yading@10: /* yading@10: * ARM NEON optimised DSP functions yading@10: * Copyright (c) 2008 Mans Rullgard 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: function ff_clear_block_neon, export=1 yading@10: vmov.i16 q0, #0 yading@10: .rept 8 yading@10: vst1.16 {q0}, [r0,:128]! yading@10: .endr yading@10: bx lr yading@10: endfunc yading@10: yading@10: function ff_clear_blocks_neon, export=1 yading@10: vmov.i16 q0, #0 yading@10: .rept 8*6 yading@10: vst1.16 {q0}, [r0,:128]! yading@10: .endr yading@10: bx lr yading@10: endfunc yading@10: yading@10: function ff_put_pixels_clamped_neon, export=1 yading@10: vld1.16 {d16-d19}, [r0,:128]! yading@10: vqmovun.s16 d0, q8 yading@10: vld1.16 {d20-d23}, [r0,:128]! yading@10: vqmovun.s16 d1, q9 yading@10: vld1.16 {d24-d27}, [r0,:128]! yading@10: vqmovun.s16 d2, q10 yading@10: vld1.16 {d28-d31}, [r0,:128]! yading@10: vqmovun.s16 d3, q11 yading@10: vst1.8 {d0}, [r1,:64], r2 yading@10: vqmovun.s16 d4, q12 yading@10: vst1.8 {d1}, [r1,:64], r2 yading@10: vqmovun.s16 d5, q13 yading@10: vst1.8 {d2}, [r1,:64], r2 yading@10: vqmovun.s16 d6, q14 yading@10: vst1.8 {d3}, [r1,:64], r2 yading@10: vqmovun.s16 d7, q15 yading@10: vst1.8 {d4}, [r1,:64], r2 yading@10: vst1.8 {d5}, [r1,:64], r2 yading@10: vst1.8 {d6}, [r1,:64], r2 yading@10: vst1.8 {d7}, [r1,:64], r2 yading@10: bx lr yading@10: endfunc yading@10: yading@10: function ff_put_signed_pixels_clamped_neon, export=1 yading@10: vmov.u8 d31, #128 yading@10: vld1.16 {d16-d17}, [r0,:128]! yading@10: vqmovn.s16 d0, q8 yading@10: vld1.16 {d18-d19}, [r0,:128]! yading@10: vqmovn.s16 d1, q9 yading@10: vld1.16 {d16-d17}, [r0,:128]! yading@10: vqmovn.s16 d2, q8 yading@10: vld1.16 {d18-d19}, [r0,:128]! yading@10: vadd.u8 d0, d0, d31 yading@10: vld1.16 {d20-d21}, [r0,:128]! yading@10: vadd.u8 d1, d1, d31 yading@10: vld1.16 {d22-d23}, [r0,:128]! yading@10: vadd.u8 d2, d2, d31 yading@10: vst1.8 {d0}, [r1,:64], r2 yading@10: vqmovn.s16 d3, q9 yading@10: vst1.8 {d1}, [r1,:64], r2 yading@10: vqmovn.s16 d4, q10 yading@10: vst1.8 {d2}, [r1,:64], r2 yading@10: vqmovn.s16 d5, q11 yading@10: vld1.16 {d24-d25}, [r0,:128]! yading@10: vadd.u8 d3, d3, d31 yading@10: vld1.16 {d26-d27}, [r0,:128]! yading@10: vadd.u8 d4, d4, d31 yading@10: vadd.u8 d5, d5, d31 yading@10: vst1.8 {d3}, [r1,:64], r2 yading@10: vqmovn.s16 d6, q12 yading@10: vst1.8 {d4}, [r1,:64], r2 yading@10: vqmovn.s16 d7, q13 yading@10: vst1.8 {d5}, [r1,:64], r2 yading@10: vadd.u8 d6, d6, d31 yading@10: vadd.u8 d7, d7, d31 yading@10: vst1.8 {d6}, [r1,:64], r2 yading@10: vst1.8 {d7}, [r1,:64], r2 yading@10: bx lr yading@10: endfunc yading@10: yading@10: function ff_add_pixels_clamped_neon, export=1 yading@10: mov r3, r1 yading@10: vld1.8 {d16}, [r1,:64], r2 yading@10: vld1.16 {d0-d1}, [r0,:128]! yading@10: vaddw.u8 q0, q0, d16 yading@10: vld1.8 {d17}, [r1,:64], r2 yading@10: vld1.16 {d2-d3}, [r0,:128]! yading@10: vqmovun.s16 d0, q0 yading@10: vld1.8 {d18}, [r1,:64], r2 yading@10: vaddw.u8 q1, q1, d17 yading@10: vld1.16 {d4-d5}, [r0,:128]! yading@10: