diff options
| author | Ronald S. Bultje <rsbultje@gmail.com> | 2013-01-20 15:41:52 -0800 |
|---|---|---|
| committer | Ronald S. Bultje <rsbultje@gmail.com> | 2013-01-22 11:55:42 -0800 |
| commit | d56668bd80075615b89aff652fe8a576bf853ceb (patch) | |
| tree | 9da3ed036b716dbaf33f5c9869578bedb6e393a2 /libavutil/arm | |
| parent | 5959bfaca396ecaf63a8123055f499688b79cae3 (diff) | |
| download | ffmpeg-d56668bd80075615b89aff652fe8a576bf853ceb.tar.gz | |
floatdsp: move scalarproduct_float from dsputil to avfloatdsp.
This makes the aac decoder and all voice codecs independent of dsputil.
Diffstat (limited to 'libavutil/arm')
| -rw-r--r-- | libavutil/arm/float_dsp_init_neon.c | 3 | ||||
| -rw-r--r-- | libavutil/arm/float_dsp_neon.S | 13 |
2 files changed, 16 insertions, 0 deletions
diff --git a/libavutil/arm/float_dsp_init_neon.c b/libavutil/arm/float_dsp_init_neon.c index b3644e82a2..a7245ad92b 100644 --- a/libavutil/arm/float_dsp_init_neon.c +++ b/libavutil/arm/float_dsp_init_neon.c @@ -43,6 +43,8 @@ void ff_vector_fmul_reverse_neon(float *dst, const float *src0, void ff_butterflies_float_neon(float *v1, float *v2, int len); +float ff_scalarproduct_float_neon(const float *v1, const float *v2, int len); + void ff_float_dsp_init_neon(AVFloatDSPContext *fdsp) { fdsp->vector_fmul = ff_vector_fmul_neon; @@ -52,4 +54,5 @@ void ff_float_dsp_init_neon(AVFloatDSPContext *fdsp) fdsp->vector_fmul_add = ff_vector_fmul_add_neon; fdsp->vector_fmul_reverse = ff_vector_fmul_reverse_neon; fdsp->butterflies_float = ff_butterflies_float_neon; + fdsp->scalarproduct_float = ff_scalarproduct_float_neon; } diff --git a/libavutil/arm/float_dsp_neon.S b/libavutil/arm/float_dsp_neon.S index 4acc406d33..559b565628 100644 --- a/libavutil/arm/float_dsp_neon.S +++ b/libavutil/arm/float_dsp_neon.S @@ -256,3 +256,16 @@ function ff_butterflies_float_neon, export=1 bgt 1b bx lr endfunc + +function ff_scalarproduct_float_neon, export=1 + vmov.f32 q2, #0.0 +1: vld1.32 {q0},[r0,:128]! + vld1.32 {q1},[r1,:128]! + vmla.f32 q2, q0, q1 + subs r2, r2, #4 + bgt 1b + vadd.f32 d0, d4, d5 + vpadd.f32 d0, d0, d0 +NOVFP vmov.32 r0, d0[0] + bx lr +endfunc |