ARM: fix NEON synth_filter_float with hardfp calls
[ffmpeg.git] / libavcodec / arm / synth_filter_neon.S
1 /*
2  * Copyright (c) 2010 Mans Rullgard <mans@mansr.com>
3  *
4  * This file is part of FFmpeg.
5  *
6  * FFmpeg is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU Lesser General Public
8  * License as published by the Free Software Foundation; either
9  * version 2.1 of the License, or (at your option) any later version.
10  *
11  * FFmpeg is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14  * Lesser General Public License for more details.
15  *
16  * You should have received a copy of the GNU Lesser General Public
17  * License along with FFmpeg; if not, write to the Free Software
18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19  */
20
21 #include "asm.S"
22
23         preserve8
24
25 function ff_synth_filter_float_neon, export=1
26         push            {r3-r11,lr}
27
28         ldr             r4,  [r2]               @ synth_buf_offset
29         add             r1,  r1,  r4,  lsl #2   @ synth_buf
30         sub             r12, r4,  #32
31         bfc             r12, #9,  #23
32         bic             r4,  r4,  #63
33         str             r12, [r2]
34
35         ldr             r2,  [sp, #12*4]        @ in
36         mov             r9,  r1                 @ synth_buf
37
38 VFP     vpush           {d0}
39         bl              ff_imdct_half_neon
40 VFP     vpop            {d0}
41         pop             {r3}
42
43         ldr             r5,  [sp, #9*4]         @ window
44         ldr             r2,  [sp, #10*4]        @ out
45 NOVFP   vldr            d0,  [sp, #12*4]        @ scale, bias
46         add             r8,  r9,  #12*4
47
48         mov             lr,  #64*4
49         mov             r1,  #4
50 1:
51         add             r10, r9,  #16*4         @ synth_buf
52         add             r11, r8,  #16*4
53         add             r0,  r5,  #16*4         @ window
54         add             r6,  r5,  #32*4
55         add             r7,  r5,  #48*4
56
57         vld1.32         {q10},    [r3,:128]     @ a
58         add             r3,  r3,  #16*4
59         vld1.32         {q1},     [r3,:128]     @ b
60         vmov.f32        q2,  #0.0               @ c
61         vmov.f32        q3,  #0.0               @ d
62
63         mov             r12, #512
64 2:
65         vld1.32         {q9},     [r8, :128], lr
66         vrev64.32       q9,  q9
67         vld1.32         {q8},     [r5, :128], lr
68         vmls.f32        d20, d16, d19
69         vld1.32         {q11},    [r0, :128], lr
70         vmls.f32        d21, d17, d18
71         vld1.32         {q12},    [r9, :128], lr
72         vmla.f32        d2,  d22, d24
73         vld1.32         {q8},     [r6, :128], lr
74         vmla.f32        d3,  d23, d25
75         vld1.32         {q9},     [r10,:128], lr
76         vmla.f32        d4,  d16, d18
77         vld1.32         {q12},    [r11,:128], lr
78         vmla.f32        d5,  d17, d19
79         vrev64.32       q12, q12
80         vld1.32         {q11},    [r7, :128], lr
81         vmla.f32        d6,  d22, d25
82         vmla.f32        d7,  d23, d24
83         subs            r12, r12, #64
84         beq             3f
85         cmp             r12, r4
86         bne             2b
87         sub             r8,  r8,  #512*4
88         sub             r9,  r9,  #512*4
89         sub             r10, r10, #512*4
90         sub             r11, r11, #512*4
91         b               2b
92 3:
93         vdup.32         q8,  d0[1]
94         vdup.32         q9,  d0[1]
95         vmla.f32        q8,  q10, d0[0]
96         vmla.f32        q9,  q1,  d0[0]
97         vst1.32         {q3},     [r3,:128]
98         sub             r3,  r3,  #16*4
99         vst1.32         {q2},     [r3,:128]
100         vst1.32         {q8},     [r2,:128]
101         add             r2,  r2,  #16*4
102         vst1.32         {q9},     [r2,:128]
103
104         subs            r1,  r1,  #1
105         popeq           {r4-r11,pc}
106
107         cmp             r4,  #0
108         subeq           r8,  r8,  #512*4
109         subeq           r9,  r9,  #512*4
110         sub             r5,  r5,  #512*4
111         sub             r2,  r2,  #12*4         @ out
112         add             r3,  r3,  #4*4          @ synth_buf2
113         add             r5,  r5,  #4*4          @ window
114         add             r9,  r9,  #4*4          @ synth_buf
115         sub             r8,  r8,  #4*4          @ synth_buf
116         b               1b
117 endfunc