yading@10: /*
yading@10:  * ARM NEON optimised integer operations
yading@10:  * Copyright (c) 2009 Kostya Shishkov
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with FFmpeg; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: #include "libavutil/arm/asm.S"
yading@10: 
yading@10:         .fpu neon
yading@10: 
yading@10: function ff_scalarproduct_int16_neon, export=1
yading@10:         vmov.i16        q0,  #0
yading@10:         vmov.i16        q1,  #0
yading@10:         vmov.i16        q2,  #0
yading@10:         vmov.i16        q3,  #0
yading@10: 1:      vld1.16         {d16-d17}, [r0]!
yading@10:         vld1.16         {d20-d21}, [r1,:128]!
yading@10:         vmlal.s16       q0,  d16,  d20
yading@10:         vld1.16         {d18-d19}, [r0]!
yading@10:         vmlal.s16       q1,  d17,  d21
yading@10:         vld1.16         {d22-d23}, [r1,:128]!
yading@10:         vmlal.s16       q2,  d18,  d22
yading@10:         vmlal.s16       q3,  d19,  d23
yading@10:         subs            r2,  r2,   #16
yading@10:         bne             1b
yading@10: 
yading@10:         vpadd.s32       d16, d0,   d1
yading@10:         vpadd.s32       d17, d2,   d3
yading@10:         vpadd.s32       d10, d4,   d5
yading@10:         vpadd.s32       d11, d6,   d7
yading@10:         vpadd.s32       d0,  d16,  d17
yading@10:         vpadd.s32       d1,  d10,  d11
yading@10:         vpadd.s32       d2,  d0,   d1
yading@10:         vpaddl.s32      d3,  d2
yading@10:         vmov.32         r0,  d3[0]
yading@10:         bx              lr
yading@10: endfunc
yading@10: 
yading@10: @ scalarproduct_and_madd_int16(/*aligned*/v0,v1,v2,order,mul)
yading@10: function ff_scalarproduct_and_madd_int16_neon, export=1
yading@10:         vld1.16         {d28[],d29[]}, [sp]
yading@10:         vmov.i16        q0,  #0
yading@10:         vmov.i16        q1,  #0
yading@10:         vmov.i16        q2,  #0
yading@10:         vmov.i16        q3,  #0
yading@10:         mov             r12, r0
yading@10: 
yading@10: 1:      vld1.16         {d16-d17}, [r0,:128]!
yading@10:         vld1.16         {d18-d19}, [r1]!
yading@10:         vld1.16         {d20-d21}, [r2]!
yading@10:         vld1.16         {d22-d23}, [r0,:128]!
yading@10:         vld1.16         {d24-d25}, [r1]!
yading@10:         vld1.16         {d26-d27}, [r2]!
yading@10:         vmul.s16        q10, q10,  q14
yading@10:         vmul.s16        q13, q13,  q14
yading@10:         vmlal.s16       q0,  d16,  d18
yading@10:         vmlal.s16       q1,  d17,  d19
yading@10:         vadd.s16        q10, q8,   q10
yading@10:         vadd.s16        q13, q11,  q13
yading@10:         vmlal.s16       q2,  d22,  d24
yading@10:         vmlal.s16       q3,  d23,  d25
yading@10:         vst1.16         {q10},     [r12,:128]!
yading@10:         subs            r3,  r3,   #16
yading@10:         vst1.16         {q13},     [r12,:128]!
yading@10:         bne             1b
yading@10: 
yading@10:         vpadd.s32       d16, d0,   d1
yading@10:         vpadd.s32       d17, d2,   d3
yading@10:         vpadd.s32       d10, d4,   d5
yading@10:         vpadd.s32       d11, d6,   d7
yading@10:         vpadd.s32       d0,  d16,  d17
yading@10:         vpadd.s32       d1,  d10,  d11
yading@10:         vpadd.s32       d2,  d0,   d1
yading@10:         vpaddl.s32      d3,  d2
yading@10:         vmov.32         r0,  d3[0]
yading@10:         bx              lr
yading@10: endfunc