OSDN Git Service

android-x86/frameworks-av.git
11 years agoFix master build.
Ben Murdoch [Wed, 28 Nov 2012 13:58:29 +0000 (13:58 +0000)]
Fix master build.

Change-Id: Ia362f74d8cd7df76292473c26c112dffe190c599

11 years agoam 5d7b2778: resolved conflicts for merge of 41829f30 to jb-mr1-dev-plus-aosp
Glenn Kasten [Wed, 28 Nov 2012 02:43:32 +0000 (18:43 -0800)]
am 5d7b2778: resolved conflicts for merge of 41829f30 to jb-mr1-dev-plus-aosp

* commit '5d7b2778d0e9849fa601d722ec2efcee7d032d4f':
  New VHQ resampler

11 years agoam b8fa1240: am aa9e3e01: Camera: Play shutter sound iff enableShutterSound(true...
Igor Murashkin [Wed, 28 Nov 2012 01:34:23 +0000 (17:34 -0800)]
am b8fa1240: am aa9e3e01: Camera: Play shutter sound iff enableShutterSound(true) && ShutterCallback !null

* commit 'b8fa1240eedeb487a7ee0cf7d60e17bed9b25cf4':
  Camera: Play shutter sound iff enableShutterSound(true) && ShutterCallback !null

11 years agoresolved conflicts for merge of 41829f30 to jb-mr1-dev-plus-aosp
Glenn Kasten [Tue, 27 Nov 2012 23:05:48 +0000 (15:05 -0800)]
resolved conflicts for merge of 41829f30 to jb-mr1-dev-plus-aosp

Change-Id: I952d333c2e88b6b28663793046a136822d1b6838

11 years agoMerge "New VHQ resampler" into jb-mr1.1-dev
Glenn Kasten [Tue, 27 Nov 2012 22:51:19 +0000 (14:51 -0800)]
Merge "New VHQ resampler" into jb-mr1.1-dev

11 years agoNew VHQ resampler
Glenn Kasten [Mon, 19 Nov 2012 17:00:47 +0000 (09:00 -0800)]
New VHQ resampler

Squashed commit of the following:

commit 12b6952da9f25e94d06dd7185bce255924e7e791
Author: Mathias Agopian <mathias@google.com>
Date:   Mon Nov 19 15:27:26 2012 -0800

    fix a typo in SINC resampler that prevented tracks to be mixed

    we were always erasing the current mix instead of mixing into it.

    Change-Id: Ib229245f9e5a0d384f1727640a59e9f0469211a2

commit 0019ce082df430278f14ab922e900ce33b64897d
Author: Dave Bort <dbort@google.com>
Date:   Tue Nov 13 01:30:32 2007 -0800

    Rename "TARGET" to "MODULE" in the build system.

    Part one of the grand renaming.

    API_CHANGE: Third parties may need to update their makefiles.
    Any variables with "LOCAL" and "TARGET" in their names
    should now use "MODULE" instead of "TARGET"; e.g., LOCAL_MODULE,
    LOCAL_MODULE_TAGS.

    PRESUBMIT=passed
    OCL=39840

    Change-Id: Ica9a7937d3d9552ab84db46ac6eea8a290e404fe
Signed-off-by: Glenn Kasten <gkasten@google.com>
commit f01adc0cef0e39e75c76d9195ac26a94cac0a100
Author: Glenn Kasten <gkasten@google.com>
Date:   Wed Nov 14 08:32:08 2012 -0800

    Fix build warnings

    Change-Id: Ic43bcca166a529a6431711b05a7fa21849b6a38b

commit 9bb031a565c753a03d9c9397edea318947d80528
Author: Mathias Agopian <mathias@google.com>
Date:   Sat Nov 10 04:44:30 2012 -0800

    more optimizations...

    calculate the offsets from the phase differently, this happens
    to reduce the register pressure in the main loop, which in turns
    allows the compiler to generate much better code (doesn't need
    to spill a lot of stuff on the stack).

    this gives another 15% performance increase

    Change-Id: I2ce3479dd48b9e6941adb80e6d443d6e14d64d96

commit 5a951598f31217b8cd2babd0720c9608ee17291a
Author: Mathias Agopian <mathias@google.com>
Date:   Sat Nov 10 03:26:39 2012 -0800

    refactor code to improve neon code

    we want to make sure we don't transfer data from the
    neon unit to the arm register file, as this can be quite
    slow. instead we do all the calculation on the neon side
    and write the result directly to main memory.

    Change-Id: Ibb56664d3ab03098ae2798b75e2b6927ac900187

commit b381ee9e83bc9fd18986e79c7809841514ed590e
Author: Mathias Agopian <mathias@google.com>
Date:   Sun Nov 4 15:16:13 2012 -0800

    NEON optimized SINC resampler

    this currently gives us a 60% to 80% boost depending
    on the quality level selected.

    Change-Id: I7db385007e811ed7bffe5fd3403b44e300894f5b

commit bea077354210242ea193a50b0dbab0fedab25df3
Author: Mathias Agopian <mathias@google.com>
Date:   Mon Nov 5 01:51:37 2012 -0800

    minor cleanups

    Change-Id: Ia12ee4fb59e90221761bec85e6450db29197591f

commit 8f4ed7decbe161a5ff38200b218f5216d80aba46
Author: Mathias Agopian <mathias@google.com>
Date:   Sun Nov 4 18:49:14 2012 -0800

    improve resample test

    - handle stereo input
    - input file can now be ommited, in this case
      a linear chirp will be used automatically
    - better usage information

    Change-Id: I5d62a6c26a9054a1c1a517a065b4df5a2cdcda22

commit 5fcd634ea6cb4df27c495abe20f5f9b8ff55d128
Author: Mathias Agopian <mathias@google.com>
Date:   Sun Nov 4 02:03:49 2012 -0800

    change how we store the FIR coefficients

    The coefficient table is now transposed and shows
    much better its polyphase nature: we now have a FIR
    per line, each line corresponding to a phase.

    This doesn't change at all the results produced by
    the filter, but allows us to make slightly better
    use of the data cache and improves performance a bit
    (although not as much as I thought it would).

    The main benefit is that it is the first step
    before we can make much larger optimizations
    (like using NEON).

    Change-Id: Iebf7695825dcbd41f25861efcaefbaa3365ecb43

commit d652231abf4c7e2ea1fc89caae730cec1f7259a1
Author: Mathias Agopian <mathias@google.com>
Date:   Sat Nov 3 23:37:53 2012 -0700

    improve SINC resampler performance

    The improvement is about 60% by just tweaking a few
    things to help the compiler generate better code.
    It turns out that inlining too much stuff manually was hurting us.

