diff options
| author | Michael Niedermayer <michaelni@gmx.at> | 2011-07-03 02:56:12 +0200 |
|---|---|---|
| committer | Michael Niedermayer <michaelni@gmx.at> | 2011-07-03 03:14:10 +0200 |
| commit | cbfdfbe846f872d4283579c8cbca42c90896905c (patch) | |
| tree | adc2aebfe0b12ffa1a889dc72ab57bcb1e0ad9e0 /libavcodec/arm | |
| parent | 59bd0fef6691f5b23eebbd7dd2f3e74c8d60c6a7 (diff) | |
| parent | 5dd045ebc11933dca4d6af06e6b1e62be56802f7 (diff) | |
Merge remote-tracking branch 'qatar/master'
* qatar/master:
ARM: ac3: update ff_ac3_extract_exponents_neon per 8b7b2d6
ARM: NEON optimised vector_clip_int32()
swscale: disable full_chroma_int when converting to non-24/32bpp RGB.
suggest to use av_get_bytes_per_sample() in av_get_bits_per_sample_format() doxy
ffmpeg: use av_get_bytes_per_sample() in place of av_get_bits_per_sample_fmt()
put_bits: remove ALT_BITSTREAM_WRITER
put_bits: always use intreadwrite.h macros
libavformat: Add an example how to use the metadata API
doxygen: Prefer member groups over grouping into modules
doxygen: be more permissive when searching for API examples
avformat: doxify the Metadata API
lavf: restore old behavior for custom AVIOContex with an AVFMT_NOFILE format.
lavf: use the correct pointer in av_open_input_stream().
avidec: infer absolute vs relative index from first packet
Conflicts:
libavformat/Makefile
libavformat/avidec.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
Diffstat (limited to 'libavcodec/arm')
| -rw-r--r-- | libavcodec/arm/ac3dsp_neon.S | 7 | ||||
| -rw-r--r-- | libavcodec/arm/dsputil_init_neon.c | 3 | ||||
| -rw-r--r-- | libavcodec/arm/dsputil_neon.S | 16 |
3 files changed, 20 insertions, 6 deletions
diff --git a/libavcodec/arm/ac3dsp_neon.S b/libavcodec/arm/ac3dsp_neon.S index fdf1deabc9..e97197c27a 100644 --- a/libavcodec/arm/ac3dsp_neon.S +++ b/libavcodec/arm/ac3dsp_neon.S @@ -95,19 +95,14 @@ function ff_float_to_fixed24_neon, export=1 endfunc function ff_ac3_extract_exponents_neon, export=1 - vmov.i32 q14, #24 vmov.i32 q15, #8 1: - vld1.32 {q0}, [r1,:128] + vld1.32 {q0}, [r1,:128]! vabs.s32 q1, q0 vclz.i32 q3, q1 vsub.i32 q3, q3, q15 - vcge.s32 q2, q3, q14 - vbit q3, q14, q2 - vbic q0, q0, q2 vmovn.i32 d6, q3 vmovn.i16 d6, q3 - vst1.32 {q0}, [r1,:128]! vst1.32 {d6[0]}, [r0,:32]! subs r2, r2, #4 bgt 1b diff --git a/libavcodec/arm/dsputil_init_neon.c b/libavcodec/arm/dsputil_init_neon.c index 6faf3dc8d0..c818b33fef 100644 --- a/libavcodec/arm/dsputil_init_neon.c +++ b/libavcodec/arm/dsputil_init_neon.c @@ -160,6 +160,8 @@ void ff_vector_fmul_add_neon(float *dst, const float *src0, const float *src1, void ff_vector_clipf_neon(float *dst, const float *src, float min, float max, int len); +void ff_vector_clip_int32_neon(int32_t *dst, const int32_t *src, int32_t min, + int32_t max, unsigned int len); void ff_vorbis_inverse_coupling_neon(float *mag, float *ang, int blocksize); @@ -316,6 +318,7 @@ void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx) c->vector_fmul_reverse = ff_vector_fmul_reverse_neon; c->vector_fmul_add = ff_vector_fmul_add_neon; c->vector_clipf = ff_vector_clipf_neon; + c->vector_clip_int32 = ff_vector_clip_int32_neon; c->vector_fmul_sv_scalar[0] = ff_vector_fmul_sv_scalar_2_neon; c->vector_fmul_sv_scalar[1] = ff_vector_fmul_sv_scalar_4_neon; diff --git a/libavcodec/arm/dsputil_neon.S b/libavcodec/arm/dsputil_neon.S index 2147658af6..250a32af6e 100644 --- a/libavcodec/arm/dsputil_neon.S +++ b/libavcodec/arm/dsputil_neon.S @@ -815,3 +815,19 @@ function ff_apply_window_int16_neon, export=1 pop {r4,pc} endfunc + +function ff_vector_clip_int32_neon, export=1 + vdup.32 q0, r2 + vdup.32 q1, r3 + ldr r2, [sp] +1: + vld1.32 {q2-q3}, [r1,:128]! + vmin.s32 q2, q2, q1 + vmin.s32 q3, q3, q1 + vmax.s32 q2, q2, q0 + vmax.s32 q3, q3, q0 + vst1.32 {q2-q3}, [r0,:128]! + subs r2, r2, #8 + bgt 1b + bx lr +endfunc |