Mercurial > libavcodec.hg
view arm/h264pred_neon.S @ 10644:5da7180afadf libavcodec
refactor and optimize scalarproduct
29-105% faster apply_filter, 6-90% faster ape decoding on core2
(Any x86 other than core2 probably gets much less, since this is mostly due to ssse3 cachesplit avoidance and I haven't written the full gamut of other cachesplit modes.)
9-123% faster ape decoding on G4.
| author | lorenm |
|---|---|
| date | Sat, 05 Dec 2009 15:09:10 +0000 |
| parents | f52d07b169b4 |
| children | 361a5fcb4393 |
line wrap: on
line source
/* * Copyright (c) 2009 Mans Rullgard <mans@mansr.com> * * This file is part of FFmpeg. * * FFmpeg is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * * FFmpeg is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public * License along with FFmpeg; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ #include "asm.S" .macro ldcol.8 rd, rs, rt, n=8, hi=0 .if \n == 8 || \hi == 0 vld1.8 {\rd[0]}, [\rs], \rt vld1.8 {\rd[1]}, [\rs], \rt vld1.8 {\rd[2]}, [\rs], \rt vld1.8 {\rd[3]}, [\rs], \rt .endif .if \n == 8 || \hi == 1 vld1.8 {\rd[4]}, [\rs], \rt vld1.8 {\rd[5]}, [\rs], \rt vld1.8 {\rd[6]}, [\rs], \rt vld1.8 {\rd[7]}, [\rs], \rt .endif .endm .macro add16x8 dq, dl, dh, rl, rh vaddl.u8 \dq, \rl, \rh vadd.u16 \dl, \dl, \dh vpadd.u16 \dl, \dl, \dl vpadd.u16 \dl, \dl, \dl .endm function ff_pred16x16_128_dc_neon, export=1 vmov.i8 q0, #128 b .L_pred16x16_dc_end .endfunc function ff_pred16x16_top_dc_neon, export=1 sub r2, r0, r1 vld1.8 {q0}, [r2,:128] add16x8 q0, d0, d1, d0, d1 vrshrn.u16 d0, q0, #4 vdup.8 q0, d0[0] b .L_pred16x16_dc_end .endfunc function ff_pred16x16_left_dc_neon, export=1 sub r2, r0, #1 ldcol.8 d0, r2, r1 ldcol.8 d1, r2, r1 add16x8 q0, d0, d1, d0, d1 vrshrn.u16 d0, q0, #4 vdup.8 q0, d0[0] b .L_pred16x16_dc_end .endfunc function ff_pred16x16_dc_neon, export=1 sub r2, r0, r1 vld1.8 {q0}, [r2,:128] sub r2, r0, #1 ldcol.8 d2, r2, r1 ldcol.8 d3, r2, r1 vaddl.u8 q0, d0, d1 vaddl.u8 q1, d2, d3 vadd.u16 q0, q0, q1 vadd.u16 d0, d0, d1 vpadd.u16 d0, d0, d0 vpadd.u16 d0, d0, d0 vrshrn.u16 d0, q0, #5 vdup.8 q0, d0[0] .L_pred16x16_dc_end: mov r3, #8 6: vst1.8 {q0}, [r0,:128], r1 vst1.8 {q0}, [r0,:128], r1 subs r3, r3, #1 bne 6b bx lr .endfunc function ff_pred16x16_hor_neon, export=1 sub r2, r0, #1 mov r3, #16 1: vld1.8 {d0[],d1[]},[r2], r1 vst1.8 {q0}, [r0,:128], r1 subs r3, r3, #1 