    Change-Id: I8068f0f75051f95ac600e50ce552572dd1e8c304

commit 9dc68ef5b94c700c4ee68790e8cbb334c90a538d
Author: Mathias Agopian <mathias@google.com>
Date:   Thu Nov 1 21:03:46 2012 -0700

    new coefficients for the vhq resampler

    previous coefficients were provided by a 3rd party and didn't have a
    way to re-generate them. we're now using the 'fir' utility.

    the performance of the filter is virtually identical, except for
    the down-sampling case which seems slightly better now:
       It looks like both the previous and new coefficients are generating
    some sort of clipping for full-scale signals in the down-sampling case
    (although the new ones seem better), the reason for that is
    unknown (see bug: 7453062)

    Also updated the HQ coefficients for the down-samplers, previous ones
    were a little bit too conservative -- the new ones push the cut-off
    frequency up by about 1 KHz.

    Change-Id: I54a827b5c707c7cc41268ed01283758dce1d7647

commit 38e0b8560a6fc1b7124e22e0e09a84a285182f8e
Author: Mathias Agopian <mathias@google.com>
Date:   Tue Oct 30 13:51:44 2012 -0700

    fix SINC resampler on non ARM architectures

    make sure the C version of the code generates the same
    output than the ARM assemply version.

    Change-Id: Ide218785c35d02598b2d7278e646b1b178148698

commit a1878128b182696ba508569b4d211d0dfae92463
Author: Mathias Agopian <mathias@google.com>
Date:   Tue Oct 30 12:49:07 2012 -0700

    fix another issue with generating FIR coefficients

    the impulse response of a low-pass is 2*f*sinc(2*pi*f*k), we were
    missing the 2*f scale factor. This explains why we were seeing
    clipping and had to manually scale the filter down.

    Change-Id: I86d0bb82ecdd99681c8ba5a8112a8257bf6f0186

commit 1a0fb993430acc9f601e6c538305bc407c20ac5d
Author: Mathias Agopian <mathias@google.com>
Date:   Mon Oct 29 17:13:20 2012 -0700

    fir a typo that caused up-sampling coefficiens to be wrong

    up-sample coefficient were generated with a cut-off frequency of 24KHz
    intead of ~20KHz, which caused more aliasing in the audible band.

    also increased the attenuation to 1.3 dB on both up and down
    sampling coefficient to avoid clipping.

    Change-Id: Ie8aeecf1429190541b656810c6716b6aae5ece2e

commit 9520ad6862bd682ad075a9d9e3e94ada9f6e58b6
Author: Mathias Agopian <mathias@google.com>
Date:   Mon Oct 29 17:13:16 2012 -0700

    test-resample: clip instead of overflowing

    Change-Id: I550e5a59e51c11e1095ca338222b094f92b96878

commit ba36656300f250f7f1fdeb75149749344260e6cb
Author: Mathias Agopian <mathias@google.com>
Date:   Sun Oct 21 01:01:38 2012 -0700

    a test app for the resamplers

    Change-Id: I66852d90d384f1d9e77b51ad1a1ebdbaf61d0607

commit 056a08b9bfd33cf27228c992adc8293a56b01be8
Author: Mathias Agopian <mathias@google.com>
Date:   Fri Oct 26 14:11:01 2012 -0700

    reenable the cubic resampler

    cubic resampler was disabled because it hadn't been qualified,
    however after I did some tests, it does improve significantly
    the sound quality over the order-1 resampler, even if it is
    still quite bad.

    also HIGH_QUALITY resampler was partially disabled, it's now
    fully enabled. It's a big improvement over the cubic resampler
    in terms of aliasing noise (it's not as good in the pass-band).

    Change-Id: I70e3658c255896588642697be9eb594ff4ec0f8b

commit 8c0241d3ff50ae85167f69b3bd369244894cfa44
Author: Mathias Agopian <mathias@google.com>
Date:   Fri Oct 26 13:48:42 2012 -0700

    improve SINC resampler coefficients

    - we increase the interpolation precision from 4 to 7 bits
    this doesn't increase CPU power required, it only increases the
    size of the filter table but significantly reduces the noise
    introduced by the quantization of the impulse response.

    - the parameters of the filter are set such that aliasing is
    rejected at 80 dB below 20 KHz. Because we don't use a lot of
    coefficient (to save compute power), there are quite a bit of
    attenuation in the pass-band: starting at 9KHz for the
    down-sampler (48 to 44.1), and starting at 13 KHz for the
    up-sampler (44.1 to 48) -- the transition band is about 15 KHz.

    Change-Id: I855548d2aab8a0fb0d2a2da3a364b6842d7d3838

commit 69e7dab2192adc1f780464146810629ebd01b145
Author: Pixelflinger <mathias.agopian@gmail.com>
Date:   Thu Oct 25 19:43:49 2012 -0700

    improve fir tool: cleanup, better default, bug fixes

    - all parameters can be changed on the command-line
    - added float output
    - added debug option
    - added an option to generate a polyphase filter coefficients
    - added an attenuation option in dBFS
    - added a lot of comments and references
    - fixed kaiser window parameter

    also the default should generate a filter with 80 dB rejection
    (of the 24 KHz aliasing) above 20 KHz and a 15 KHz transition
    band around ~20 KHz (for 48 KHz sampling rate).
    It's not very good but corresponds to the current code.

commit 8347499d105a50257c18e9dac652e750b06428b1
Author: Glenn Kasten <gkasten@google.com>
Date:   Mon Oct 22 17:09:27 2012 -0700

    Increase allowed number of VHQ resamplers to 3

    Bug: 7378660
    Change-Id: I69e33ca2eb4bb9bd38e2c63df62cd1130d68baf6

commit f91cf3cad7f5c4d52614271c89ab468741c5d24c
Author: Mathias Agopian <mathias@google.com>
Date:   Sun Oct 21 03:04:05 2012 -0700

    Fix a typo that caused the high quality resampler to produce garbage

    the problem is that if libaudio_resampler is present, it is those
    coefficients that will always be selected, but the correct
    meta-data.

