Disable the VC-1 motion compensation NEON asm which is the only part
using that directive. The integrated assembler in the upcoming clang 3.5
does not support .dn/.qn without plans to change that. Too much effort
to implement it while it is rarely used.
http://llvm.org/bugs/show_bug.cgi?id=18199.
Not supported by Clang's integrated assembler. Since it just adds
debug information it can safely omitted.
---
configure | 5 +
libavutil/aarch64/asm.S | 12 +---
libavutil/arm/asm.S | 12 +---
3 files changed, 23 insertions(+), 6 deletions(-)
diff --git a/c
Hi,
following patches make our arm and aarch64 asm ready for the integrated
assembler in llvm/clang 3.5.
AArch64 needs one patch in llvm to support .req on AArch64 which is not
yet submitted. I need to write test cases for it. It seems to work fine
except fate failures for monkeyaudio (not relat
Clang's integrated assembler does not support it and it is optional
for GNU as.
---
libavutil/aarch64/asm.S | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/libavutil/aarch64/asm.S b/libavutil/aarch64/asm.S
index 8ac5c18..86f24d9 100644
--- a/libavutil/aarch64/asm.S
+++ b/libavu
Clang's integrated assembler (after 3.4) does not yet support the
'.altmacro' directive which is only used in arm asm. Support is planned:
http://llvm.org/bugs/show_bug.cgi?id=18918
---
configure | 6 ++
1 file changed, 6 insertions(+)
diff --git a/configure b/configure
index 6e2ab24..e5deaf0
On 2014-05-29 14:39:21 +0300, Martin Storsjö wrote:
> The gas manual shows using a #, while gas itself actually doesn't
> support that syntax. Make it optional to keep supporting building
> earlier versions of libav which used this syntax.
> ---
> gas-preprocessor.pl | 4 ++--
> 1 file changed, 2
On 2014-05-29 14:39:12 +0300, Martin Storsjö wrote:
> This fixes building in PIC mode with gas. The examples in the gas
> manual showed using a # here even though gas itself actually didn't
> support that syntax (and the gas test suite only tests it without
> the extra hash sign).
>
> CC: libav-st
On 2014-05-19 17:53:38 +0300, Martin Storsjö wrote:
> This fixes compilation with MSVC.
> ---
> libavutil/display.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/libavutil/display.c b/libavutil/display.c
> index ae6c1c2..b2b9832 100644
> --- a/libavutil/display.c
> +++ b/libavutil/disp
---
now the version which actually compiles, forgot to amend the commit
number of side data elements in AVStream is nv_side_data instead of
side_data_elems in AVPacket
doc/APIchanges | 4
libavformat/avformat.h | 11 +++
libavformat/utils.c| 15 +++
libavform
---
doc/APIchanges | 4
libavformat/avformat.h | 11 +++
libavformat/utils.c| 15 +++
libavformat/version.h | 4 ++--
4 files changed, 32 insertions(+), 2 deletions(-)
diff --git a/doc/APIchanges b/doc/APIchanges
index b5defe4..92b8fda 100644
--- a/doc/APIc
On 2014-05-19 12:03:13 +0200, Anton Khirnov wrote:
>
> On Mon, 19 May 2014 11:49:27 +0200, Janne Grunau
> wrote:
> > On 2014-05-18 21:30:26 +0200, Anton Khirnov wrote:
> > > This also handles side data properly.
> > > ---
> > > libavcodec/pthread_fram
On 2014-05-18 21:19:16 +0200, Anton Khirnov wrote:
> Side data count is incremented by by calling av_packet_new_side_data()
> in the following loop, setting it explicitly results in the resulting
> value being twice what it should be.
>
> CC: libav-sta...@libav.org
> ---
> libavcodec/avpacket.c |
On 2014-05-18 21:30:26 +0200, Anton Khirnov wrote:
> This also handles side data properly.
> ---
> libavcodec/pthread_frame.c | 17 +++--
> 1 file changed, 3 insertions(+), 14 deletions(-)
>
> diff --git a/libavcodec/pthread_frame.c b/libavcodec/pthread_frame.c
> index 34ca9a6..66b1
On 2014-05-19 06:35:38 +0200, Anton Khirnov wrote:
> From: Vittorio Giovara
>
> Add AV_PKT_DATA_DISPLAYMATRIX and AV_FRAME_DATA_DISPLAYMATRIX as stream and
> frame side data (respectively) to describe a display transformation matrix
> for linear transformation operations on the decoded video.
