All *neon.S files in aecm and ns modules have been removed. We need no
assembly offset generation now.
Pass byte to byte conformance test for aecm and ns test in audioproc
between new NEON (written in intrinsics) version and C version on both
ARMv7 and ARM64.
BUG=3580
R=andrew@webrtc.org, jridges@masque.com
Change-Id: I05d43d0c04d00bead65ca8c8fda25f0a42394b2b
Review URL: https://webrtc-codereview.appspot.com/32229004
Patch from Zhongwei Yai <zhongwei.yao@arm.com>.
git-svn-id: http://webrtc.googlecode.com/svn/trunk@7800 4adac7df-926f-26a2-2b94-8c16560cd09d
Pass compilation on both ARMv7 and ARM64. The generated
binary (audioproc) is byte to byte (with symbol striped) same as
before. The output of audioproc -aecm is also byte to byte same between
C and NEON version on ARMv7 and ARM64.
Change-Id: Ibdf40fe085f6bad1311f59bf9318bbcf37dd7ce5
BUG=3850
R=andrew@webrtc.org, jridges@masque.com
Review URL: https://webrtc-codereview.appspot.com/30239004
Patch from Zhongwei Yao <zhongwei.yao@arm.com>.
git-svn-id: http://webrtc.googlecode.com/svn/trunk@7783 4adac7df-926f-26a2-2b94-8c16560cd09d
Two VP9 tests needed to be disabled (see webrtc:4059) to make all tests pass.
Relevant changes:
* src/third_party/android_tools: ea50ccc..4c47ef6
* src/third_party/icu: dd72764..866ff69
* src/third_party/libvpx: 2e5ced5..429874c
* src/third_party/nss: 258342e..bb4e75a
* src/third_party/yasm/source/patched-yasm: c960eb1..4671120
* src/tools/gyp: 0a381c0..fe00999
* src/tools/swarming_client: 5b827c9..1d4965c
Details: 309cf65..24b4c73/DEPS
Clang version was not updated in this roll.
BUG=4059
TBR=pbos@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/26289004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@7778 4adac7df-926f-26a2-2b94-8c16560cd09d
Implemented the 3 bands splitting filter bank by:
1. Upsample by 4/3.
2. Split twice into 2 bands.
3. Discard upper most band, because it is empty anyway.
A unittest was also implemented:
1. Generate a signal from presence or absence of sine waves of different frequencies.
2. Split into 3 bands and check their presence or absence.
3. Recombine the bands.
4. Calculate delay (as it is an IIR it depends on frequency).
5. Check that the cross correlation of input and output is high enough at that delay.
BUG=webrtc:3146
R=andrew@webrtc.org, bjornv@webrtc.org, kwiberg@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/31029004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@7754 4adac7df-926f-26a2-2b94-8c16560cd09d
It was discovered that if remaining_bytes is an exact multiple of
max_payload_len in RtpPacketizerVp8::CalcNextSize, then the packetizer
will produce too many packets (i.e., split the payload into more
packets than needed).
This CL adds a test to trigger the problem.
BUG=4019
R=stefan@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/24289004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@7739 4adac7df-926f-26a2-2b94-8c16560cd09d
This CL fixes the problem described in issue 4021. In summary, of the
very first packet coming in to NetEq gets rejected because the RTP
payload type is unknown, subsequent GetAudio calls would trigger asserts
(in debug builds). The problem was that the first_packet_ variable was
reset and new_codec_ was set, even though the packet was discarded
further down the line. Now, these variables are modified after the
packet has been verified.
BUG=4021
R=tina.legrand@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/29089004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@7724 4adac7df-926f-26a2-2b94-8c16560cd09d
This doesn't change the behavior at all.
The logic behind this is having one class which manages all the splitting filters, because in the future we plan to add a 3 band one for 48kHz support.
It also breaks the dependency of the AudioBuffer with the filter states of these filters (which are going to be different for the 3 band one). The AudioBuffer is complicated enough and is going to need changes to support 3 bands in the future, so any simplification is a good idea.
On top of that it eliminates repeated code in the APM (now only iterating over channels, but then also deciding in how many bands to split). This should be managed by the AudioBuffer directly.
BUG=webrtc:3146
R=bjornv@webrtc.org, kwiberg@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/32469004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@7705 4adac7df-926f-26a2-2b94-8c16560cd09d
One debug recording with non matching sample rates between render and capture revealed a bug in modules/audio_processing/test/process_test.cc
The far_frame (render audio frame) used was loaded with the capture rate instead of the render rate with a data length mismatch error as result.
BUG=N/A
TESTED=manually on linux
R=kwiberg@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/27169004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@7695 4adac7df-926f-26a2-2b94-8c16560cd09d
This should speed up test execution on Android since only
the files needed by the test will be processed (instead
of the whole data + resources directories).
A few files for modules_unittests had to be explicitly added
for Android, since they were previously a part of the
add-whole-directories entries for the resources and data
directories.
BUG=webrtc:3741
TEST=Passing android+android_rel trybots.
R=phoglund@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/22559004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@7694 4adac7df-926f-26a2-2b94-8c16560cd09d
The modification only uses the unique part of the CalcLinearEnergies
function. Pass byte to byte conformance test both on ARMv7 and ARM64,
and the single function performance is similar with original assembly
version on different platforms. If not specified, the code is compiled
by GCC 4.6. The result is the "X version / C version" ratio, and the
less is better.
| run 100k times | cortex-a7 | cortex-a9 | cortex-a15 |
| use C as the base on each | (1.2Ghz) | (1.0Ghz) | (1.7Ghz) |
| CPU target | | | |
|----------------------------+-----------+-----------+------------|
| Neon asm | 19.48% | 19.26% | 13.68% |
| Neon inline | 27.90% | 28.87% | 17.79% |
| Neon intrinsics (GCC 4.8) | 18.69% | 20.18% | 14.69% |
| Neon intrinsics (LLVM 3.4) | 18.52% | 21.15% | 13.56% |
BUG=3580
R=andrew@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/23349004
Patch from Zhongwei Yao <zhongwei.yao@arm.com>.
git-svn-id: http://webrtc.googlecode.com/svn/trunk@7686 4adac7df-926f-26a2-2b94-8c16560cd09d
The modification only uses the unique part of the StoreAdaptiveChannel
function. Pass byte to byte conformance test both on ARM32 and ARM64,
and the single function performance is similar with original assembly
version on different platforms. If not specified, the code is compiled
by GCC 4.6. The result is the "X version / C version" ratio, and the
less is better.
| run 100k times | cortex-a7 | cortex-a9 | cortex-a15 |
| use C as the base on each | (1.2Ghz) | (1.0Ghz) | (1.7Ghz) |
| CPU target | | | |
|----------------------------+-----------+-----------+------------|
| Neon asm | 20.97% | 37.70% | 25.41% |
| Neon inline | 36.93% | 51.80% | 38.14% |
| Neon intrinsics (GCC 4.6) | 27.78% | 43.71% | 26.50% |
| Neon intrinsics (GCC 4.8) | 27.16% | 38.22% | 26.87% |
| Neon intrinsics (LLVM 3.4) | 27.82% | 39.90% | 26.69% |
Change-Id: Ia55d8a268a70164b50676c604ae40b68fc183106
BUG=3580
R=andrew@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/30029004
Patch from Zhongwei Yao <zhongwei.yao@arm.com>.
git-svn-id: http://webrtc.googlecode.com/svn/trunk@7685 4adac7df-926f-26a2-2b94-8c16560cd09d