    Bug: 7385994
    Change-Id: Ieebeb37b4dfb62a1a051bc29fae2ce056dbc6621

commit e158a9e4262a174c59469a205658bc3ca4078234
Author: Dan Bornstein <danfuzz@google.com>
Date:   Fri Oct 3 10:34:57 2008 -0700

    Manually merge change #111620 from tc3 to mainline, to keep the

    automerger from choking on it.

    p4 sync
    p4 integrate -r -b android_to_tc3 //...@111620,111620
    p4 resolve -a
    p4 resolve     # resolve a couple merge travesties

    PRESUBMIT=passed
    BUG=1399648
    TBR=edheyl
    OCL=111902

    Change-Id: I854b01553dd92bbf9c864f5a9bd51a3d665f0ac2
Signed-off-by: Glenn Kasten <gkasten@google.com>
commit b9f3c26032be7a6ea01a10d93d94826f449e68ab
Author: Dave Bort <dbort@google.com>
Date:   Fri Jan 18 14:51:05 2008 -0800

    Rename "Makefile" to "Android.mk" throughout the tree.

    For <http://b/issue?id=960416>.

    I've tested this as much as I can, but 1500 open files =
    easy to mess things up.  Please let me know if there's
    a problem rather than rolling back this change.

    PRESUBMIT=passed
    BUG=960416
    TBR=joeo
    OCL=46537

    Change-Id: I5a404caf0f398a7afa7ae7abaf2f2a1c6ab490eb
Signed-off-by: Glenn Kasten <gkasten@google.com>
commit 0c22a9a44c4103483fba1d944acf1354c5eb1617
Author: Mathias Agopian <mathias@google.com>
Date:   Mon Oct 29 23:44:25 2007 -0700

    Tweak the SINC resampler parameters and double the performance. It's using about 10% CPU in the worse case now.

    Change-Id: I50ac7e6c6702a427fa36ab6d976c507155057507
Signed-off-by: Glenn Kasten <gkasten@google.com>
commit b85e41487983ad085b859acf8251e7e54480308a
Author: Mathias Agopian <mathias@google.com>
Date:   Mon Oct 29 04:34:36 2007 -0700

    A sinc resampler for Audioflinger. It's not enabled yet, but fully functional and apparently working. It need more "quality" tests. In the 48->44 KHz, it takes about 25% of the CPU time.

    Change-Id: I80eb5185e13ebdb907e0b85c49ba1272c23d60ec
Signed-off-by: Glenn Kasten <gkasten@google.com>
commit ba3949ef17cac2ba71cc3096c413782a49c922e5
Author: Mathias Agopian <mathias@google.com>
Date:   Thu Aug 23 21:01:28 2007 -0700

    fix a few small typos in the FIR computation

    Change-Id: I6e56b514fe520f30f7487f85c64ea5d2a7c19b40
Signed-off-by: Glenn Kasten <gkasten@google.com>
commit 7474bfa7de2604021963794dddfe44985648db6a
Author: Mathias Agopian <mathias@google.com>
Date:   Thu Aug 23 03:16:02 2007 -0700

    This is a tool to compute the the reconstruction filter coefficients for a sinc audio resampler.

    Change-Id: I99be2505139b8e0e7647200e1647509d4f7e6067
Signed-off-by: Glenn Kasten <gkasten@google.com>
Bug: 7577965
Change-Id: I2c84a9283a1668723bad83e1a119c849c88c3e6b

11 years agoam aa9e3e01: Camera: Play shutter sound iff enableShutterSound(true) && ShutterCallba...
Igor Murashkin [Tue, 27 Nov 2012 19:39:22 +0000 (11:39 -0800)]
am aa9e3e01: Camera: Play shutter sound iff enableShutterSound(true) && ShutterCallback !null

* commit 'aa9e3e01b86bd9bfb5ac36c0f360d5fe478cbb2d':
  Camera: Play shutter sound iff enableShutterSound(true) && ShutterCallback !null

11 years agoam 8b5985d9: am 5768fa03: Merge "Crash in android::MyHandler::parsePlayResponse"
Andreas Huber [Tue, 27 Nov 2012 17:41:12 +0000 (09:41 -0800)]
am 8b5985d9: am 5768fa03: Merge "Crash in android::MyHandler::parsePlayResponse"

* commit '8b5985d99becc2d5126b8a26afb6f5798b560007':
  Crash in android::MyHandler::parsePlayResponse

11 years agoam 9abbd0fb: am 3eb46d17: Merge "Fix of uninitialized mIsDrm variable."
Andreas Huber [Tue, 27 Nov 2012 17:41:11 +0000 (09:41 -0800)]
am 9abbd0fb: am 3eb46d17: Merge "Fix of uninitialized mIsDrm variable."

* commit '9abbd0fb822aa0076f6de03f2092db47760b924d':
  Fix of uninitialized mIsDrm variable.

11 years agoam 5768fa03: Merge "Crash in android::MyHandler::parsePlayResponse"
Andreas Huber [Tue, 27 Nov 2012 17:39:12 +0000 (09:39 -0800)]
am 5768fa03: Merge "Crash in android::MyHandler::parsePlayResponse"

* commit '5768fa034ede834656697d3612c525595ff85ef9':
  Crash in android::MyHandler::parsePlayResponse

11 years agoam 3eb46d17: Merge "Fix of uninitialized mIsDrm variable."
Andreas Huber [Tue, 27 Nov 2012 17:39:11 +0000 (09:39 -0800)]
am 3eb46d17: Merge "Fix of uninitialized mIsDrm variable."

* commit '3eb46d179b1f62cde21077fde466925d4c5c79ad':
  Fix of uninitialized mIsDrm variable.

11 years agoMerge "Crash in android::MyHandler::parsePlayResponse"
Andreas Huber [Tue, 27 Nov 2012 16:50:52 +0000 (08:50 -0800)]
Merge "Crash in android::MyHandler::parsePlayResponse"

11 years agoMerge "Fix of uninitialized mIsDrm variable."
Andreas Huber [Tue, 27 Nov 2012 16:49:34 +0000 (08:49 -0800)]
Merge "Fix of uninitialized mIsDrm variable."

11 years agoCrash in android::MyHandler::parsePlayResponse
Patric Frederiksen [Mon, 26 Sep 2011 08:51:35 +0000 (10:51 +0200)]
Crash in android::MyHandler::parsePlayResponse

This fix handles problems with several asynchronous calls
within streaming. This case is when the phone has sent a
request to the server and while the response is being sent
back by the server the request is aborted by the user.
The fix is an if case that checks if we have aborted while
waiting for a response from the server. If we have aborted
we should ignore the late response instead of continuing.

Change-Id: I1264bb992f6abcaee1f10a89479e08b54a95e3c2

11 years agoFix of uninitialized mIsDrm variable.
Henrik B Andersson [Wed, 31 Oct 2012 12:02:47 +0000 (13:02 +0100)]
Fix of uninitialized mIsDrm variable.

The mIsDrm is a bool that isn't initialized.
This causes it to be true in most default cases.

