We currently hit asserts in AECM where the output of WebRtcSpl_NormW16() on armv7 is incorrect.
I've verified that it outputs -17 for negative values. Internally that means that clz returns 0 after a two's complement operation on a int16_t.
There is a mismatch between the int16_t input and otherwise 32 bit assumptions. Explicitly casting to int32_t makes the two's complement do the correct thing.
The CL also extends the unit tests by running through a larger set of values.
BUG=4486
TESTED=locally on Android Nexus 7 and trybots
R=aluebs@webrtc.org, kwiberg@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/49549004
Cr-Commit-Position: refs/heads/master@{#8897}
The macro is defined as
#define WEBRTC_SPL_MUL_16_16_RSFT(a, b, c) \
(WEBRTC_SPL_MUL_16_16(a, b) >> (c))
where the latter macro is in C defined as
#define WEBRTC_SPL_MUL_16_16(a, b) \
((int32_t) (((int16_t)(a)) * ((int16_t)(b))))
(For definitions on ARMv7 and MIPS, see common_audio/signal_processing/include/spl_inl_{armv7,mips}.h)
The replacement consists of
- avoiding casts to int16_t if inputs already are int16_t
- adding explicit cast to <type> if result is assigned to <type> (other than int or int32_t)
- minor cleanups like remove of unnecessary parentheses and style changes
BUG=3347, 3348, 3353
TESTED=locally on Linux for both fixed and floating point and trybots
R=kwiberg@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/44799004
Cr-Commit-Position: refs/heads/master@{#8857}
The macro is defined as
#define WEBRTC_SPL_MUL_16_16_RSFT(a, b, c) \
(WEBRTC_SPL_MUL_16_16(a, b) >> (c))
where the latter macro is in C defined as
#define WEBRTC_SPL_MUL_16_16(a, b) \
((int32_t) (((int16_t)(a)) * ((int16_t)(b))))
(For definitions on ARMv7 and MIPS, see common_audio/signal_processing/include/spl_inl_{armv7,mips}.h)
The replacement consists of
- avoiding casts to int16_t if inputs already are int16_t
- adding explicit cast to <type> if result is assigned to <type> (other than int or int32_t)
- minor cleanups like remove of unnecessary parentheses and style changes
BUG=3348, 3353
TESTED=locally on Linux for both fixed and floating point and trybots
R=kwiberg@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/49499004
Cr-Commit-Position: refs/heads/master@{#8853}
We are using the Ooura FFT in a few places:
- AGC
- Transient suppression
- Noise suppression
The optimized OpenMAX DL FFT is considerably faster, but currently does
not compile everywhere, notably on iOS. This change will allow us to use
Openmax when possible and otherwise fall back to Ooura.
(Unfortunately, noise suppression won't be able to take advantage of it
since it's not C++. Upgrade time?)
R=aluebs@webrtc.org, mgraczyk@chromium.org
Review URL: https://webrtc-codereview.appspot.com/45789004
Cr-Commit-Position: refs/heads/master@{#8798}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8798 4adac7df-926f-26a2-2b94-8c16560cd09d
- Remove unnecessary window member from lapped_transform.
- Add comment indicated that Blocker does not take ownership of
the window passed to its constructor.
- Streamline LappedTransform constructor so members can be const.
Also use a range-based for loop in audio_processing_impl.cc for clarity.
R=aluebs@webrtc.org, andrew@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/41229004
Cr-Commit-Position: refs/heads/master@{#8708}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8708 4adac7df-926f-26a2-2b94-8c16560cd09d
Clang version changed 223108:230914
Details: e144d30..6fdb142/tools/clang/scripts/update.sh
Removes the OVERRIDE macro defined in:
* webrtc/base/common.h
* webrtc/typedefs.h
The majority of the source changes were done by running this in src/:
perl -0pi -e "s/virtual\s([^({;]*(\([^({;]*\)[^({;]*))(OVERRIDE|override)/\1override/sg" `find {talk,webrtc} -name "*.h" -o -name "*.cc*" -o -name "*.mm*"`
which converted all:
virtual Foo() OVERRIDE
functions to:
Foo() override
Then I manually edited:
* talk/media/webrtc/fakewebrtccommon.h
* webrtc/test/fake_common.h
Remaining uses of OVERRIDE was fixed by search+replace.
Manual edits were done to fix virtual destructors that were
overriding inherited ones.
Finally a build error related to the pure virtual definitions of
Read, Write and Rewind in common_types.h required a bit of
refactoring in:
* webrtc/common_types.cc
* webrtc/common_types.h
* webrtc/system_wrappers/interface/file_wrapper.h
* webrtc/system_wrappers/source/file_impl.cc
This roll should make it possible for us to finally re-enable deadlock
detection for TSan on the buildbots.
BUG=4106
R=pbos@webrtc.org, tommi@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/41069004
Cr-Commit-Position: refs/heads/master@{#8596}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8596 4adac7df-926f-26a2-2b94-8c16560cd09d
Mostly, it's about moving constructors and descructors to the .cc
files, so that they won't be inlined everywhere.
The reason this CL is so big is that a lot of code was using
common_types.h without declaring a dependency on webrtc_common, which
broke the build once common_types.h started to depend on
common_types.cc.
BUG=163
R=kjellander@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/26089004
Cr-Commit-Position: refs/heads/master@{#8516}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8516 4adac7df-926f-26a2-2b94-8c16560cd09d
As part of the review, refactored AudioConverter into internal derived
classes, each focused on one type of conversion. A factory method
returns the correct converter (or chain of converters, via
CompositionConverter).
BUG=b/18938079
R=rojer@google.com
Review URL: https://webrtc-codereview.appspot.com/35699004
Cr-Commit-Position: refs/heads/master@{#8322}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8322 4adac7df-926f-26a2-2b94-8c16560cd09d
Now the ChannelBuffer has 2 separate arrays, one for the full-band data and one for the splitted one. The corresponding accessors are added to the ChannelBuffer.
This is done to avoid having to refresh the bands pointers in AudioBuffer. It will also allow us to have a general accessor like data()[band][channel][sample].
All the files using the ChannelBuffer needed to be re-factored.
Tested with modules_unittests, common_audio_unittests, audioproc, audioproc_f, voe_cmd_test.
R=andrew@webrtc.org, kwiberg@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/36999004
Cr-Commit-Position: refs/heads/master@{#8318}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8318 4adac7df-926f-26a2-2b94-8c16560cd09d
The modification only uses the unique part of the WebRtcSpl_MaxAbsValue
function. Pass Spltest.MinMaxOperationTest conformance test on both
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-a15 |
| use C as the base on each | (1.2Ghz) | (1.7Ghz) |
| CPU target | | |
|----------------------------+-----------+------------|
| Neon asm | 32% | 15% |
| Neon intrinsics (GCC 4.6) | 36% | 37% |
| Neon intrinsics (GCC 4.8) | 35% | 18% |
BUG=3580
R=andrew@webrtc.org, jridges@masque.com
Change-Id: Ia2f6822ec58774b401cc440b6751a97e540b5048
Review URL: https://webrtc-codereview.appspot.com/30109004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@7803 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
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
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