>
On 2014-05-18 21:34:33 +0200, Anton Khirnov wrote:
> From: Vittorio Giovara
>
> Signed-off-by: Anton Khirnov
> ---
> libavformat/isom.h |2 ++
> libavformat/mov.c | 61
> +++-
> 2 files changed, 53 insertions(+), 10 deletions(-)
ok
Janne
On 2014-05-15 22:06:50 -0400, Vittorio Giovara wrote:
> ---
> libavformat/isom.h | 2 ++
> libavformat/mov.c | 42 +++---
> 2 files changed, 41 insertions(+), 3 deletions(-)
>
> diff --git a/libavformat/isom.h b/libavformat/isom.h
> index bf0792c..b2d5574 100
On 2014-05-15 22:06:49 -0400, Vittorio Giovara wrote:
> Add AV_PKT_DATA_DISPLAYMATRIX and AV_FRAME_DATA_DISPLAYMATRIX as stream and
> frame side data (respectively) to describe a display transformation matrix
> for linear transformation operations on the decoded video.
>
> Add APIs to easily conve
On 2014-05-03 21:13:30 +0200, Anton Khirnov wrote:
> ---
> avconv.c| 91
> ++-
> avconv.h|1 +
> avconv_filter.c |8 ++---
> avconv_opt.c| 25 +--
> 4 files changed, 69 insertions(+), 56 deletions(-)
On 2014-05-03 21:13:21 +0200, Anton Khirnov wrote:
> This function should not modify any global state, so there should be no
> reason for any locking.
> ---
> libavcodec/utils.c | 18 --
> 1 file changed, 18 deletions(-)
>
> diff --git a/libavcodec/utils.c b/libavcodec/utils.c
>
On 2014-05-03 21:13:29 +0200, Anton Khirnov wrote:
> ---
> avconv.c| 66
> ---
> avconv.h|1 +
> avconv_filter.c | 18 +++
> avconv_opt.c| 28 ---
> 4 files changed, 64 insertions(+),
On 2014-05-15 18:27:54 +0200, Christophe Gisquet wrote:
> Hi,
>
> (message addressed to the people in the thread rather than just the poster)
>
> 2014-05-13 17:25 GMT+02:00 :
> > +static void inline ff_dct_unquantize_h263_neon(int qmul, int qadd,
>
> I think the vpx, rv40 and h264 decoders at l
---
configure | 8
1 file changed, 8 insertions(+)
diff --git a/configure b/configure
index 6f84dfa..d9ec8ae 100755
--- a/configure
+++ b/configure
@@ -1361,6 +1361,7 @@ HAVE_LIST_PUB="
HEADERS_LIST="
alsa_asoundlib_h
altivec_h
+arm_neon_h
arpa_inet_h
cdio_parano
On 2014-05-15 18:10:58 +0200, Diego Biurrun wrote:
> On Thu, May 15, 2014 at 12:09:46PM +0200, Janne Grunau wrote:
> > On 2014-05-14 23:20:12 +0200, Diego Biurrun wrote:
> > > On Wed, May 14, 2014 at 03:15:15PM +0200, Janne Grunau wrote:
> > > > --- a/M
Removes an ugly $(if) to avoid conflicts with $(FFLIBS-) in the
top-level MAkefile.
---
common.mak | 2 +-
configure | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/common.mak b/common.mak
index c17b5bf..9244fd3 100644
--- a/common.mak
+++ b/common.mak
@@ -8,7 +8,7 @@ all:
---
libavcodec/Makefile | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/libavcodec/Makefile b/libavcodec/Makefile
index 60c3f32..66426d3 100644
--- a/libavcodec/Makefile
+++ b/libavcodec/Makefile
@@ -698,7 +698,7 @@ SKIPHEADERS-$(CONFIG_DXVA2)+= dxva2.h
dxva2_inter
On 2014-05-15 17:08:48 +0200, Diego Biurrun wrote:
> The feature is of doubtful utility and has no known users.
ask pandora.be if they still exists and need this ;)
> ---
> common.mak | 3 +--
> configure | 38 +++---
> library.mak | 2 +-
> 3 files changed,
avcodec might depend on avresample and with --as-needed required symbols
might be get removed if avresample is linked before avcodec.
