Merge commit 'cebdedca57d95834a8f8098c7b6a322a1163e26b'
[ffmpeg.git] / libavcodec / mips / compute_antialias_float.h
1 /*
2  * Copyright (c) 2012
3  *      MIPS Technologies, Inc., California.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the MIPS Technologies, Inc., nor the names of its
14  *    contributors may be used to endorse or promote products derived from
15  *    this software without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE MIPS TECHNOLOGIES, INC. ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE MIPS TECHNOLOGIES, INC. BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  * Author:  Bojan Zivkovic (bojan@mips.com)
30  *
31  * Compute antialias function optimised for MIPS floating-point architecture
32  *
33  * This file is part of FFmpeg.
34  *
35  * FFmpeg is free software; you can redistribute it and/or
36  * modify it under the terms of the GNU Lesser General Public
37  * License as published by the Free Software Foundation; either
38  * version 2.1 of the License, or (at your option) any later version.
39  *
40  * FFmpeg is distributed in the hope that it will be useful,
41  * but WITHOUT ANY WARRANTY; without even the implied warranty of
42  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
43  * Lesser General Public License for more details.
44  *
45  * You should have received a copy of the GNU Lesser General Public
46  * License along with FFmpeg; if not, write to the Free Software
47  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
48  */
49
50 /**
51  * @file
52  * Reference: libavcodec/mpegaudiodec.c
53  */
54
55 #ifndef AVCODEC_MIPS_COMPUTE_ANTIALIAS_FLOAT_H
56 #define AVCODEC_MIPS_COMPUTE_ANTIALIAS_FLOAT_H
57
58 #if HAVE_INLINE_ASM
59 static void compute_antialias_mips_float(MPADecodeContext *s,
60                                         GranuleDef *g)
61 {
62     float *ptr, *ptr_end;
63     float *csa = &csa_table[0][0];
64     int n;
65     /* temporary variables */
66     float in1, in2, in3, in4, in5, in6, in7, in8;
67     float out1, out2, out3, out4;
68
69     ptr = g->sb_hybrid + 18;
70     /* we antialias only "long" bands */
71     if (g->block_type == 2) {
72         if (!g->switch_point)
73             return;
74         /* XXX: check this for 8000Hz case */
75         n = 1;
76         ptr_end = ptr + 18;
77     } else {
78         n = 31;
79         ptr_end = ptr + 558;
80     }
81
82     /**
83     * instructions are scheduled to minimize pipeline stall.
84     */
85
86     __asm__ volatile (
87         "compute_antialias_float_loop%=:                                \t\n"
88         "lwc1    %[in1],  -1*4(%[ptr])                                  \t\n"
89         "lwc1    %[in2],  0(%[csa])                                     \t\n"
90         "lwc1    %[in3],  1*4(%[csa])                                   \t\n"
91         "lwc1    %[in4],  0(%[ptr])                                     \t\n"
92         "lwc1    %[in5],  -2*4(%[ptr])                                  \t\n"
93         "lwc1    %[in6],  4*4(%[csa])                                   \t\n"
94         "mul.s   %[out1], %[in1],  %[in2]                               \t\n"
95         "mul.s   %[out2], %[in1],  %[in3]                               \t\n"
96         "lwc1    %[in7],  5*4(%[csa])                                   \t\n"
97         "lwc1    %[in8],  1*4(%[ptr])                                   \t\n"
98         "nmsub.s %[out1], %[out1], %[in3], %[in4]                       \t\n"
99         "madd.s  %[out2], %[out2], %[in2], %[in4]                       \t\n"
100         "mul.s   %[out3], %[in5],  %[in6]                               \t\n"
101         "mul.s   %[out4], %[in5],  %[in7]                               \t\n"
102         "lwc1    %[in1],  -3*4(%[ptr])                                  \t\n"
103         "swc1    %[out1], -1*4(%[ptr])                                  \t\n"
104         "swc1    %[out2], 0(%[ptr])                                     \t\n"
105         "nmsub.s %[out3], %[out3], %[in7], %[in8]                       \t\n"
106         "madd.s  %[out4], %[out4], %[in6], %[in8]                       \t\n"
107         "lwc1    %[in2],  8*4(%[csa])                                   \t\n"
108         "swc1    %[out3], -2*4(%[ptr])                                  \t\n"
109         "swc1    %[out4], 1*4(%[ptr])                                   \t\n"
110         "lwc1    %[in3],  9*4(%[csa])                                   \t\n"
111         "lwc1    %[in4],  2*4(%[ptr])                                   \t\n"
112         "mul.s   %[out1], %[in1],  %[in2]                               \t\n"
113         "lwc1    %[in5],  -4*4(%[ptr])                                  \t\n"
114         "lwc1    %[in6],  12*4(%[csa])                                  \t\n"
115         "mul.s   %[out2], %[in1],  %[in3]                               \t\n"
116         "lwc1    %[in7],  13*4(%[csa])                                  \t\n"