bne 1b bx lr .endfunc function ff_pred16x16_vert_neon, export=1 sub r0, r0, r1 vld1.8 {q0}, [r0,:128], r1 mov r3, #8 1: vst1.8 {q0}, [r0,:128], r1 vst1.8 {q0}, [r0,:128], r1 subs r3, r3, #1 bne 1b bx lr .endfunc function ff_pred16x16_plane_neon, export=1 sub r3, r0, r1 add r2, r3, #8 sub r3, r3, #1 vld1.8 {d0}, [r3] vld1.8 {d2}, [r2,:64], r1 ldcol.8 d1, r3, r1 add r3, r3, r1 ldcol.8 d3, r3, r1 vrev64.8 q0, q0 vaddl.u8 q8, d2, d3 vsubl.u8 q2, d2, d0 vsubl.u8 q3, d3, d1 movrel r3, p16weight vld1.8 {q0}, [r3,:128] vmul.s16 q2, q2, q0 vmul.s16 q3, q3, q0 vadd.i16 d4, d4, d5 vadd.i16 d5, d6, d7 vpadd.i16 d4, d4, d5 vpadd.i16 d4, d4, d4 vshl.i16 d5, d4, #2 vaddl.s16 q2, d4, d5 vrshrn.s32 d4, q2, #6 mov r3, #0 vtrn.16 d4, d5 vadd.i16 d2, d4, d5 vshl.i16 d3, d2, #3 vrev64.16 d16, d17 vsub.i16 d3, d3, d2 vadd.i16 d16, d16, d0 vshl.i16 d2, d16, #4 vsub.i16 d2, d2, d3 vshl.i16 d3, d4, #4 vext.16 q0, q0, q0, #7 vsub.i16 d6, d5, d3 vmov.16 d0[0], r3 vmul.i16 q0, q0, d4[0] vdup.16 q1, d2[0] vdup.16 q2, d4[0] vdup.16 q3, d6[0] vshl.i16 q2, q2, #3 vadd.i16 q1, q1, q0 vadd.i16 q3, q3, q2 mov r3, #16 1: vqshrun.s16 d0, q1, #5 vadd.i16 q1, q1, q2 vqshrun.s16 d1, q1, #5 vadd.i16 q1, q1, q3 vst1.8 {q0}, [r0,:128], r1 subs r3, r3, #1 bne 1b bx lr .endfunc .section .rodata .align 4 p16weight: .short 1,2,3,4,5,6,7,8 .text function ff_pred8x8_hor_neon, export=1 sub r2, r0, #1 mov r3, #8 1: vld1.8 {d0[]}, [r2], r1 vst1.8 {d0}, [r0,:64], r1 subs r3, r3, #1 bne 1b bx lr .endfunc function ff_pred8x8_vert_neon, export=1 sub r0, r0, r1 vld1.8 {d0}, [r0,:64], r1 mov r3, #4 1: vst1.8 {d0}, [r0,:64], r1 vst1.8 {d0}, [r0,:64], r1 subs r3, r3, #1 bne 1b bx lr .endfunc function ff_pred8x8_plane_neon, export=1 sub r3, r0, r1 add r2, r3, #4 sub r3, r3, #1 vld1.32 {d0[0]}, [r3] vld1.32 {d2[0]}, [r2,:32], r1 ldcol.8 d0, r3, r1, 4, hi=1 add r3, r3, r1 ldcol.8 d3, r3, r1, 4 vaddl.u8 q8, d2, d3 vrev32.8 d0, d0 vtrn.32 d2, d3 vsubl.u8 q2, d2, d0 movrel r3, p16weight vld1.16 {q0}, [r3,:128] vmul.s16 d4, d4, d0 vmul.s16 d5, d5, d0 vpadd.i16 d4, d4, d5 vpaddl.s16 d4, d4 vshl.i32 d5, d4, #4 vadd.s32 d4, d4, d5 vrshrn.s32 d4, q2, #5 mov r3, #0 vtrn.16 d4, d5 vadd.i16 d2, d4, d5 vshl.i16 d3, d2, #2 vrev64.16 d16, d16 vsub.i16 d3, d3, d2 vadd.i16 d16, d16, d0 vshl.i16 d2, d16, #4 vsub.i16 d2, d2, d3 vshl.i16 d3, d4, #3 vext.16 q0, q0, q0, #7 vsub.i16 d6, d5, d3 vmov.16 