Fixes link failures of avprobe and avplay on aarch64 with
--enable-neon-clobber-test.
---
re-indented and added linking order comment
Makefile | 11 ++-
1
On 2014-05-15 21:41:29 +0800, James Yu wrote:
> On 15 May 2014 17:08, Janne Grunau wrote:
>
> > > >
> > > > If you have a real target, the build directory needs to be available on
> > > > target. If you can't mount to the same path use
> >
avcodec might depend on avresample and with --as-needed required symbols
might be get removed if avresample is linked before avcodec.
Fixes link failures of avprobe and avplay on aarch64 with
--enable-neon-clobber-test.
---
Makefile | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git
On 2014-05-14 23:20:12 +0200, Diego Biurrun wrote:
> On Wed, May 14, 2014 at 03:15:15PM +0200, Janne Grunau wrote:
> > --- a/Makefile
> > +++ b/Makefile
> > @@ -130,6 +130,9 @@ $(foreach V,$(SUBDIR_VARS),$(eval $(call RESET,$(V
> > SUBDIR := $(1)/
> >
On 2014-05-15 10:13:01 +0800, James Yu wrote:
> On 15 May 2014 02:13, Janne Grunau wrote:
>
> > On 2014-05-15 00:16:48 +0800, James Yu wrote:
> > > On 14 May 2014 21:12, James Yu wrote:
> > >
> > > I found this fails even in master branch. Could yo
On 2014-05-15 00:16:48 +0800, James Yu wrote:
> On 14 May 2014 21:12, James Yu wrote:
>
> I found this fails even in master branch. Could you help check? Thanks
>
> commit 7e90133f6420b1c53652f972b9561600822881ee
> Author: Janne Grunau
> Date: Wed May 14 00:05:27 2014
---
Would be nice to avoid the ifdef/endif in the Makefile
just testing for the header i not enough, since clang 3.4 fails over
its own arm_neon.h
Makefile | 3 +++
configure | 11 +++
2 files changed, 14 insertions(+)
diff --git a/Makefile b/Makefile
index fd60521..1a55045 100644
---
On 2014-05-13 23:25:12 +0800, james...@linaro.org wrote:
> From: James Yu
>
> Using NEON intrinsics to support ARMv7 and ARMv8 simultaneously.
> Rewrite ARMv7 assembly to intinsics version.
Thanks for the patch. We prefer asm over intrinsics especially for functions
with existing ARMv7 NEON asm
Avoids including disabled library Makefiles.
---
uninteded side effect of renaming the required FFLIBS for each lib
FFLIBS-$(NAME)
common.mak | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/common.mak b/common.mak
index cdcdc8a..c17b5bf 100644
--- a/common.mak
+++ b/common.mak
On 2014-05-13 20:30:09 +0300, Martin Storsjö wrote:
> On Mon, 12 May 2014, Janne Grunau wrote:
>
> >Opus celt decoding 11% faster and the iMDCT over 2.5 times faster on
> >Apple's A7.
> >---
> >Untangled the 3 * 5 * 2^x imdct/fft from opus. Only file, str
On 2014-05-13 20:23:55 +0300, Martin Storsjö wrote:
> On Tue, 13 May 2014, Janne Grunau wrote:
>
> >Values are positive powers of two, so just replace it with right shift.
> >---
> >libavcodec/aarch64/fft_neon.S | 2 +-
> >1 file changed, 1 insertion(+), 1 d
On 2014-05-13 19:28:24 +0200, Diego Biurrun wrote:
> On Tue, May 13, 2014 at 11:55:38AM +0200, Janne Grunau wrote:
> > --- a/configure
> > +++ b/configure
> > @@ -655,6 +655,15 @@ prepend(){
> >
> > +unique(){
> > +var=$1
> > +uniq_list=&
---
gas-preprocessor.pl | 8
1 file changed, 8 insertions(+)
diff --git a/gas-preprocessor.pl b/gas-preprocessor.pl
index ab4294c..b597a0e 100755
--- a/gas-preprocessor.pl
+++ b/gas-preprocessor.pl
@@ -788,6 +788,14 @@ sub handle_serialized_line {
if ($line =~
/^\s*(s|u)xtl(2)?
