diff options
| author | Ronald S. Bultje <rsbultje@gmail.com> | 2013-01-19 22:26:58 -0800 |
|---|---|---|
| committer | Ronald S. Bultje <rsbultje@gmail.com> | 2013-01-22 11:55:42 -0800 |
| commit | 55aa03b9f8f11ebb7535424cc0e5635558590f49 (patch) | |
| tree | e13b12d715aa32cb1d81ed066735af40ea6dbb8a /libavutil/arm | |
| parent | 0881cbf314982cce8448bd12644ce2a6e0b8c576 (diff) | |
floatdsp: move vector_fmul_add from dsputil to avfloatdsp.
Diffstat (limited to 'libavutil/arm')
| -rw-r--r-- | libavutil/arm/float_dsp_init_neon.c | 4 | ||||
| -rw-r--r-- | libavutil/arm/float_dsp_neon.S | 27 |
2 files changed, 31 insertions, 0 deletions
diff --git a/libavutil/arm/float_dsp_init_neon.c b/libavutil/arm/float_dsp_init_neon.c index 16ea47154a..41e513fcdc 100644 --- a/libavutil/arm/float_dsp_init_neon.c +++ b/libavutil/arm/float_dsp_init_neon.c @@ -35,10 +35,14 @@ void ff_vector_fmul_scalar_neon(float *dst, const float *src, float mul, void ff_vector_fmul_window_neon(float *dst, const float *src0, const float *src1, const float *win, int len); +void ff_vector_fmul_add_neon(float *dst, const float *src0, const float *src1, + const float *src2, int len); + void ff_float_dsp_init_neon(AVFloatDSPContext *fdsp) { fdsp->vector_fmul = ff_vector_fmul_neon; fdsp->vector_fmac_scalar = ff_vector_fmac_scalar_neon; fdsp->vector_fmul_scalar = ff_vector_fmul_scalar_neon; fdsp->vector_fmul_window = ff_vector_fmul_window_neon; + fdsp->vector_fmul_add = ff_vector_fmul_add_neon; } diff --git a/libavutil/arm/float_dsp_neon.S b/libavutil/arm/float_dsp_neon.S index 540cfc6907..100eb02455 100644 --- a/libavutil/arm/float_dsp_neon.S +++ b/libavutil/arm/float_dsp_neon.S @@ -193,3 +193,30 @@ function ff_vector_fmul_window_neon, export=1 vst1.32 {d22,d23},[ip,:128], r5 pop {r4,r5,pc} endfunc + +function ff_vector_fmul_add_neon, export=1 + ldr r12, [sp] + vld1.32 {q0-q1}, [r1,:128]! + vld1.32 {q8-q9}, [r2,:128]! + vld1.32 {q2-q3}, [r3,:128]! + vmul.f32 q10, q0, q8 + vmul.f32 q11, q1, q9 +1: vadd.f32 q12, q2, q10 + vadd.f32 q13, q3, q11 + pld [r1, #16] + pld [r2, #16] + pld [r3, #16] + subs r12, r12, #8 + beq 2f + vld1.32 {q0}, [r1,:128]! + vld1.32 {q8}, [r2,:128]! + vmul.f32 q10, q0, q8 + vld1.32 {q1}, [r1,:128]! + vld1.32 {q9}, [r2,:128]! + vmul.f32 q11, q1, q9 + vld1.32 {q2-q3}, [r3,:128]! + vst1.32 {q12-q13},[r0,:128]! + b 1b +2: vst1.32 {q12-q13},[r0,:128]! + bx lr +endfunc |