Change-Id: I41b534514bf6a3ca88a9f0994b814d55fcd7453b

11 years agoCamera: Play shutter sound iff enableShutterSound(true) && ShutterCallback !null
Igor Murashkin [Mon, 26 Nov 2012 18:50:55 +0000 (10:50 -0800)]
Camera: Play shutter sound iff enableShutterSound(true) && ShutterCallback !null

Bug: 7564718
Change-Id: Ie7821cdee57966d88af048759578439a3e6ecb2e

11 years agoam 5bc5bf39: am 79fd6853: Merge "MediaCodec: Add a method for getting the component...
Andreas Huber [Mon, 26 Nov 2012 22:39:20 +0000 (14:39 -0800)]
am 5bc5bf39: am 79fd6853: Merge "MediaCodec: Add a method for getting the component name"

* commit '5bc5bf39120ae27ef46a8b13f85bf44ea19c7d5e':
  MediaCodec: Add a method for getting the component name

11 years agoam 79fd6853: Merge "MediaCodec: Add a method for getting the component name"
Andreas Huber [Mon, 26 Nov 2012 22:37:22 +0000 (14:37 -0800)]
am 79fd6853: Merge "MediaCodec: Add a method for getting the component name"

* commit '79fd685323e34e0fde22d17fd6848d33f171f4ae':
  MediaCodec: Add a method for getting the component name

11 years agoMerge "MediaCodec: Add a method for getting the component name"
Andreas Huber [Mon, 26 Nov 2012 21:48:41 +0000 (13:48 -0800)]
Merge "MediaCodec: Add a method for getting the component name"

11 years agoam 21006fa5: (-s ours) am 5865ddf7: Merge "AudioTrack::dump null mCblk check test"
Marco Nelissen [Mon, 26 Nov 2012 19:20:33 +0000 (11:20 -0800)]
am 21006fa5: (-s ours) am 5865ddf7: Merge "AudioTrack::dump null mCblk check test"

* commit '21006fa5fa180d1eb1513a5ae297211a24312021':
  AudioTrack::dump null mCblk check test

11 years agoam 5865ddf7: Merge "AudioTrack::dump null mCblk check test"
Marco Nelissen [Mon, 26 Nov 2012 18:00:09 +0000 (10:00 -0800)]
am 5865ddf7: Merge "AudioTrack::dump null mCblk check test"

* commit '5865ddf769d368d714af630aba18392ea1387bc7':
  AudioTrack::dump null mCblk check test

11 years agoMerge "AudioTrack::dump null mCblk check test"
Marco Nelissen [Mon, 26 Nov 2012 16:59:50 +0000 (08:59 -0800)]
Merge "AudioTrack::dump null mCblk check test"

11 years agoAudioTrack::dump null mCblk check test
Zbigniew Mazur [Thu, 11 Oct 2012 11:56:41 +0000 (13:56 +0200)]
AudioTrack::dump null mCblk check test

This fix is protecting AudioTrack::dump from SIGSEGEV
when calling dumpsys shell command.

Change-Id: I30d136e195a12d6fdad41e14f557b0ad9e59b8a2

11 years agoam b96c4b4c: am 2da6e4ae: Merge "Avoid memory leaks when handling metadata strings"
Marco Nelissen [Tue, 20 Nov 2012 20:49:52 +0000 (12:49 -0800)]
am b96c4b4c: am 2da6e4ae: Merge "Avoid memory leaks when handling metadata strings"

* commit 'b96c4b4ce8eb90a6bdb162681affe1e50fe6aafc':
  Avoid memory leaks when handling metadata strings

11 years agoam 2da6e4ae: Merge "Avoid memory leaks when handling metadata strings"
Marco Nelissen [Tue, 20 Nov 2012 20:47:56 +0000 (12:47 -0800)]
am 2da6e4ae: Merge "Avoid memory leaks when handling metadata strings"

* commit '2da6e4ae488896df10b22166b0aa0b2cc15492f1':
  Avoid memory leaks when handling metadata strings

11 years agoMerge "Avoid memory leaks when handling metadata strings"
Marco Nelissen [Tue, 20 Nov 2012 20:28:27 +0000 (12:28 -0800)]
Merge "Avoid memory leaks when handling metadata strings"

11 years agoAvoid memory leaks when handling metadata strings
David Williams [Mon, 19 Nov 2012 08:52:16 +0000 (09:52 +0100)]
Avoid memory leaks when handling metadata strings

Don't duplicate strings when retrieveing metadata from media
files. As any requests for metadata strings would pass through
the binder, this would cause the reference to the duplicate string
to be lost, causing a memory leak as the duplicate would not be
freed.

Change-Id: I2593733472b1bb589bc502b2c11080f581015bb5

11 years agoMerge "AudioFlinger files reorganization"
Eric Laurent [Tue, 20 Nov 2012 18:37:47 +0000 (10:37 -0800)]
Merge "AudioFlinger files reorganization"

11 years agoam 14dda623: am 7013209c: Merge "Handle large AVCC chunks"
Andreas Huber [Tue, 20 Nov 2012 17:27:22 +0000 (09:27 -0800)]
am 14dda623: am 7013209c: Merge "Handle large AVCC chunks"

* commit '14dda623c8db5f991b8a22dce4f19f8d8b47fea2':
  Handle large AVCC chunks

11 years agoam 343f9c81: am dca2b5d7: Merge "Changed parsing of trkn and disk from 8 bits to...
Marco Nelissen [Tue, 20 Nov 2012 17:27:21 +0000 (09:27 -0800)]
am 343f9c81: am dca2b5d7: Merge "Changed parsing of trkn and disk from 8 bits to 16 bits"

* commit '343f9c81f293f56e09b1cc1921844ecd3372e435':
  Changed parsing of trkn and disk from 8 bits to 16 bits

11 years agoam 7013209c: Merge "Handle large AVCC chunks"
Andreas Huber [Tue, 20 Nov 2012 17:24:25 +0000 (09:24 -0800)]
am 7013209c: Merge "Handle large AVCC chunks"

* commit '7013209cdf393b3d958ddd46ed50394349378826':
  Handle large AVCC chunks

11 years agoam dca2b5d7: Merge "Changed parsing of trkn and disk from 8 bits to 16 bits"
Marco Nelissen [Tue, 20 Nov 2012 17:24:24 +0000 (09:24 -0800)]
am dca2b5d7: Merge "Changed parsing of trkn and disk from 8 bits to 16 bits"

* commit 'dca2b5d7c29ee06f3c82527dd7264fcc21cac9a6':
  Changed parsing of trkn and disk from 8 bits to 16 bits

11 years agoMerge "Handle large AVCC chunks"
Andreas Huber [Tue, 20 Nov 2012 16:52:44 +0000 (08:52 -0800)]
Merge "Handle large AVCC chunks"

11 years agoMerge "Changed parsing of trkn and disk from 8 bits to 16 bits"
Marco Nelissen [Tue, 20 Nov 2012 16:18:38 +0000 (08:18 -0800)]
Merge "Changed parsing of trkn and disk from 8 bits to 16 bits"

11 years agoAudioFlinger files reorganization
Eric Laurent [Mon, 19 Nov 2012 22:55:58 +0000 (14:55 -0800)]
AudioFlinger files reorganization

Audioflinger.cpp and Audioflinger.h files must be split to
improve readability and maintainability.