Values are positive powers of two, so just replace it with right shift.
---
libavcodec/aarch64/fft_neon.S | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/libavcodec/aarch64/fft_neon.S b/libavcodec/aarch64/fft_neon.S
index 5189bfb..9802349 100644
--- a/libavcodec/aarch64/fft_neo
---
gas-preprocessor.pl | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/gas-preprocessor.pl b/gas-preprocessor.pl
index f4a55a9..ab4294c 100755
--- a/gas-preprocessor.pl
+++ b/gas-preprocessor.pl
@@ -784,8 +784,9 @@ sub handle_serialized_line {
$line = "
Instead of setting FFLIBS in each library Makefile configure
exports FFLIBS-$library in config.mak.
---
all comments addressed, changed lib_version to printf instead of echo
to avoid newlines in $requires
common.mak | 2 +-
configure | 55
On 2014-05-13 00:59:14 +0200, Diego Biurrun wrote:
> On Thu, May 08, 2014 at 11:27:07PM +0200, Janne Grunau wrote:
> > On 2014-05-08 17:23:26 +0200, Diego Biurrun wrote:
> > > On Sat, May 03, 2014 at 07:20:24PM +0200, Janne Grunau wrote:
> > --- a/configure
> >
on.S
b/libavcodec/aarch64/opus_imdct_neon.S
new file mode 100644
index 000..6b06396
--- /dev/null
+++ b/libavcodec/aarch64/opus_imdct_neon.S
@@ -0,0 +1,647 @@
+/*
+ * Copyright (c) 2014 Janne Grunau
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/
On 2014-05-11 12:43:37 +0200, Luca Barbato wrote:
> It would reduce the boilerplate code users have to write.
> ---
>
> Now more complete and with the right message-id.
>
> libavcodec/avcodec.h | 5 +
> libavcodec/utils.c | 7 +++
> 2 files changed, 12 insertions(+)
>
> diff --git a/l
On 2014-05-11 08:23:25 +0200, Anton Khirnov wrote:
>
> On Thu, 8 May 2014 15:15:06 +0200, Luca Barbato wrote:
> > It would reduce the boilerplate code users have to write.
> > ---
> > libavcodec/avcodec.h | 5 +
> > libavcodec/pthread_frame.c | 4
> > 2 files changed, 9 insertion
On 2014-04-15 05:16:31 +0200, Vittorio Giovara wrote:
> ---
> libavfilter/vf_showinfo.c | 24
> 1 file changed, 24 insertions(+)
>
> diff --git a/libavfilter/vf_showinfo.c b/libavfilter/vf_showinfo.c
> index 031d8fe..de1c5d9 100644
> --- a/libavfilter/vf_showinfo.c
> +++
On 2014-05-10 10:10:20 -0400, Reinhard Tartler wrote:
> ---
> src/download | 24 +---
> src/news | 24
> 2 files changed, 37 insertions(+), 11 deletions(-)
looks good
Janne
___
libav-devel mailing list
l
---
libavcodec/opus_celt.c | 1 -
1 file changed, 1 deletion(-)
diff --git a/libavcodec/opus_celt.c b/libavcodec/opus_celt.c
index 3c86f66..1e7fadb 100644
--- a/libavcodec/opus_celt.c
+++ b/libavcodec/opus_celt.c
@@ -2105,7 +2105,6 @@ int ff_celt_decode_frame(CeltContext *s, OpusRangeCoder
*rc,
---
libavcodec/opus_imdct.c | 12 ++--
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/libavcodec/opus_imdct.c b/libavcodec/opus_imdct.c
index aa1b0d6..09037da 100644
--- a/libavcodec/opus_imdct.c
+++ b/libavcodec/opus_imdct.c
@@ -262,11 +262,11 @@ static void celt_imdct_half
Hi,
changed $name_FFLIBS to FFLIBS-$name which matches the other make
variables better
On 2014-05-08 17:23:26 +0200, Diego Biurrun wrote:
> On Sat, May 03, 2014 at 07:20:24PM +0200, Janne Grunau wrote:
>
> > --- a/configure
> > +++ b/configure
> > @@ -655,6 +655,16 @@ pr
On 2014-05-08 12:17:51 +0300, Martin Storsjö wrote:
> The -MD option (that specifies using a dynamically linked CRT)
> can be interpreted by cpp to create a bogus dependency file named
> '-.d'.
