This change is needed by ChromeOS as it introduces -fno-omit-frame-pointer
flag (see code.google.com/p/chromium/issues/detail?id=477749). This causes
compile error for MIPS, as some MIPS optimization blocks use maximum possible
number of available registers.
Also, this change contains minor GN build fix for MIPS platform regarding the
pitch_filter_mips.c / pitch_filter_c.c file inclusion.
BUG=477749
R=andrew@webrtc.org, djordje.pesut@imgtec.com, tina.legrand@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/48139004
Patch from Ljubomir Papuga <lpapuga@mips.com>.
Cr-Commit-Position: refs/heads/master@{#9047}
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
- removed commented code lines used during development
- excluded fft.c since there are neon optimizations used and a removal may cause a performance regression
BUG=3348, 3353
TESTED=locally on linux and trybots
R=kwiberg@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/48799004
Cr-Commit-Position: refs/heads/master@{#8967}
Now that android_webview_build is no longer supported, remove build
conditionals referencing it and also remove the extra level of
indirection used to reference the cpufeatures target.
BUG=chromium:440793
R=henrika@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/44119005
Patch from Richard Coles <torne@chromium.org>.
Cr-Commit-Position: refs/heads/master@{#8963}
The macro is defined as
#define WEBRTC_SPL_LSHIFT_W32(a, b) ((a) << (b))
hence trivial.
The macro name may in fact mislead the user to assume a cast/truncation to int32_t is done.
- Removing usage of it.
- Some style changes.
BUG=3348, 3353
TESTED=locally on linux and trybots
R=kwiberg@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/46749005
Cr-Commit-Position: refs/heads/master@{#8918}
Pass content_browsertests in Chromium. Performance test result (lower is
better):
C version: 100%
old intrinsics Neon version (with bug): 16.5%
new intrinsics Neon version: 18.0%
asm Neon version: 23.3%
BUG=4002
R=andrew@webrtc.org, jridges@masque.com
Change-Id: Ia0a96ac237216b635fc528f67d39319cdf246281
Review URL: https://webrtc-codereview.appspot.com/46739004
Cr-Commit-Position: refs/heads/master@{#8907}
We want to crate the illusion that iSAC supports 48000 Hz decoding,
while in fact it outputs 32000 Hz. This is the iSAC fullband mode.
Currently this is (also) handled by higher layers, but in future
changes this will not be the case.
R=tina.legrand@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/47809004
Cr-Commit-Position: refs/heads/master@{#8889}
A codec's packet-loss concealer is called once from NetEq before
decoding the first packet after a packet loss. The purpose is not to
use the PLC output, but to prepare the state of the decoder such that
it may recover faster after the loss. However, this effect is not
achieved by calling iSAC's PLC. Also, there are some problems with the
fixed-point implementation of the PLC (see the associated bug).
BUG=4423
R=kwiberg@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/42849004
Cr-Commit-Position: refs/heads/master@{#8827}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8827 4adac7df-926f-26a2-2b94-8c16560cd09d
The macro is defined as
#define WEBRTC_SPL_LSHIFT_W32(a, b) ((a) << (b))
It is a trivial operation that need no macro. In fact it may be confusing for to the user, since it can be interpreted as having an implicit cast to int32_t.
BUG=3348,3353
TESTED=locally on linux and trybots
R=kwiberg@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/44659004
Cr-Commit-Position: refs/heads/master@{#8801}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8801 4adac7df-926f-26a2-2b94-8c16560cd09d
NetEQ can crash when decoder gives too many output samples than it can handle. A practical case this happens is when multiple opus packets are combined.
The best solution is to pass the max size to the ACM decode function and let it return a failure if the max size if too small.