d0[0], r3 vmul.i16 q0, q0, d4[0] vdup.16 q1, d2[0] vdup.16 q2, d4[0] vdup.16 q3, d6[0] vshl.i16 q2, q2, #3 vadd.i16 q1, q1, q0 vadd.i16 q3, q3, q2 mov r3, #8 1: vqshrun.s16 d0, q1, #5 vadd.i16 q1, q1, q3 vst1.8 {d0}, [r0,:64], r1 subs r3, r3, #1 bne 1b bx lr .endfunc function ff_pred8x8_128_dc_neon, export=1 vmov.i8 q0, #128 b .L_pred8x8_dc_end .endfunc function ff_pred8x8_top_dc_neon, export=1 sub r2, r0, r1 vld1.8 {d0}, [r2,:64] vpaddl.u8 d0, d0 vpadd.u16 d0, d0, d0 vrshrn.u16 d0, q0, #2 vdup.8 d1, d0[1] vdup.8 d0, d0[0] vtrn.32 d0, d1 b .L_pred8x8_dc_end .endfunc function ff_pred8x8_left_dc_neon, export=1 sub r2, r0, #1 ldcol.8 d0, r2, r1 vpaddl.u8 d0, d0 vpadd.u16 d0, d0, d0 vrshrn.u16 d0, q0, #2 vdup.8 d1, d0[1] vdup.8 d0, d0[0] b .L_pred8x8_dc_end .endfunc function ff_pred8x8_dc_neon, export=1 sub r2, r0, r1 vld1.8 {d0}, [r2,:64] sub r2, r0, #1 ldcol.8 d1, r2, r1 vtrn.32 d0, d1 vpaddl.u8 q0, q0 vpadd.u16 d0, d0, d1 vpadd.u16 d1, d0, d0 vrshrn.u16 d2, q0, #3 vrshrn.u16 d3, q0, #2 vdup.8 d0, d2[4] vdup.8 d1, d3[3] vdup.8 d4, d3[2] vdup.8 d5, d2[5] vtrn.32 q0, q2 .L_pred8x8_dc_end: mov r3, #4 add r2, r0, r1, lsl #2 6: vst1.8 {d0}, [r0,:64], r1 vst1.8 {d1}, [r2,:64], r1 subs r3, r3, #1 bne 6b bx lr .endfunc function ff_pred8x8_l0t_dc_neon, export=1 sub r2, r0, r1 vld1.8 {d0}, [r2,:64] sub r2, r0, #1 ldcol.8 d1, r2, r1, 4 vtrn.32 d0, d1 vpaddl.u8 q0, q0 vpadd.u16 d0, d0, d1 vpadd.u16 d1, d0, d0 vrshrn.u16 d2, q0, #3 vrshrn.u16 d3, q0, #2 vdup.8 d0, d2[4] vdup.8 d1, d3[0] vdup.8 q2, d3[2] vtrn.32 q0, q2 b .L_pred8x8_dc_end .endfunc function ff_pred8x8_l00_dc_neon, export=1 sub r2, r0, #1 ldcol.8 d0, r2, r1, 4 vpaddl.u8 d0, d0 vpadd.u16 d0, d0, d0 vrshrn.u16 d0, q0, #2 vmov.i8 d1, #128 vdup.8 d0, d0[0] b .L_pred8x8_dc_end .endfunc function ff_pred8x8_0lt_dc_neon, export=1 sub r2, r0, r1 vld1.8 {d0}, [r2,:64] add r2, r0, r1, lsl #2 sub r2, r2, #1 ldcol.8 d1, r2, r1, 4, hi=1 vtrn.32 d0, d1 vpaddl.u8 q0, q0 vpadd.u16 d0, d0, d1 vpadd.u16 d1, d0, d0 vrshrn.u16 d3, q0, #2 vrshrn.u16 d2, q0, #3 vdup.8 d0, d3[0] vdup.8 d1, d3[3] vdup.8 d4, d3[2] vdup.8 d5, d2[5] vtrn.32 q0, q2 b .L_pred8x8_dc_end .endfunc function ff_pred8x8_0l0_dc_neon, export=1 add r2, r0, r1, lsl #2 sub r2, r2, #1 ldcol.8 d1, r2, r1, 4 vpaddl.u8 d2, d1 vpadd.u16 d2, d2, d2 vrshrn.u16 d1, q1, #2 vmov.i8 d0, #128 vdup.8 d1, d1[0] b .L_pred8x8_dc_end .endfunc