> ---
> gas-preprocessor.pl | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/gas-preprocessor.p
On 2014-05-07 18:49:41 +0300, Martin Storsjö wrote:
> On Wed, 7 May 2014, Janne Grunau wrote:
>
> >On 2014-05-07 16:29:54 +0300, Martin Storsjö wrote:
> >>On Wed, 7 May 2014, Janne Grunau wrote:
> >>
> >>>On 2014-05-07 14:00:11 +0300, Martin Storsjö wrot
On 2014-05-07 16:29:54 +0300, Martin Storsjö wrote:
> On Wed, 7 May 2014, Janne Grunau wrote:
>
> >On 2014-05-07 14:00:11 +0300, Martin Storsjö wrote:
> >>the normal compiler. If specified manually (with --as="gaspp armasm"
> >>or so), configure t
On 2014-05-07 14:00:11 +0300, Martin Storsjö wrote:
> On Wed, 7 May 2014, Janne Grunau wrote:
>
> >---
> >gas-preprocessor.pl | 12 +++-
> >1 file changed, 11 insertions(+), 1 deletion(-)
> >
> >diff --git a/gas-preprocessor.pl b/gas-preprocessor
---
gas-preprocessor.pl | 12 +++-
1 file changed, 11 insertions(+), 1 deletion(-)
diff --git a/gas-preprocessor.pl b/gas-preprocessor.pl
index d6d12c3..f4a55a9 100755
--- a/gas-preprocessor.pl
+++ b/gas-preprocessor.pl
@@ -126,7 +126,17 @@ if ((grep /^-c$/, @gcc_cmd) && !(grep /^-o/, @gc
use stddev test for any sample using the silk decoder
add fate-opus-celt/hybrid/silk for tests which use one mode exclusively
---
tests/fate/opus.mak | 28 ++--
1 file changed, 22 insertions(+), 6 deletions(-)
diff --git a/tests/fate/opus.mak b/tests/fate/opus.mak
index 31
Instead of setting FFLIBS in each library Makefile configure
exports $LIBNAME_FFLIBS in config.mak.
---
common.mak | 2 +-
configure | 55 +++---
libavcodec/Makefile| 1 -
libavdevice/Makefile | 1 -
libavfilter/Makefile
Required for the conditional libavresample dependency added by the
native opus decoder to libavcodec.
---
Dependencies could be declared in configure and written to config.mak
that would avoid the reverse DOSUBDIR handling in Makefile. They are
already required for the pkg-config files so we could
Fixes at least fate-opus on arm.
---
pushed to oracle
libavcodec/opus.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/libavcodec/opus.h b/libavcodec/opus.h
index 5aa151b..02e5ad3 100644
--- a/libavcodec/opus.h
+++ b/libavcodec/opus.h
@@ -124,7 +124,7 @@ typedef struct OpusSt
On 2014-04-29 15:22:02 +0200, Anton Khirnov wrote:
> ---
> tests/tiny_psnr.c |5 +
> 1 file changed, 5 insertions(+)
>
> diff --git a/tests/tiny_psnr.c b/tests/tiny_psnr.c
> index 66eaf82..0055ffc 100644
> --- a/tests/tiny_psnr.c
> +++ b/tests/tiny_psnr.c
> @@ -223,6 +223,11 @@ int main(i
On 2014-04-29 15:22:01 +0200, Anton Khirnov wrote:
> This function is missing at least on MSVC.
> ---
> configure|1 +
> libavutil/libm.h |7 +++
> 2 files changed, 8 insertions(+)
>
> diff --git a/configure b/configure
> index 30f90eb..764d6a0 100755
> --- a/configure
> +++ b
On 2014-04-28 23:03:16 +0200, Anton Khirnov wrote:
> Move the dxva check after the dependencies have been detected, make sure
> the ole32 library exists.
