Starting with version 4.7, gcc properly supports unaligned
memory accesses on ARM. Not using the inline asm with these
compilers results in better code.
Signed-off-by: Mans Rullgard <mans@mansr.com>
With a dereferenced type-cast pointer as memory operand, gcc 4.6
and later will sometimes copy the data to a temporary location,
the address of which is used as the operand value, if it thinks
the target address might be misaligned. Using a pointer to a
packed struct type instead does the right thing.
The 16-bit case is special since the ldrh instruction addressing
modes are limited compared to ldr. The "Uq" constraint produces a
memory reference suitable for an ldrsb instruction, which supports
the same addressing modes as ldrh. However, the restrictions appear
to apply only when the operand addresses a single byte. The memory
reference must thus be split into two operands each targeting one
byte. Finally, the "Uq" constraint is only available in ARM mode.
The Thumb-2 ldrh instruction supports most addressing modes so the
normal "m" constraint can be used there.
Signed-off-by: Mans Rullgard <mans@mansr.com>
* qatar/master:
ARM: allow runtime masking of CPU features
dsputil: remove unused functions
mov: Treat keyframe indexes as 1-origin if starting at non-zero.
mov: Take stps entries into consideration also about key_off.
Remove lowres video decoding
Conflicts:
ffmpeg.c
ffplay.c
libavcodec/arm/vp8dsp_init_arm.c
libavcodec/libopenjpegdec.c
libavcodec/mjpegdec.c
libavcodec/mpegvideo.c
libavcodec/utils.c
libavformat/mov.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
This allows masking CPU features with the -cpuflags avconv option
which is useful for testing different optimisations without rebuilding.
Signed-off-by: Mans Rullgard <mans@mansr.com>
* qatar/master:
pcm-mpeg: convert to bytestream2 API
Revert "h264: clear trailing bits in partially parsed NAL units"
remove iwmmxt optimizations
mimic: do not continue if swap_buf_size is 0
mimic: convert to bytestream2 API
frwu: use MKTAG to check marker instead of AV_RL32
txd: port to bytestream2 API
c93: convert to bytestream2 API
iff: make .long_name more descriptive
FATE: add test for cdxl demuxer
rtsp: Fix a typo
Conflicts:
libavcodec/arm/dsputil_iwmmxt.c
libavcodec/arm/dsputil_iwmmxt_rnd_template.c
libavcodec/arm/mpegvideo_iwmmxt.c
libavcodec/c93.c
libavcodec/txd.c
libavutil/arm/cpu.c
tests/fate/demux.mak
Merged-by: Michael Niedermayer <michaelni@gmx.at>
The were broken since August of 2010 without anyone noticing until
three weeks ago. Nobody cares about it anymore and hopefully Marvell
will support NEON like in the PXA978 from now on.
With these changes, gcc 4.5 and later recognise it as a bswap
and use the proper instructions on ARM and x86. On x86, the
16-bit bswap is recognised from gcc 4.1.
Signed-off-by: Mans Rullgard <mans@mansr.com>
* qatar/master:
doxygen: Consistently use '@' instead of '\' for Doxygen markup.
Use av_printf_format to check the usage of printf style functions
Add av_printf_format, for marking printf style format strings and their parameters
ARM: enable thumb for Cortex-M* CPUs
nsvdec: Propagate error values instead of returning 0 in nsv_read_header().
build: remove SRC_PATH_BARE variable
build: move basic rules and variables to main Makefile
build: move special targets to end of main Makefile
lavdev: improve feedback in case of invalid frame rate/size
vfwcap: prefer "framerate_q" over "fps" in vfw_read_header()
v4l2: prefer "framerate_q" over "fps" in v4l2_set_parameters()
fbdev: prefer "framerate_q" over "fps" in device context
bktr: prefer "framerate" over "fps" for grab_read_header()
ALSA: implement channel layout for playback.
alsa: support unsigned variants of already supported signed formats.
alsa: add support for more formats.
ARM: allow building in Thumb2 mode
Conflicts:
common.mak
doc/APIchanges
libavcodec/vdpau.h
libavdevice/alsa-audio-common.c
libavdevice/fbdev.c
libavdevice/libdc1394.c
libavutil/avutil.h
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* qatar/master: (27 commits)
ac3enc: fix LOCAL_ALIGNED usage in count_mantissa_bits()
ac3dsp: do not use the ff_* prefix when referencing ff_ac3_bap_bits.
ac3dsp: fix loop condition in ac3_update_bap_counts_c()
ARM: unbreak build
ac3enc: modify mantissa bit counting to keep bap counts for all values of bap instead of just 0 to 4.
ac3enc: split mantissa bit counting into a separate function.
ac3enc: store per-block/channel bap pointers by reference block in a 2D array rather than in the AC3Block struct.
get_bits: add av_unused tag to cache variable
sws: replace all long with int.
