically, if either the specified elapsed or offset are non-zero,
it indicates a mid-track resume.
To resume by time only, set elapsed to nonzero and offset to zero.
To resume by offset only, set offset to nonzero and elapsed to zero.
Which one the codec uses and which has priority is up to the codec;
however, using an elapsed time covers more cases:
* Codecs not able to use an offset such as VGM or other atomic
formats
* Starting playback at a nonzero elapsed time from a source that
contains no offset, such as a cuesheet
The change re-versions pretty much everything from tagcache to nvram.
Change-Id: Ic7aebb24e99a03ae99585c5e236eba960d163f38
Reviewed-on: http://gerrit.rockbox.org/516
Reviewed-by: Michael Sevakis <jethead71@rockbox.org>
Tested: Michael Sevakis <jethead71@rockbox.org>