>
> Fixes build on non-Windows.
> ---
> Makefile |2 +-
> avconv_opt.c |2 +-
> configure| 10 +++---
> 3 files changed,
---
typo fixed, rebased onto the v2 patch and using FFMAX(N, 19)
libavcodec/opus_imdct.c | 10 +++---
1 file changed, 7 insertions(+), 3 deletions(-)
diff --git a/libavcodec/opus_imdct.c b/libavcodec/opus_imdct.c
index 671f8d5..d4ffd95 100644
--- a/libavcodec/opus_imdct.c
+++ b/libavcodec/op
---
patch against my tree still based on v1 of the patch. so fuzz is expected.
Unsure if I should replace the (N + 4) in the malloc with FFMAX(N, 19),
reasoning for the '+ 4' was that it allows over reads but contrary to my
initial thinking it's not required for the NEON version.
libavcodec/opus_
Make the dependency explicit and add lavresample to lavcodec's
pkg-config file if the opus decoder is enabled.
Change the linking order for our programs. Only noticed that one after
enabling the avresample neon clobber test.
---
Makefile | 2 +-
configure | 6 +-
2 files changed, 6 insertion
On 2014-04-24 15:51:06 +0200, Niels Möller wrote:
> ni...@lysator.liu.se (Niels Möller) writes:
>
> > For a start, I implemented a *very* crude up-sampling, simply repeating
> > each core sample twice.
>
> I've now tried the following experiment:
>
> Decode the core stream using unmodified libav
On 2014-04-20 21:38:35 +0200, Anton Khirnov wrote:
> Initial implementation by Andrew D'Addesio during
> GSoC 2012.
>
> Completion by Anton Khirnov , sponsored by the
> Mozilla Corporation.
> ---
> Changelog|1 +
> libavcodec/Makefile |4 +
> libavcodec/allcodecs.c
On 2014-04-24 12:33:43 +0200, Niels Möller wrote:
> Janne Grunau writes:
>
> > the raw audio muxer have separate names for the sample formats.
> >
> > -f f32le $FILE at the end of the command will output little endian float
> > samples.
>
> Nice, -f s24le see
On 2014-04-24 11:13:23 +0200, Niels Möller wrote:
> When testing the dca code, I would really like to generate an
> uncompressed file (.wav or .au) with more than 16 bits per sample,
> ideally 24 bits signed integers.
>
> avconv -sample_fmts lists s32 and flt, and flt is what dcadec.c actually
> g
On 2014-04-24 09:29:28 +0200, Niels Möller wrote:
> ni...@lysator.liu.se (Niels Möller) writes:
>
> > Maybe there's also an X96 extension in the stream, I'll have to check.
> > Otherwise, the core data has to be up-sampled in some unspecified way
> > before the residual can be added to it.
>
> As
On 2014-04-24 07:48:57 +0200, Anton Khirnov wrote:
> In case start_frame() fails, this potentially invalid frame can still be
> output to the caller.
>
> Bug-Id: 672; debian/741240; ubuntu/1288206
> ---
> libavcodec/h264.c |2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/libavcodec/
---
libavcodec/arm/asm-offsets.h | 7 ---
libavcodec/arm/mpegvideo_arm.c | 14 --
libavutil/internal.h | 7 +++
3 files changed, 15 insertions(+), 13 deletions(-)
diff --git a/libavcodec/arm/asm-offsets.h b/libavcodec/arm/asm-offsets.h
index 8999456..fdc33ce 1006
Hi,
changes from the previous patchset:
- move struct ResampleContext definition to resample.h (suggested by Anton)
- replaced the hardcoded struct offsets with defines which are compile-time
checked.
- moves CHK_OFFS macro from libavcodec/arm/asm-offsets.h to
libavutil/internal.h and ren
ll
+++ b/libavresample/aarch64/resample_neon.S
@@ -0,0 +1,233 @@
+/*
+ * Copyright (c) 2014 Janne Grunau
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Fre
Required for arch optimized resampling.