This CL splits the files as follows:

AudioFlinger.cpp split into:
- AudioFlinger.cpp: implementation of IAudioflinger interface and global methods
- AFThreads.cpp: implementation of ThreadBase, PlaybackThread, MixerThread,
DuplicatingThread, DirectOutputThread and RecordThread.
- AFTracks.cpp: implementation of TrackBase, Track, TimedTrack, OutputTrack,
RecordTrack, TrackHandle and RecordHandle.
- AFEffects.cpp: implementation of EffectModule, EffectChain and EffectHandle.

AudioFlinger.h is modified by inline inclusion of header files containing
the declaration of complex inner classes:
- AFThreads.h: ThreadBase, PlaybackThread, MixerThread, DuplicatingThread,
DirectOutputThread and RecordThread
- AFEffects.h: EffectModule, EffectChain and EffectHandle

AFThreads.h includes the follownig headers inline:
- AFTrackBase.h: TrackBase
- AFPlaybackTracks: Track, TimedTrack, OutputTrack
- AFRecordTracks: RecordTrack

Change-Id: I512ebc3a51813ab7a4afccc9a538b18125165c4c

11 years agofix a typo in SINC resampler that prevented tracks to be mixed
Mathias Agopian [Mon, 19 Nov 2012 23:27:26 +0000 (15:27 -0800)]
fix a typo in SINC resampler that prevented tracks to be mixed

we were always erasing the current mix instead of mixing into it.

Change-Id: Ib229245f9e5a0d384f1727640a59e9f0469211a2

11 years agoMerge "Clean up channel count and channel mask"
Glenn Kasten [Mon, 19 Nov 2012 19:53:52 +0000 (11:53 -0800)]
Merge "Clean up channel count and channel mask"

11 years agoClean up channel count and channel mask
Glenn Kasten [Wed, 14 Nov 2012 20:47:55 +0000 (12:47 -0800)]
Clean up channel count and channel mask

Channel count is uint32_t.
Remove redundant mask parameter to AudioTrack::createTrack_l()
    and AudioRecord::openRecord_l().

Change-Id: I5dc2b18eb609b2c0dc3091994cbaa4628062c17f

11 years agodelete -> free
Marco Nelissen [Mon, 19 Nov 2012 17:49:18 +0000 (09:49 -0800)]
delete -> free

Strings duplicated with strdup() should be free()d, not deleted.

Change-Id: I42bb3df9625bb8d35f80b02d15364b94c36496f8

11 years agoMediaCodec: Add a method for getting the component name
Martin Storsjo [Tue, 25 Sep 2012 08:43:02 +0000 (11:43 +0300)]
MediaCodec: Add a method for getting the component name

If the codec was chosen based on mime type, the caller does
not know what component actually was chosen. This allows
getting essential information (such as supported color formats,
for a video encoder) for this component.

Change-Id: Ie471f40f8104b37d27ced3dba5a54facc6504b1b

11 years agoChanged parsing of trkn and disk from 8 bits to 16 bits
Andreas Lillvik [Wed, 13 Oct 2010 13:37:01 +0000 (15:37 +0200)]
Changed parsing of trkn and disk from 8 bits to 16 bits

The MPEG4Extractor was parsing 8 bits instead of 16 bits when parsing
'trkn' and 'disk'. Also added support for 16 bytes size 'disk'.

Change-Id: I22b4de2ac800881884d5759776cb380917522a87

11 years agoHandle large AVCC chunks
Jan Olof Svensson [Wed, 26 Sep 2012 07:08:11 +0000 (09:08 +0200)]
Handle large AVCC chunks

If enabling seq_scaling_matrix_present_flag = 1 the AVCC chunk can
be larger than the original buffer size. Changed to using ABuffer
instead.

Change-Id: Idacc14b45ea2634c5e608919f3ce567f23363135

11 years agoMerge ""if" statements use curly braces per media style"
Glenn Kasten [Fri, 16 Nov 2012 23:47:51 +0000 (15:47 -0800)]
Merge ""if" statements use curly braces per media style"

11 years agoMerge "Fix time vs. bytes units bug in getRenderPosition"
Glenn Kasten [Fri, 16 Nov 2012 23:38:29 +0000 (15:38 -0800)]
Merge "Fix time vs. bytes units bug in getRenderPosition"

11 years agoMerge "Don't use control block frame count after create"
Glenn Kasten [Fri, 16 Nov 2012 23:23:42 +0000 (15:23 -0800)]
Merge "Don't use control block frame count after create"

11 years agoMerge "Don't explicitly log tid"
Glenn Kasten [Fri, 16 Nov 2012 23:03:51 +0000 (15:03 -0800)]
Merge "Don't explicitly log tid"

11 years agoDon't use control block frame count after create
Glenn Kasten [Wed, 14 Nov 2012 21:42:25 +0000 (13:42 -0800)]
Don't use control block frame count after create

This is part of a series to clean up the control block.