vaddw.u8 q2, q2, d18 yading@10: vst1.8 {d0}, [r3,:64], r2 yading@10: vqmovun.s16 d2, q1 yading@10: vld1.8 {d19}, [r1,:64], r2 yading@10: vld1.16 {d6-d7}, [r0,:128]! yading@10: vaddw.u8 q3, q3, d19 yading@10: vqmovun.s16 d4, q2 yading@10: vst1.8 {d2}, [r3,:64], r2 yading@10: vld1.8 {d16}, [r1,:64], r2 yading@10: vqmovun.s16 d6, q3 yading@10: vld1.16 {d0-d1}, [r0,:128]! yading@10: vaddw.u8 q0, q0, d16 yading@10: vst1.8 {d4}, [r3,:64], r2 yading@10: vld1.8 {d17}, [r1,:64], r2 yading@10: vld1.16 {d2-d3}, [r0,:128]! yading@10: vaddw.u8 q1, q1, d17 yading@10: vst1.8 {d6}, [r3,:64], r2 yading@10: vqmovun.s16 d0, q0 yading@10: vld1.8 {d18}, [r1,:64], r2 yading@10: vld1.16 {d4-d5}, [r0,:128]! yading@10: vaddw.u8 q2, q2, d18 yading@10: vst1.8 {d0}, [r3,:64], r2 yading@10: vqmovun.s16 d2, q1 yading@10: vld1.8 {d19}, [r1,:64], r2 yading@10: vqmovun.s16 d4, q2 yading@10: vld1.16 {d6-d7}, [r0,:128]! yading@10: vaddw.u8 q3, q3, d19 yading@10: vst1.8 {d2}, [r3,:64], r2 yading@10: vqmovun.s16 d6, q3 yading@10: vst1.8 {d4}, [r3,:64], r2 yading@10: vst1.8 {d6}, [r3,:64], r2 yading@10: bx lr yading@10: endfunc yading@10: yading@10: function ff_vector_clipf_neon, export=1 yading@10: VFP vdup.32 q1, d0[1] yading@10: VFP vdup.32 q0, d0[0] yading@10: NOVFP vdup.32 q0, r2 yading@10: NOVFP vdup.32 q1, r3 yading@10: NOVFP ldr r2, [sp] yading@10: vld1.f32 {q2},[r1,:128]! yading@10: vmin.f32 q10, q2, q1 yading@10: vld1.f32 {q3},[r1,:128]! yading@10: vmin.f32 q11, q3, q1 yading@10: 1: vmax.f32 q8, q10, q0 yading@10: vmax.f32 q9, q11, q0 yading@10: subs r2, r2, #8 yading@10: beq 2f yading@10: vld1.f32 {q2},[r1,:128]! yading@10: vmin.f32 q10, q2, q1 yading@10: vld1.f32 {q3},[r1,:128]! yading@10: vmin.f32 q11, q3, q1 yading@10: vst1.f32 {q8},[r0,:128]! yading@10: vst1.f32 {q9},[r0,:128]! yading@10: b 1b yading@10: 2: vst1.f32 {q8},[r0,:128]! yading@10: vst1.f32 {q9},[r0,:128]! yading@10: bx lr yading@10: endfunc yading@10: yading@10: function ff_apply_window_int16_neon, export=1 yading@10: push {r4,lr} yading@10: add r4, r1, r3, lsl #1 yading@10: add lr, r0, r3, lsl #1 yading@10: sub r4, r4, #16 yading@10: sub lr, lr, #16 yading@10: mov r12, #-16 yading@10: 1: yading@10: vld1.16 {q0}, [r1,:128]! yading@10: vld1.16 {q2}, [r2,:128]! yading@10: vld1.16 {q1}, [r4,:128], r12 yading@10: vrev64.16 q3, q2 yading@10: vqrdmulh.s16 q0, q0, q2 yading@10: vqrdmulh.s16 d2, d2, d7 yading@10: vqrdmulh.s16 d3, d3, d6 yading@10: vst1.16 {q0}, [r0,:128]! yading@10: vst1.16 {q1}, [lr,:128], r12 yading@10: subs r3, r3, #16 yading@10: bgt 1b yading@10: yading@10: pop {r4,pc} yading@10: endfunc yading@10: yading@10: function ff_vector_clip_int32_neon, export=1 yading@10: vdup.32 q0, r2 yading@10: vdup.32 q1, r3 yading@10: ldr r2, [sp] yading@10: 1: yading@10: vld1.32 {q2-q3}, [r1,:128]! yading@10: vmin.s32 q2, q2, q1 yading@10: vmin.s32 q3, q3, q1 yading@10: vmax.s32 q2, q2, q0 yading@10: vmax.s32 q3, q3, q0 yading@10: vst1.32 {q2-q3}, [r0,:128]! yading@10: subs r2, r2, #8 yading@10: bgt 1b yading@10: bx lr yading@10: endfunc