---
libavresample/resample.c | 29 -
libavresample/resample.h | 29 +
2 files changed, 29 insertions(+), 29 deletions(-)
diff --git a/libavresample/resample.c b/libavresample/resample.c
index 5fd6b90..
On 2014-04-23 11:21:27 +0200, Christophe Gisquet wrote:
> Hi,
>
> 2014-04-22 23:46 GMT+02:00 Janne Grunau :
> > The assembler source is due to extensive macro use uglier than I would
> > like it to be but it's 4 functions in one. It looked nicer before
On 2014-04-23 12:29:26 +0300, Martin Storsjö wrote:
> On Tue, 22 Apr 2014, Janne Grunau wrote:
>
> >+#include "libavutil/aarch64/asm.S"
> >+
> >+.macro resample_one fmt, es=2
> >+function ff_resample_one_\fmt\()_neon, export=1
> >+sxtw
Required for arch optimized resampling.
---
libavresample/internal.h | 29 +
libavresample/resample.c | 29 -
2 files changed, 29 insertions(+), 29 deletions(-)
diff --git a/libavresample/internal.h b/libavresample/internal.h
index 057f89a..
+}
+}
+}
+}
diff --git a/libavresample/aarch64/resample_neon.S
b/libavresample/aarch64/resample_neon.S
new file mode 100644
index 000..548c820
--- /dev/null
+++ b/libavresample/aarch64/resample_neon.S
@@ -0,0 +1,219 @@
+/*
+ * Copyright (c) 2014 Janne Grunau
+ *
+ * This file is
On 2014-04-22 17:01:42 -0400, Derek Buitenhuis wrote:
> From: Michael Niedermayer
>
> Fix compilation
>
> Signed-off-by: Michael Niedermayer
> Signed-off-by: Derek Buitenhuis
> ---
> I guess luca moved it?
> ---
> libavutil/avutil.h | 10 +-
> 1 file changed, 5 insertions(+), 5 deleti
From: Janne Grunau
Get rid of the malloc. The product of N and B is at most
MAX(celt_freq_range) * 8.
Approximately 5% faster celt decoding.
---
feel free to squash it
libavcodec/opus_celt.c | 45 ++---
1 file changed, 30 insertions(+), 15 deletions
On 2014-04-20 21:38:35 +0200, Anton Khirnov wrote:
> Initial implementation by Andrew D'Addesio during
> GSoC 2012.
>
> Completion by Anton Khirnov , sponsored by the
> Mozilla Corporation.
> ---
> Changelog|1 +
> libavcodec/Makefile |4 +
> libavcodec/allcodecs.c
>From the ARMv7 NEON version. 16 times faster as the C version, overall
more than 12% faster vorbis decoding on Apple's A7.
---
libavcodec/aarch64/Makefile | 3 ++
libavcodec/aarch64/vorbisdsp_init.c | 34 +++
libavcodec/aarch64/vorbisdsp_neon.S | 82 ++
dex 000..5176b8d
--- /dev/null
+++ b/libavcodec/aarch64/mpegaudiodsp_neon.S
@@ -0,0 +1,224 @@
+/*
+ * Copyright (c) 2014 Janne Grunau
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General P
nvert_neon.S
new file mode 100644
index 000..198cf86
--- /dev/null
+++ b/libavresample/aarch64/audio_convert_neon.S
@@ -0,0 +1,363 @@
+/*
+ * Copyright (c) 2008 Mans Rullgard
+ * Copyright (c) 2014 Janne Grunau
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribu
Hi,
this patch series contains various optimisations aimed at lossy audio
decoding. I've benchmarked only aac, ac3, mp3 and vorbis but other
floating point or mpeg audio based codecs should profit too.