BUG=4361
R=henrik.lundin@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/45619004
Cr-Commit-Position: refs/heads/master@{#8730}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8730 4adac7df-926f-26a2-2b94-8c16560cd09d
The macro is in C defined as
#define WEBRTC_SPL_MUL_16_16(a, b) ((int32_t) (((int16_t)(a)) * ((int16_t)(b))))
(For definition on ARMv7 and MIPS, see
common_audio/signal_processing/include/spl_inl_armv7.h and
common_audio/signal_processing/include/spl_inl_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)
Some other minor code cleanup also exists.
BUG=3348,3353
TESTED=locally on Mac and trybots
R=henrik.lundin@webrtc.org, kwiberg@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/42639004
Cr-Commit-Position: refs/heads/master@{#8717}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8717 4adac7df-926f-26a2-2b94-8c16560cd09d
The current way that iSAC RCU is packetized and sent as a RED packet,
with the same payload type for primary and redundant payloads, does
not follow the specification for RED. As it is now, it is impossible
for a receiver to know if an incoming RED packet with iSAC payloads
inside consists of two "primary" (but time-shifted) payloads, or one
primary and one RCU payload. The RED standard stipulates that the
former option is the correct interpretation, while our implementation
currently applies the latter.
This CL removes support for iSAC RCU from Audio Coding Module, but
leaves it in the iSAC codec itself (i.e., in the C implementation).
BUG=4402
COAUTHOR=kwiberg@webrtc.orgR=tina.legrand@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/45569004
Cr-Commit-Position: refs/heads/master@{#8713}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8713 4adac7df-926f-26a2-2b94-8c16560cd09d
Added method AudioEncoder::MaxEncodedBytes() and provided implementations in derived encoders. This method returns the number of bytes that can be produced by the encoder at each Encode() call.
Unit tests were updated to use the new method.
Buffer allocation was not changed in AudioCodingModuleImpl::Encode(). It will be done after additional investigation.
Other refactoring work that was done, that may not be obvious why:
1. Moved some code into AudioEncoderCng::EncodePassive() to make it more consistent with EncodeActive().
2. Changed the order of NumChannels() and RtpTimestampRateHz() declarations in AudioEncoderG722 and AudioEncoderCopyRed classes. It just bothered me that the order was not the same as in AudioEncoder class and its other derived classes.
R=kwiberg@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/40259005
Cr-Commit-Position: refs/heads/master@{#8671}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8671 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
This should be safe to land now that issue 4143 was resolved (in r8492).
This change effectively reverts 8488.
TBR=kwiberg@webrtc.org
Original commit message:
This CL changes the way the decoder sample rate is set and updated. In
practice, it only concerns the iSAC (float) codec.
One single iSAC decoder instance is used for both wideband and
super-wideband decoding, and the instance must be told to switch
output frequency if the payload type changes. This used to be done
through a call to UpdateDecoderSampleRate, but is now instead done in
the Decode call as an extra parameter.
Review URL: https://webrtc-codereview.appspot.com/39289004
Cr-Commit-Position: refs/heads/master@{#8496}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8496 4adac7df-926f-26a2-2b94-8c16560cd09d
Without this patch, Valgrind's Memcheck was complaining that the test
for whether we should return -1 following the call to
WebRtcIsac_DecodeLb made a conditional branch or move based on the
value of numSamplesLB, which was uninitialized if WebRtcIsac_DecodeLb
failed.