Change-Id: I7f4cb05aef63053f8e2ab05b286d302260ef4758

11 years agoam abae71d3: am d983364b: Static AudioTrack plays twice initially
Glenn Kasten [Fri, 16 Nov 2012 22:46:42 +0000 (14:46 -0800)]
am abae71d3: am d983364b: Static AudioTrack plays twice initially

* commit 'abae71d37d4860e297de7ee06f49efa5254b90ee':
  Static AudioTrack plays twice initially

11 years agoam d983364b: Static AudioTrack plays twice initially
Glenn Kasten [Fri, 16 Nov 2012 22:44:05 +0000 (14:44 -0800)]
am d983364b: Static AudioTrack plays twice initially

* commit 'd983364b3655a547b55bb11dbe148103198c011d':
  Static AudioTrack plays twice initially

11 years agoFix a crash when the stop might be called due to some error before start in RTSPSource
James Dong [Fri, 16 Nov 2012 22:31:15 +0000 (14:31 -0800)]
Fix a crash when the stop might be called due to some error before start in RTSPSource

o related-to-bug: 7507224

Change-Id: Ic8bfec13097b824ba337a01c9b00c98af2a33f43

11 years ago"if" statements use curly braces per media style
Glenn Kasten [Sat, 23 Jun 2012 00:21:07 +0000 (17:21 -0700)]
"if" statements use curly braces per media style

Change-Id: I130e7849fd1da7a0b7fe56c3c53919d26e3843b8

11 years agoDon't explicitly log tid
Glenn Kasten [Fri, 2 Nov 2012 17:00:06 +0000 (10:00 -0700)]
Don't explicitly log tid

If needed, it can be obtained with adb logcat -v threadtime

Change-Id: I91b3911d20f7bcfc3361db4052db21ff9181f1cf

11 years agoFix time vs. bytes units bug in getRenderPosition
Glenn Kasten [Fri, 16 Nov 2012 20:01:44 +0000 (12:01 -0800)]
Fix time vs. bytes units bug in getRenderPosition

Rename correctLatency since it requires thread to be locked.
Use size_t for byte and frame counts.

Change-Id: I178fdd18bdb823813b9563927bdff8c0d28ca5a5

11 years agoStatic AudioTrack plays twice initially
Glenn Kasten [Thu, 15 Nov 2012 22:13:16 +0000 (14:13 -0800)]
Static AudioTrack plays twice initially

Bug: 7528721
Change-Id: I10bc16a26f33dba6572b730a170cb3bf00e68e30

11 years agoMerge "Only pass the surface to the video decoder."
Andreas Huber [Fri, 16 Nov 2012 21:00:45 +0000 (13:00 -0800)]
Merge "Only pass the surface to the video decoder."

11 years agoresolved conflicts for merge of 205d7249 to master
Marco Nelissen [Fri, 16 Nov 2012 19:30:43 +0000 (11:30 -0800)]
resolved conflicts for merge of 205d7249 to master

Change-Id: I3df408b6e30e0c0b2a19a3336134ce49fb73a8bb

11 years agoOnly pass the surface to the video decoder.
Andreas Huber [Fri, 16 Nov 2012 19:15:44 +0000 (11:15 -0800)]
Only pass the surface to the video decoder.

Change-Id: Ice0cfc0021fdd9fe053be6ee324cbc64226ed122

11 years agoam da33d66e: Merge "Add .mpga to acceptable file name extensions list."
Marco Nelissen [Fri, 16 Nov 2012 17:34:19 +0000 (09:34 -0800)]
am da33d66e: Merge "Add .mpga to acceptable file name extensions list."

* commit 'da33d66e68791d0bfeccebc8253a59467b5ef670':
  Add .mpga to acceptable file name extensions list.

11 years agoMerge "Add .mpga to acceptable file name extensions list."
Marco Nelissen [Fri, 16 Nov 2012 16:59:14 +0000 (08:59 -0800)]
Merge "Add .mpga to acceptable file name extensions list."

11 years agoMerge "Add GSM 6.10 decoder"
Marco Nelissen [Fri, 16 Nov 2012 16:19:30 +0000 (08:19 -0800)]
Merge "Add GSM 6.10 decoder"

11 years agoMerge "Use size_t for frame counts"
Glenn Kasten [Fri, 16 Nov 2012 16:14:40 +0000 (08:14 -0800)]
Merge "Use size_t for frame counts"

11 years agoAdd .mpga to acceptable file name extensions list.
Jan Bjernler [Fri, 16 Nov 2012 15:40:42 +0000 (16:40 +0100)]
Add .mpga to acceptable file name extensions list.

The *.mpga files are playable, but are not correctly scanned.
This is because they are prevented from being scanned in
StagefrightMediaScanner.cpp.
What this fix does is to add the extension to the list of valid
file extensions so that the scanner handles the filetype properly.
We have previously added the .mpga extension to the framework to
make it playable, but not added it so that the scanner scans it.

Change-Id: I02a44d770adb80d38e8bed77d0d76efa1b28a861

11 years agoAdd GSM 6.10 decoder
Marco Nelissen [Thu, 15 Nov 2012 22:31:56 +0000 (14:31 -0800)]
Add GSM 6.10 decoder

Supports Microsoft frame packing only, since that's what the sample
file used.
b/6620569

Change-Id: Ia89d95bcbf0f8dcbaad42148a7401728f60e079d

11 years agoUse size_t for frame counts
Glenn Kasten [Wed, 14 Nov 2012 20:54:39 +0000 (12:54 -0800)]
Use size_t for frame counts

Also fix typo: bufferCount should be frameCount.

Change-Id: Ibed539504db75ef99dc21c8ff1bf2987122063a5

11 years agoMerge "Static AudioTrack plays twice initially"
Glenn Kasten [Thu, 15 Nov 2012 23:05:51 +0000 (15:05 -0800)]
Merge "Static AudioTrack plays twice initially"

11 years agoStatic AudioTrack plays twice initially
Glenn Kasten [Thu, 15 Nov 2012 22:13:16 +0000 (14:13 -0800)]
Static AudioTrack plays twice initially

Bug: 7528721
Change-Id: I10bc16a26f33dba6572b730a170cb3bf00e68e30

11 years agoMerge "wfd sink update."
Andreas Huber [Thu, 15 Nov 2012 20:48:41 +0000 (12:48 -0800)]
Merge "wfd sink update."

11 years agowfd sink update.
Andreas Huber [Thu, 15 Nov 2012 19:16:30 +0000 (11:16 -0800)]
wfd sink update.

Change-Id: Ib4e41ec1524d045699543536acdddc9a243db741

11 years agoThe length information of the chunks making up vorbis codec specific info
Andreas Huber [Wed, 14 Nov 2012 23:24:53 +0000 (15:24 -0800)]
The length information of the chunks making up vorbis codec specific info

are "Xiph-style-lacing encoded" instead of individual bytes.

Change-Id: Ic1274a5bd8f082197bae6831da04002762a920c5
related-to-bug: 7401329

11 years agoMerge "Stagefright command line tool: input file name last"
Jean-Michel Trivi [Thu, 15 Nov 2012 17:32:05 +0000 (09:32 -0800)]
Merge "Stagefright command line tool: input file name last"

11 years agoClean up frame size in AudioTrack and AudioFlinger
Glenn Kasten [Thu, 21 Jun 2012 19:56:37 +0000 (12:56 -0700)]
Clean up frame size in AudioTrack and AudioFlinger

TrackBase::mFrameSize, mChannelMask, and mChannelCount are now const.
Use TrackBase::mFrameSize instead of re-calculating frame size.
AudioFlinger only sees 16-bit PCM format, conversion from 8-bit is
  now entirely on the client side.  Previously a small part of the
  responsibility was on server side also.
size_t is unsigned, so use %u in logs.
Fix theoretical bug where TrackBase constructor was over-allocating space
  for non-linear AudioTrack or 8-bit PCM AudioRecord (probably benign).