The FFT and MDCT have probably potential for further optimisations
since ARMv7 version is sligh
file mode 100644
index 000..bccd832
--- /dev/null
+++ b/libavcodec/aarch64/mdct_neon.S
@@ -0,0 +1,323 @@
+/*
+ * AArch64 NEON optimised MDCT
+ * Copyright (c) 2009 Mans Rullgard
+ * Copyright (c) 2014 Janne Grunau
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can
de 100644
index 000..5189bfb
--- /dev/null
+++ b/libavcodec/aarch64/fft_neon.S
@@ -0,0 +1,442 @@
+/*
+ * ARM NEON optimised FFT
+ *
+ * Copyright (c) 2009 Mans Rullgard
+ * Copyright (c) 2009 Naotoshi Nojiri
+ * Copyright (c) 2014 Janne Grunau
+ *
+ * This algorithm (though not any of th
On 2014-04-18 12:37:32 +0200, Hendrik Leppkes wrote:
> On Fri, Apr 18, 2014 at 12:27 PM, Janne Grunau wrote:
> > On 2014-04-17 20:38:27 +0200, Anton Khirnov wrote:
> >>
> >> Any progress?
> >
> > Not fully convinced from patch below. Without changes in
The most compelling place to initialize the hardware of AVHWAccel-backed
decoders is the get_format callback. To avoid unneeded reinitializations
ff_get_format() calls get_format() only if the previously returned pixel
format is not in the new pixel format list.
On a frame size change codecs are e
avctx->coded_{height,width} will always equal h->{height,width} since
init_dimensions() does that explicitly, Size changes are detected by
changes in mb_{height,width} earlier and propagated through the
needs_reinit variable.
---
libavcodec/h264_slice.c | 5 +
1 file changed, 1 insertion(+), 4
On 2014-04-17 20:17:42 +0200, Anton Khirnov wrote:
>
> On Thu, 17 Apr 2014 12:57:35 +0200, Janne Grunau
> wrote:
> > On 2014-04-17 12:46:37 +0200, Anton Khirnov wrote:
> > > ---
> > > tests/tiny_psnr.c |5 +
> > > 1 file changed, 5 ins
On 2014-04-17 20:38:27 +0200, Anton Khirnov wrote:
>
> On Tue, 25 Mar 2014 11:13:37 +0100, Janne Grunau
> wrote:
> > On 2014-03-24 21:20:27 +0100, Luca Barbato wrote:
> > > On 24/03/14 20:26, wm4 wrote:
> > > > In practice, it's called when the decoder
On 2014-04-17 12:46:37 +0200, Anton Khirnov wrote:
> ---
> tests/tiny_psnr.c |5 +
> 1 file changed, 5 insertions(+)
>
> diff --git a/tests/tiny_psnr.c b/tests/tiny_psnr.c
> index 66eaf82..5f819bb 100644
> --- a/tests/tiny_psnr.c
> +++ b/tests/tiny_psnr.c
> @@ -223,6 +223,11 @@ int main(i
On 2014-04-07 01:03:42 +0200, Luca Barbato wrote:
> On 07/04/14 00:56, Janne Grunau wrote:
> > Sets cc_default based on $toolchain. As effect adds -fPIC for
> > clang-tsan.
> > ---
> > or even shorter with shell substring processing
> >
> > configure | 27
On 2014-04-07 00:55:41 +0200, Luca Barbato wrote:
> Makes passing to configure
> --toolchain=gcc-asan --cross-prefix=armv7a-hardfloat-linux-gnueabi-
> work as intended.
> ---
>
> It is mostly lazy-friendly, the other toolchain options do not expect
> people to use cross-prefix at all or so I hope.
Sets cc_default based on $toolchain. As effect adds -fPIC for
clang-tsan.
---
or even shorter with shell substring processing
configure | 27 ++-
1 file changed, 6 insertions(+), 21 deletions(-)
diff --git a/configure b/configure
index 4d005f1..7487cb2 100755
--- a/config
On 2014-03-31 20:46:30 +0200, wm4 wrote:
> ff_MPV_common_init initializes Picture.f, so call it.
>
> s->m.mb_type must be set to NULL before ff_MPV_common_end, because it
> attempts to free it.
> ---
> Fixes a crash. Probably should be squashed with the previous commit,
> or moved before it.
> ---
On 2014-03-31 20:46:29 +0200, wm4 wrote:
> The only interesting parts are initialization in ff_MPV_common_init and
> uninitialization in ff_MPV_common_end.
>
> ff_mpeg_unref_picture and ff_thread_release_buffer have additional NULL
> checks for Picture.f, because these functions can be called on
>
601 - 700 of 2711 matches
Mail list logo