117         "nmsub.s %[out1], %[out1], %[in3], %[in4]                       \t\n"
118         "lwc1    %[in8],  3*4(%[ptr])                                   \t\n"
119         "mul.s   %[out3], %[in5],  %[in6]                               \t\n"
120         "madd.s  %[out2], %[out2], %[in2], %[in4]                       \t\n"
121         "mul.s   %[out4], %[in5],  %[in7]                               \t\n"
122         "swc1    %[out1], -3*4(%[ptr])                                  \t\n"
123         "lwc1    %[in1],  -5*4(%[ptr])                                  \t\n"
124         "nmsub.s %[out3], %[out3], %[in7], %[in8]                       \t\n"
125         "swc1    %[out2], 2*4(%[ptr])                                   \t\n"
126         "madd.s  %[out4], %[out4], %[in6], %[in8]                       \t\n"
127         "lwc1    %[in2],  16*4(%[csa])                                  \t\n"
128         "lwc1    %[in3],  17*4(%[csa])                                  \t\n"
129         "swc1    %[out3], -4*4(%[ptr])                                  \t\n"
130         "lwc1    %[in4],  4*4(%[ptr])                                   \t\n"
131         "swc1    %[out4], 3*4(%[ptr])                                   \t\n"
132         "mul.s   %[out1], %[in1],  %[in2]                               \t\n"
133         "mul.s   %[out2], %[in1],  %[in3]                               \t\n"
134         "lwc1    %[in5],  -6*4(%[ptr])                                  \t\n"
135         "lwc1    %[in6],  20*4(%[csa])                                  \t\n"
136         "lwc1    %[in7],  21*4(%[csa])                                  \t\n"
137         "nmsub.s %[out1], %[out1], %[in3], %[in4]                       \t\n"
138         "madd.s  %[out2], %[out2], %[in2], %[in4]                       \t\n"
139         "lwc1    %[in8],  5*4(%[ptr])                                   \t\n"
140         "mul.s   %[out3], %[in5],  %[in6]                               \t\n"
141         "mul.s   %[out4], %[in5],  %[in7]                               \t\n"
142         "swc1    %[out1], -5*4(%[ptr])                                  \t\n"
143         "swc1    %[out2], 4*4(%[ptr])                                   \t\n"
144         "lwc1    %[in1],  -7*4(%[ptr])                                  \t\n"
145         "nmsub.s %[out3], %[out3], %[in7], %[in8]                       \t\n"
146         "madd.s  %[out4], %[out4], %[in6], %[in8]                       \t\n"
147         "lwc1    %[in2],  24*4(%[csa])                                  \t\n"
148         "lwc1    %[in3],  25*4(%[csa])                                  \t\n"
149         "lwc1    %[in4],  6*4(%[ptr])                                   \t\n"
150         "swc1    %[out3], -6*4(%[ptr])                                  \t\n"
151         "swc1    %[out4], 5*4(%[ptr])                                   \t\n"
152         "mul.s   %[out1], %[in1],  %[in2]                               \t\n"
153         "lwc1    %[in5],  -8*4(%[ptr])                                  \t\n"
154         "mul.s   %[out2], %[in1],  %[in3]                               \t\n"
155         "lwc1    %[in6],  28*4(%[csa])                                  \t\n"
156         "lwc1    %[in7],  29*4(%[csa])                                  \t\n"
157         "nmsub.s %[out1], %[out1], %[in3], %[in4]                       \t\n"
158         "lwc1    %[in8],  7*4(%[ptr])                                   \t\n"
159         "madd.s  %[out2], %[out2], %[in2], %[in4]                       \t\n"
160         "mul.s   %[out3], %[in5],  %[in6]                               \t\n"
161         "mul.s   %[out4], %[in5],  %[in7]                               \t\n"
162         "swc1    %[out1], -7*4(%[ptr])                                  \t\n"
163         "swc1    %[out2], 6*4(%[ptr])                                   \t\n"
164         "addiu   %[ptr],  %[ptr],  72                                   \t\n"
165         "nmsub.s %[out3], %[out3], %[in7], %[in8]                       \t\n"
166         "madd.s  %[out4], %[out4], %[in6], %[in8]                       \t\n"
167         "swc1    %[out3], -26*4(%[ptr])                                 \t\n"
168         "swc1    %[out4], -11*4(%[ptr])                                 \t\n"
169         "bne     %[ptr],  %[ptr_end],  compute_antialias_float_loop%=   \t\n"
170
171         : [ptr] "+r" (ptr),
172           [in1] "=&f" (in1), [in2] "=&f" (in2),
173           [in3] "=&f" (in3), [in4] "=&f" (in4),
174           [in5] "=&f" (in5), [in6] "=&f" (in6),
175           [in7] "=&f" (in7), [in8] "=&f" (in8),
176           [out1] "=&f" (out1), [out2] "=&f" (out2),
177           [out3] "=&f" (out3), [out4] "=&f" (out4)
178         : [csa] "r" (csa), [ptr_end] "r" (ptr_end)
179         : "memory"
180     );
181 }
182 #define compute_antialias compute_antialias_mips_float
183 #endif /* HAVE_INLINE_ASM */
184
185 #endif /* AVCODEC_MIPS_COMPUTE_ANTIALIAS_FLOAT_H */