Change-Id: I7cbbba0bf4dba29ea751d8af341ab8e5cbbdc206

11 years agoUse uint32_t for sample rate
Glenn Kasten [Wed, 14 Nov 2012 16:44:39 +0000 (08:44 -0800)]
Use uint32_t for sample rate

Change-Id: Ie240b48fb54b08359f69ecd4e5f8bda3d15cbe80

11 years agoMerge "Update audio comments"
Glenn Kasten [Thu, 15 Nov 2012 00:14:31 +0000 (16:14 -0800)]
Merge "Update audio comments"

11 years agoMerge "Remove deprecated AudioSystem methods"
Glenn Kasten [Wed, 14 Nov 2012 23:43:41 +0000 (15:43 -0800)]
Merge "Remove deprecated AudioSystem methods"

11 years agoMerge "Fix build warnings"
Glenn Kasten [Wed, 14 Nov 2012 23:43:18 +0000 (15:43 -0800)]
Merge "Fix build warnings"

11 years agoMerge changes I2ce3479d,Ibb56664d
Mathias Agopian [Wed, 14 Nov 2012 22:51:36 +0000 (14:51 -0800)]
Merge changes I2ce3479d,Ibb56664d

* changes:
  more optimizations...
  refactor code to improve neon code

11 years agoUpdate audio comments
Glenn Kasten [Wed, 7 Nov 2012 22:03:00 +0000 (14:03 -0800)]
Update audio comments

Change-Id: I85d7d2f6381b251db5695202fec75128883a8662

11 years agoam dbb74f4e: am f0937247: Merge "Properly signal an error if codec configuration...
Andreas Huber [Wed, 14 Nov 2012 19:51:31 +0000 (11:51 -0800)]
am dbb74f4e: am f0937247: Merge "Properly signal an error if codec configuration goes wrong." into jb-mr1.1-dev

* commit 'dbb74f4ee1a971da71f26739d870fc9334100499':
  Properly signal an error if codec configuration goes wrong.

11 years agoam 911d5f93: am 0224bf17: Various improvements of wifi display code
Andreas Huber [Wed, 14 Nov 2012 19:51:30 +0000 (11:51 -0800)]
am 911d5f93: am 0224bf17: Various improvements of wifi display code

* commit '911d5f937adbd177c69bd6959603b8a3b776097e':
  Various improvements of wifi display code

11 years agoam f0937247: Merge "Properly signal an error if codec configuration goes wrong."...
Andreas Huber [Wed, 14 Nov 2012 19:49:55 +0000 (11:49 -0800)]
am f0937247: Merge "Properly signal an error if codec configuration goes wrong." into jb-mr1.1-dev

* commit 'f0937247b6d92b7d9457c64e36fe4c10927685ac':
  Properly signal an error if codec configuration goes wrong.

11 years agoam 0224bf17: Various improvements of wifi display code
Andreas Huber [Wed, 14 Nov 2012 19:49:54 +0000 (11:49 -0800)]
am 0224bf17: Various improvements of wifi display code

* commit '0224bf170a3904576bba81593eaab113c5d3a4e7':
  Various improvements of wifi display code

11 years agoMerge "Properly signal an error if codec configuration goes wrong." into jb-mr1.1-dev
Andreas Huber [Wed, 14 Nov 2012 19:45:51 +0000 (11:45 -0800)]
Merge "Properly signal an error if codec configuration goes wrong." into jb-mr1.1-dev

11 years agoFix build warnings
Glenn Kasten [Wed, 14 Nov 2012 16:32:08 +0000 (08:32 -0800)]
Fix build warnings

Change-Id: Ic43bcca166a529a6431711b05a7fa21849b6a38b

11 years agoProperly signal an error if codec configuration goes wrong.
Andreas Huber [Wed, 14 Nov 2012 17:06:33 +0000 (09:06 -0800)]
Properly signal an error if codec configuration goes wrong.

previously any error signaled by setupXXX inside ACodec::configureCodec
would be overwritten with the result of setMinBufferSize at the end
of the function.

Change-Id: Id4beb571ca52ea4646239d0af006e09ce1130268
related-to-bug: 7542181

11 years agoRemove deprecated AudioSystem methods
Glenn Kasten [Tue, 13 Nov 2012 23:01:05 +0000 (15:01 -0800)]
Remove deprecated AudioSystem methods

Change-Id: I952d504e03af9a1d3e1e0aa379c82dfb00197d9f

11 years agoVarious improvements of wifi display code
Andreas Huber [Mon, 12 Nov 2012 21:08:44 +0000 (13:08 -0800)]
Various improvements of wifi display code

- manually prepend SPS/PPS if encoder doesn't support it
- latency improvements
- support for "our" method of optional RTP retransmission
- improvements to the wfd commandline tool for testing
- make it easier to turn on/off suspension of the video pipeline on idle
- fixes an issue where an error during encryption would cause a SEGV
- add HDCP descriptor if necessary

Squashed commit of the following:

commit 1115be0ebb3b885b4f1b7dba56761ca013d0ec4a
Author: Andreas Huber <andih@google.com>
Date:   Fri Nov 9 11:32:23 2012 -0800

    Better shutdown of wfd -l sessions.

    Change-Id: Id898a14ae21efd3b065b00a729830063d39195a7

commit 0e7d106dfe4eb6e2640b0b66c65deaba265f7ff0
Author: Andreas Huber <andih@google.com>
Date:   Thu Nov 8 16:38:55 2012 -0800

    No more sending delay, create rtp packets upfront.

    Change-Id: I809a225f664fdb485c7d9a49a27886601a6a26b2

commit d399e8571b77353d59afb57508dfd2a82c1ef93a
Author: Andreas Huber <andih@google.com>
Date:   Thu Nov 8 14:19:43 2012 -0800

    Restore AudioSource buffer size, factor out TimeSeries, make

    suspending video optional.