However, as can be seen in the source, the control flow only depends
on the value of numSamplesLB if numDecodedBytesLB >= 0; i.e., if
WebRtcIsac_DecodeLb returned successfully, in which case numSamplesLB
is always initialized. The discrepancy is due to the fact that
Valgrind works on the generated machine code, which contains spurious
such dependencies. The generated code for this test:
if ((numDecodedBytesLB < 0) || (numDecodedBytesLB > lenEncodedLBBytes) ||
(numSamplesLB > MAX_FRAMESAMPLES)) {
looks like this:
95: 0f bf 45 d6 movswl -0x2a(%rbp),%eax
99: 3d c0 03 00 00 cmp $0x3c0,%eax
9e: 0f 8f 45 01 00 00 jg 1e9 <Decode+0x1e9>
a4: 44 89 f0 mov %r14d,%eax
a7: c1 e0 10 shl $0x10,%eax
aa: 0f 88 39 01 00 00 js 1e9 <Decode+0x1e9>
b0: 41 0f bf ce movswl %r14w,%ecx
b4: 89 8d 98 e1 ff ff mov %ecx,-0x1e68(%rbp)
ba: 41 0f bf c7 movswl %r15w,%eax
be: 39 c1 cmp %eax,%ecx
c0: 0f 8f 23 01 00 00 jg 1e9 <Decode+0x1e9>
Note how the compiler has seemingly ignored the C language's guarantee
that the arguments to || must be evaluated in left-to-right order, and
compares numSamplesLB (%eax) with MAX_FRAMESAMPLES (0x3c0, a.k.a. 960)
before the other two conditions! If the uninitialized value in
numSamplesLB happens to be greater than 960, we'll jump to
Decode+0x1e9 (where we'll return -1) without even looking at the other
two conditions. Has the compiler generated broken code?
Well, no. If numDecodedBytesLB is < 0 so that numSamplesLB is
uninitialized, we'll end up jumping to 1e9 whether that value is
greater than 960 or not; we'll just do it with different jump
instructions. This is entirely invisible as far as the C language is
concerned, but the dependency on the uninitialized value is visible at
the machine code level, which is why Memcheck complains.
This patch solves the problem by pragmatically initializing
numSamplesLB before the call even though it isn't necessary other than
for placating Memcheck.
BUG=4143
R=henrik.lundin@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/36309004
Cr-Commit-Position: refs/heads/master@{#8492}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8492 4adac7df-926f-26a2-2b94-8c16560cd09d
This CL changes the way the decoder sample rate is set and updated. In
practice, it only concerns the iSAC (float) codec.
One single iSAC decoder instance is used for both wideband and
super-wideband decoding, and the instance must be told to switch
output frequency if the payload type changes. This used to be done
through a call to UpdateDecoderSampleRate, but is now instead done in
the Decode call as an extra parameter.
R=kwiberg@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/34349004
Cr-Commit-Position: refs/heads/master@{#8476}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8476 4adac7df-926f-26a2-2b94-8c16560cd09d
Although sample_rate_hz(), num_channels(), and rtp_timestamp_rate_hz()
are simple accessors for almost all implementations of AudioEncoder,
they are virtual and not guaranteed to be just simple accessors. Thus,
it makes more sense to use the normal CamelCase naming scheme.
BUG=4235
R=henrik.lundin@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/34239004
Cr-Commit-Position: refs/heads/master@{#8407}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8407 4adac7df-926f-26a2-2b94-8c16560cd09d
- Add max_bit_rate and max_payload_size_bytes to config structs.
- Fix support for 48 kHz sample rate.
- Fix iSAC-RED.
- Add method UpdateDecoderSampleRate().
- Update locking structure with a separate lock for local member
variables used by the encoder methods.
BUG=3926
COAUTHOR:kwiberg@webrtc.orgR=minyue@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/41659004
Cr-Commit-Position: refs/heads/master@{#8204}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8204 4adac7df-926f-26a2-2b94-8c16560cd09d
The following three methods are added:
rtp_timestamp_rate_hz()
SetTargetBitrate()
SetProjectedPacketLossRate()
Default implementations are provided, and a few overrides are
implemented. AudioEncoderCopyRed and AudioEncoderCng propagate the new
methods to the underlying speech codec.
BUG=3926
COAUTHOR:kwiberg@webrtc.orgR=tina.legrand@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/34049004
Cr-Commit-Position: refs/heads/master@{#8171}
git-svn-id: http://webrtc.googlecode.com/svn/trunk@8171 4adac7df-926f-26a2-2b94-8c16560cd09d