ARM: aacdec: fix constraints on inline asm
ARM: remove unnecessary volatile from inline asm
ARM: add "cc" clobbers to inline asm where needed
ARM: improve FASTDIV asm
ac3enc: use LOCAL_ALIGNED macro
APIchanges: fill in git hash for av_get_pix_fmt_name (0420bd7).
lavu: add av_get_pix_fmt_name() convenience function
cmdutils: remove OPT_FUNC2
swscale: fix crash in bilinear scaling.
vpxenc: add VP8E_SET_STATIC_THRESHOLD mapping
webm: support stereo videos in matroska/webm muxer
...
Conflicts:
Changelog
cmdutils.c
cmdutils.h
doc/APIchanges
doc/muxers.texi
ffmpeg.c
ffplay.c
libavcodec/ac3enc.c
libavcodec/ac3enc_float.c
libavcodec/avcodec.h
libavcodec/get_bits.h
libavcodec/libvpxenc.c
libavcodec/version.h
libavdevice/libdc1394.c
libavformat/matroskaenc.c
libavutil/avutil.h
libswscale/rgb2rgb.c
libswscale/swscale.c
libswscale/swscale_template.c
libswscale/x86/swscale_template.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* qatar/master: (31 commits)
ARM: add ARMv6 optimised av_clip_uintp2
ARM: remove volatile from asm statements in libavutil/intmath
ARM: fix av_clipl_int32_arm()
v4l: include avdevice.h
ffserver: move close_connection() call to avoid a temporary string and copy.
lavf: initialize demuxer private options.
AVOptions: set string default values.
lavdevice: mark v4l for removal on next major bump.
swscale: fix compile on ppc.
swscale: fix compile on x86-32.
build: Remove generated .version file on distclean.
configure: Add -D_GNU_SOURCE to CPPFLAGS on OS/2.
doc: Drop hint at --enable-memalign-hack for MinGW, it is now autodetected.
ffplay: Remove disabled code.
Mark parameterless function declarations as 'void'.
swscale: use av_clip_uint8() in yuv2yuv1_c().
swscale: remove VOF/VOFW.
swscale: split chroma buffers into separate U/V planes.
swscale: replace formatConvBuffer[VOF] by allocated array.
rgb2rgb: remove duplicate mmx/mmx2/3dnow/sse2 functions.
...
Merged-by: Michael Niedermayer <michaelni@gmx.at>
The volatile qualifiers are not needed on these statements as
their effects are fully specified by constraints.
Signed-off-by: Mans Rullgard <mans@mansr.com>
On some versions of gcc, these weren't always getting inlined due to hitting
the inline cap limit in some files. This is generally bad, as most of these
functions are smaller inlined than not.
(cherry picked from commit eb3755a5aa)
On some versions of gcc, these weren't always getting inlined due to hitting
the inline cap limit in some files. This is generally bad, as most of these
functions are smaller inlined than not.
This prevents gcc inserting useless UXTH instructions, at least
in some cases.
Originally committed as revision 25212 to svn://svn.ffmpeg.org/ffmpeg/trunk
Instead of defining functions in per-arch header files included
by the main cpu.c, define them normally and call them from the
generic one.
Originally committed as revision 25084 to svn://svn.ffmpeg.org/ffmpeg/trunk
This reduces the number of false dependencies on header files and
speeds up compilation.
Originally committed as revision 22407 to svn://svn.ffmpeg.org/ffmpeg/trunk
ARMv6 and later support unaligned loads and stores for single
word/halfword but not double/multiple. GCC is ignorant of this and
will always use bytewise accesses for unaligned data. Casting to an
int32_t pointer is dangerous since a load/store double or multiple
instruction might be used (this happens with some code in FFmpeg).
Implementing the AV_[RW]* macros with inline asm using only supported
instructions gives fast and safe unaligned accesses. ARM RVCT does
the right thing with generic code.
This gives an overall speedup of up to 10%.
Originally committed as revision 18601 to svn://svn.ffmpeg.org/ffmpeg/trunk
