| 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: Nedeljko Babic (nbabic@mips.com) |
| 30 | * |
| 31 | * This file is part of FFmpeg. |
| 32 | * |
| 33 | * FFmpeg is free software; you can redistribute it and/or |
| 34 | * modify it under the terms of the GNU Lesser General Public |
| 35 | * License as published by the Free Software Foundation; either |
| 36 | * version 2.1 of the License, or (at your option) any later version. |
| 37 | * |
| 38 | * FFmpeg is distributed in the hope that it will be useful, |
| 39 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 40 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 41 | * Lesser General Public License for more details. |
| 42 | * |
| 43 | * You should have received a copy of the GNU Lesser General Public |
| 44 | * License along with FFmpeg; if not, write to the Free Software |
| 45 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 46 | */ |
| 47 | |
| 48 | /** |
| 49 | * @file |
| 50 | * Reference: libavcodec/amrwbdec.c |
| 51 | */ |
| 52 | #include "libavutil/avutil.h" |
| 53 | #include "libavcodec/amrwbdata.h" |
| 54 | #include "amrwbdec_mips.h" |
| 55 | |
| 56 | #if HAVE_INLINE_ASM |
| 57 | void hb_fir_filter_mips(float *out, const float fir_coef[HB_FIR_SIZE + 1], |
| 58 | float mem[HB_FIR_SIZE], const float *in) |
| 59 | { |
| 60 | int i; |
| 61 | float data[AMRWB_SFR_SIZE_16k + HB_FIR_SIZE]; // past and current samples |
| 62 | |
| 63 | memcpy(data, mem, HB_FIR_SIZE * sizeof(float)); |
| 64 | memcpy(data + HB_FIR_SIZE, in, AMRWB_SFR_SIZE_16k * sizeof(float)); |
| 65 | |
| 66 | for (i = 0; i < AMRWB_SFR_SIZE_16k; i++) { |
| 67 | float output; |
| 68 | float * p_data = (data+i); |
| 69 | |
| 70 | /** |
| 71 | * inner loop is entirely unrolled and instructions are scheduled |
| 72 | * to minimize pipeline stall |
| 73 | */ |
| 74 | __asm__ volatile( |
| 75 | "mtc1 $zero, %[output] \n\t" |
| 76 | "lwc1 $f0, 0(%[p_data]) \n\t" |
| 77 | "lwc1 $f1, 0(%[fir_coef]) \n\t" |
| 78 | "lwc1 $f2, 4(%[p_data]) \n\t" |
| 79 | "madd.s %[output], %[output], $f0, $f1 \n\t" |
| 80 | "lwc1 $f3, 4(%[fir_coef]) \n\t" |
| 81 | "lwc1 $f4, 8(%[p_data]) \n\t" |
| 82 | "madd.s %[output], %[output], $f2, $f3 \n\t" |
| 83 | "lwc1 $f5, 8(%[fir_coef]) \n\t" |
| 84 | |
| 85 | "lwc1 $f0, 12(%[p_data]) \n\t" |
| 86 | "lwc1 $f1, 12(%[fir_coef]) \n\t" |
| 87 | "madd.s %[output], %[output], $f4, $f5 \n\t" |
| 88 | "lwc1 $f2, 16(%[p_data]) \n\t" |
| 89 | "madd.s %[output], %[output], $f0, $f1 \n\t" |
| 90 | "lwc1 $f3, 16(%[fir_coef]) \n\t" |
| 91 | "lwc1 $f4, 20(%[p_data]) \n\t" |
| 92 | "lwc1 $f5, 20(%[fir_coef]) \n\t" |
| 93 | "madd.s %[output], %[output], $f2, $f3 \n\t" |
| 94 | |
| 95 | "lwc1 $f0, 24(%[p_data]) \n\t" |
| 96 | "lwc1 $f1, 24(%[fir_coef]) \n\t" |
| 97 | "lwc1 $f2, 28(%[p_data]) \n\t" |
| 98 | "madd.s %[output], %[output], $f4, $f5 \n\t" |
| 99 | "lwc1 $f3, 28(%[fir_coef]) \n\t" |
| 100 | "madd.s %[output], %[output], $f0, $f1 \n\t" |
| 101 | "lwc1 $f4, 32(%[p_data]) \n\t" |
| 102 | "madd.s %[output], %[output], $f2, $f3 \n\t" |
| 103 | "lwc1 $f5, 32(%[fir_coef]) \n\t" |
| 104 | "madd.s %[output], %[output], $f4, $f5 \n\t" |
| 105 | |
| 106 | "lwc1 $f0, 36(%[p_data]) \n\t" |
| 107 | "lwc1 $f1, 36(%[fir_coef]) \n\t" |
| 108 | "lwc1 $f2, 40(%[p_data]) \n\t" |
| 109 | "lwc1 $f3, 40(%[fir_coef]) \n\t" |