    Change-Id: Ifdfe4d447b901e714abf52856b4641d1d55a5d41

commit f8b649f0b8f917d59f4b8a2e8e6d7db61a684a78
Author: Andreas Huber <andih@google.com>
Date:   Thu Nov 8 09:34:06 2012 -0800

    Pull 480 frames at a time from AudioSource/AudioRecord

    Change-Id: I1e215abd329faec3da026631122c0f4c800c0ac4

commit 1bc13452eb35eebbba00f5da93fa86535be5db59
Author: Andreas Huber <andih@google.com>
Date:   Thu Nov 8 08:50:30 2012 -0800

    fixed bitrate traffic simulation

    Change-Id: Ic5efb7cbb0b5d3b4917bc77b8ba73d447249e695

commit 016cdff18e74bdc631a5679e97192645ed095aa2
Author: Andreas Huber <andih@google.com>
Date:   Wed Nov 7 14:00:03 2012 -0800

    resurrected "our" style of retransmission.

    Change-Id: I34d757aba67428437cb39b8293a9651750ad20d9

commit 384cf1a3c8fb4ec410bdf8fa5722c298e6028f3e
Author: Andreas Huber <andih@google.com>
Date:   Tue Nov 6 09:38:55 2012 -0800

    Changes to make wfd work on manta.

    Change-Id: I7a4e00cf16581fe2146edd1b359af195774090e4

commit 9628f24b22b35f28630d99dda3614babf51bc07e
Author: Andreas Huber <andih@google.com>
Date:   Wed Nov 7 09:15:44 2012 -0800

    Patch up rtp timestamps to more accurately measure network jitter.

    Change-Id: I9502a4615575f97f98a215a13131a89a6ae93c6d

commit 7c891a1a24f08bbd50f55be13f7d05f43e807eb8
Author: Andreas Huber <andih@google.com>
Date:   Tue Nov 6 09:37:24 2012 -0800

    Additions to the "wfd" tool to create a local wfd source.

    Change-Id: I99558653a70fdc703f9d13990b3ce1c4d3ae227a

Change-Id: Ia94c63fc390f597014531073485f0cfc53b3418a

11 years agoMerge "Rename TrackBase::mFrameCount to mStepCount"
Glenn Kasten [Tue, 13 Nov 2012 20:25:44 +0000 (12:25 -0800)]
Merge "Rename TrackBase::mFrameCount to mStepCount"

11 years agoRename TrackBase::mFrameCount to mStepCount
Glenn Kasten [Tue, 13 Nov 2012 17:58:55 +0000 (09:58 -0800)]
Rename TrackBase::mFrameCount to mStepCount

This prepares for adding a new field TrackBase::mFrameCount
with a different meaning.

Change-Id: I6bbe2c59f2a882be57caeec2e2e06f439a0e9e83

11 years agoSimplify AudioRecord::restoreTrack_l()
Glenn Kasten [Mon, 12 Nov 2012 23:46:10 +0000 (15:46 -0800)]
Simplify AudioRecord::restoreTrack_l()

Finish removing CBLK_RESTORING and CBLK_RESTORED from control block flags,
and remove constant RESTORE_TIMEOUT_MS.

Also minor cleanup:
 - Cache mCblk in local variable cblk and make cblk allocatable in a register.
 - Use "iMem" for sp<IMemory>.
 - Add missing error log to AudioRecord; it was already in AudioTrack.

This is part of a series to clean up the control block.

Change-Id: Ia5f5ab4763c392bc06a45851b167ddaee29e3455

11 years agoMerge "Move frame size out of the control block"
Glenn Kasten [Tue, 13 Nov 2012 16:48:57 +0000 (08:48 -0800)]
Merge "Move frame size out of the control block"

11 years agoScan .awb files too
Marco Nelissen [Wed, 7 Nov 2012 23:36:59 +0000 (15:36 -0800)]
Scan .awb files too

b/6122599

Change-Id: Ied3e0392939231447f1fc5685ca1fade1e55ce08

11 years agoMove frame size out of the control block
Glenn Kasten [Mon, 12 Nov 2012 15:58:20 +0000 (07:58 -0800)]
Move frame size out of the control block

This is part of a series to clean up the control block.

Change-Id: Ifab1c42ac0f8be704e571b292713cd2250d12a3f

11 years agoStagefright command line tool: input file name last
Jean-Michel Trivi [Thu, 25 Oct 2012 19:07:27 +0000 (12:07 -0700)]
Stagefright command line tool: input file name last

Show in usage that the source file name comes last, and is preceded
 by the options.

Change-Id: I8407fc36c8d19785cb2e6e1f7b7a352a8d86f889

11 years agoFix regression for AudioTrack::write() 8-bit PCM
Glenn Kasten [Mon, 12 Nov 2012 22:32:06 +0000 (14:32 -0800)]
Fix regression for AudioTrack::write() 8-bit PCM

Bug: 7526532
Change-Id: I8ddd1f0e9d035b54401788dcc422591281dcd97a

11 years agomore optimizations...
Mathias Agopian [Sat, 10 Nov 2012 12:44:30 +0000 (04:44 -0800)]
more optimizations...

calculate the offsets from the phase differently, this happens
to reduce the register pressure in the main loop, which in turns
allows the compiler to generate much better code (doesn't need
to spill a lot of stuff on the stack).

this gives another 15% performance increase

Change-Id: I2ce3479dd48b9e6941adb80e6d443d6e14d64d96

11 years agorefactor code to improve neon code
Mathias Agopian [Sat, 10 Nov 2012 11:26:39 +0000 (03:26 -0800)]
refactor code to improve neon code

we want to make sure we don't transfer data from the
neon unit to the arm register file, as this can be quite
slow. instead we do all the calculation on the neon side
and write the result directly to main memory.

Change-Id: Ibb56664d3ab03098ae2798b75e2b6927ac900187

11 years agoMerge "Move buffers pointer out of the control block"
Glenn Kasten [Fri, 9 Nov 2012 01:13:02 +0000 (17:13 -0800)]
Merge "Move buffers pointer out of the control block"

11 years agoNEON optimized SINC resampler
Mathias Agopian [Sun, 4 Nov 2012 23:16:13 +0000 (15:16 -0800)]
NEON optimized SINC resampler

this currently gives us a 60% to 80% boost depending
on the quality level selected.

Change-Id: I7db385007e811ed7bffe5fd3403b44e300894f5b

11 years agominor cleanups
Mathias Agopian [Mon, 5 Nov 2012 09:51:37 +0000 (01:51 -0800)]
minor cleanups

Change-Id: Ia12ee4fb59e90221761bec85e6450db29197591f

11 years agoMove buffers pointer out of the control block
Glenn Kasten [Thu, 8 Nov 2012 20:13:58 +0000 (12:13 -0800)]
Move buffers pointer out of the control block

This is part of a series to clean up the control block.

Change-Id: Ie474557db7cb360f2d9a0f11600a68f5a3d46f07