| 110 | "madd.s %[output], %[output], $f0, $f1 \n\t" |
| 111 | "lwc1 $f4, 44(%[p_data]) \n\t" |
| 112 | "lwc1 $f5, 44(%[fir_coef]) \n\t" |
| 113 | "madd.s %[output], %[output], $f2, $f3 \n\t" |
| 114 | |
| 115 | "lwc1 $f0, 48(%[p_data]) \n\t" |
| 116 | "lwc1 $f1, 48(%[fir_coef]) \n\t" |
| 117 | "lwc1 $f2, 52(%[p_data]) \n\t" |
| 118 | "madd.s %[output], %[output], $f4, $f5 \n\t" |
| 119 | "lwc1 $f3, 52(%[fir_coef]) \n\t" |
| 120 | "lwc1 $f4, 56(%[p_data]) \n\t" |
| 121 | "madd.s %[output], %[output], $f0, $f1 \n\t" |
| 122 | "lwc1 $f5, 56(%[fir_coef]) \n\t" |
| 123 | "madd.s %[output], %[output], $f2, $f3 \n\t" |
| 124 | |
| 125 | "lwc1 $f0, 60(%[p_data]) \n\t" |
| 126 | "lwc1 $f1, 60(%[fir_coef]) \n\t" |
| 127 | "lwc1 $f2, 64(%[p_data]) \n\t" |
| 128 | "madd.s %[output], %[output], $f4, $f5 \n\t" |
| 129 | "lwc1 $f3, 64(%[fir_coef]) \n\t" |
| 130 | "madd.s %[output], %[output], $f0, $f1 \n\t" |
| 131 | "lwc1 $f4, 68(%[p_data]) \n\t" |
| 132 | "madd.s %[output], %[output], $f2, $f3 \n\t" |
| 133 | "lwc1 $f5, 68(%[fir_coef]) \n\t" |
| 134 | "madd.s %[output], %[output], $f4, $f5 \n\t" |
| 135 | |
| 136 | "lwc1 $f0, 72(%[p_data]) \n\t" |
| 137 | "lwc1 $f1, 72(%[fir_coef]) \n\t" |
| 138 | "lwc1 $f2, 76(%[p_data]) \n\t" |
| 139 | "lwc1 $f3, 76(%[fir_coef]) \n\t" |
| 140 | "madd.s %[output], %[output], $f0, $f1 \n\t" |
| 141 | "lwc1 $f4, 80(%[p_data]) \n\t" |
| 142 | "lwc1 $f5, 80(%[fir_coef]) \n\t" |
| 143 | "madd.s %[output], %[output], $f2, $f3 \n\t" |
| 144 | |
| 145 | "lwc1 $f0, 84(%[p_data]) \n\t" |
| 146 | "lwc1 $f1, 84(%[fir_coef]) \n\t" |
| 147 | "lwc1 $f2, 88(%[p_data]) \n\t" |
| 148 | "madd.s %[output], %[output], $f4, $f5 \n\t" |
| 149 | "lwc1 $f3, 88(%[fir_coef]) \n\t" |
| 150 | "lwc1 $f4, 92(%[p_data]) \n\t" |
| 151 | "madd.s %[output], %[output], $f0, $f1 \n\t" |
| 152 | "lwc1 $f5, 92(%[fir_coef]) \n\t" |
| 153 | "madd.s %[output], %[output], $f2, $f3 \n\t" |
| 154 | |
| 155 | "lwc1 $f0, 96(%[p_data]) \n\t" |
| 156 | "lwc1 $f1, 96(%[fir_coef]) \n\t" |
| 157 | "lwc1 $f2, 100(%[p_data]) \n\t" |
| 158 | "madd.s %[output], %[output], $f4, $f5 \n\t" |
| 159 | "lwc1 $f3, 100(%[fir_coef]) \n\t" |
| 160 | "lwc1 $f4, 104(%[p_data]) \n\t" |
| 161 | "madd.s %[output], %[output], $f0, $f1 \n\t" |
| 162 | "lwc1 $f5, 104(%[fir_coef]) \n\t" |
| 163 | "madd.s %[output], %[output], $f2, $f3 \n\t" |
| 164 | |
| 165 | "lwc1 $f0, 108(%[p_data]) \n\t" |
| 166 | "lwc1 $f1, 108(%[fir_coef]) \n\t" |
| 167 | "madd.s %[output], %[output], $f4, $f5 \n\t" |
| 168 | "lwc1 $f2, 112(%[p_data]) \n\t" |
| 169 | "lwc1 $f3, 112(%[fir_coef]) \n\t" |
| 170 | "madd.s %[output], %[output], $f0, $f1 \n\t" |
| 171 | "lwc1 $f4, 116(%[p_data]) \n\t" |
| 172 | "lwc1 $f5, 116(%[fir_coef]) \n\t" |
| 173 | "lwc1 $f0, 120(%[p_data]) \n\t" |
| 174 | "madd.s %[output], %[output], $f2, $f3 \n\t" |
| 175 | "lwc1 $f1, 120(%[fir_coef]) \n\t" |
| 176 | "madd.s %[output], %[output], $f4, $f5 \n\t" |
| 177 | "madd.s %[output], %[output], $f0, $f1 \n\t" |
| 178 | |
| 179 | : [output]"=&f"(output) |
| 180 | : [fir_coef]"r"(fir_coef), [p_data]"r"(p_data) |
| 181 | : "$f0", "$f1", "$f2", "$f3", "$f4", "$f5", "memory" |
| 182 | ); |
| 183 | out[i] = output; |
| 184 | } |
| 185 | memcpy(mem, data + AMRWB_SFR_SIZE_16k, HB_FIR_SIZE * sizeof(float)); |
| 186 | } |
| 187 | #endif /* HAVE_INLINE_ASM */ |