downsample_fast_mips.c (10324B)
1 /* 2 * Copyright (c) 2013 The WebRTC project authors. All Rights Reserved. 3 * 4 * Use of this source code is governed by a BSD-style license 5 * that can be found in the LICENSE file in the root of the source 6 * tree. An additional intellectual property rights grant can be found 7 * in the file PATENTS. All contributing project authors may 8 * be found in the AUTHORS file in the root of the source tree. 9 */ 10 11 #include "common_audio/signal_processing/include/signal_processing_library.h" 12 13 // Version of WebRtcSpl_DownsampleFast() for MIPS platforms. 14 int WebRtcSpl_DownsampleFast_mips(const int16_t* data_in, 15 size_t data_in_length, 16 int16_t* data_out, 17 size_t data_out_length, 18 const int16_t* __restrict coefficients, 19 size_t coefficients_length, 20 int factor, 21 size_t delay) { 22 int i; 23 int j; 24 int k; 25 int32_t out_s32 = 0; 26 size_t endpos = delay + factor * (data_out_length - 1) + 1; 27 28 int32_t tmp1, tmp2, tmp3, tmp4, factor_2; 29 int16_t* p_coefficients; 30 int16_t* p_data_in; 31 int16_t* p_data_in_0 = (int16_t*)&data_in[delay]; 32 int16_t* p_coefficients_0 = (int16_t*)&coefficients[0]; 33 #if !defined(MIPS_DSP_R1_LE) 34 int32_t max_16 = 0x7FFF; 35 int32_t min_16 = 0xFFFF8000; 36 #endif // #if !defined(MIPS_DSP_R1_LE) 37 38 // Return error if any of the running conditions doesn't meet. 39 if (data_out_length == 0 || coefficients_length == 0 || 40 data_in_length < endpos) { 41 return -1; 42 } 43 #if defined(MIPS_DSP_R2_LE) 44 __asm __volatile( 45 ".set push \n\t" 46 ".set noreorder \n\t" 47 "subu %[i], %[endpos], %[delay] \n\t" 48 "sll %[factor_2], %[factor], 1 \n\t" 49 "1: \n\t" 50 "move %[p_data_in], %[p_data_in_0] \n\t" 51 "mult $zero, $zero \n\t" 52 "move %[p_coefs], %[p_coefs_0] \n\t" 53 "sra %[j], %[coef_length], 2 \n\t" 54 "beq %[j], $zero, 3f \n\t" 55 " andi %[k], %[coef_length], 3 \n\t" 56 "2: \n\t" 57 "lwl %[tmp1], 1(%[p_data_in]) \n\t" 58 "lwl %[tmp2], 3(%[p_coefs]) \n\t" 59 "lwl %[tmp3], -3(%[p_data_in]) \n\t" 60 "lwl %[tmp4], 7(%[p_coefs]) \n\t" 61 "lwr %[tmp1], -2(%[p_data_in]) \n\t" 62 "lwr %[tmp2], 0(%[p_coefs]) \n\t" 63 "lwr %[tmp3], -6(%[p_data_in]) \n\t" 64 "lwr %[tmp4], 4(%[p_coefs]) \n\t" 65 "packrl.ph %[tmp1], %[tmp1], %[tmp1] \n\t" 66 "packrl.ph %[tmp3], %[tmp3], %[tmp3] \n\t" 67 "dpa.w.ph $ac0, %[tmp1], %[tmp2] \n\t" 68 "dpa.w.ph $ac0, %[tmp3], %[tmp4] \n\t" 69 "addiu %[j], %[j], -1 \n\t" 70 "addiu %[p_data_in], %[p_data_in], -8 \n\t" 71 "bgtz %[j], 2b \n\t" 72 " addiu %[p_coefs], %[p_coefs], 8 \n\t" 73 "3: \n\t" 74 "beq %[k], $zero, 5f \n\t" 75 " nop \n\t" 76 "4: \n\t" 77 "lhu %[tmp1], 0(%[p_data_in]) \n\t" 78 "lhu %[tmp2], 0(%[p_coefs]) \n\t" 79 "addiu %[p_data_in], %[p_data_in], -2 \n\t" 80 "addiu %[k], %[k], -1 \n\t" 81 "dpa.w.ph $ac0, %[tmp1], %[tmp2] \n\t" 82 "bgtz %[k], 4b \n\t" 83 " addiu %[p_coefs], %[p_coefs], 2 \n\t" 84 "5: \n\t" 85 "extr_r.w %[out_s32], $ac0, 12 \n\t" 86 "addu %[p_data_in_0], %[p_data_in_0], %[factor_2] \n\t" 87 "subu %[i], %[i], %[factor] \n\t" 88 "shll_s.w %[out_s32], %[out_s32], 16 \n\t" 89 "sra %[out_s32], %[out_s32], 16 \n\t" 90 "sh %[out_s32], 0(%[data_out]) \n\t" 91 "bgtz %[i], 1b \n\t" 92 " addiu %[data_out], %[data_out], 2 \n\t" 93 ".set pop \n\t" 94 : [tmp1] "=&r"(tmp1), [tmp2] "=&r"(tmp2), [tmp3] "=&r"(tmp3), 95 [tmp4] "=&r"(tmp4), [p_data_in] "=&r"(p_data_in), 96 [p_data_in_0] "+r"(p_data_in_0), [p_coefs] "=&r"(p_coefficients), 97 [j] "=&r"(j), [out_s32] "=&r"(out_s32), [factor_2] "=&r"(factor_2), 98 [i] "=&r"(i), [k] "=&r"(k) 99 : [coef_length] "r"(coefficients_length), [data_out] "r"(data_out), 100 [p_coefs_0] "r"(p_coefficients_0), [endpos] "r"(endpos), 101 [delay] "r"(delay), [factor] "r"(factor) 102 : "memory", "hi", "lo"); 103 #else // #if defined(MIPS_DSP_R2_LE) 104 __asm __volatile( 105 ".set push \n\t" 106 ".set noreorder \n\t" 107 "sll %[factor_2], %[factor], 1 \n\t" 108 "subu %[i], %[endpos], %[delay] \n\t" 109 "1: \n\t" 110 "move %[p_data_in], %[p_data_in_0] \n\t" 111 "addiu %[out_s32], $zero, 2048 \n\t" 112 "move %[p_coefs], %[p_coefs_0] \n\t" 113 "sra %[j], %[coef_length], 1 \n\t" 114 "beq %[j], $zero, 3f \n\t" 115 " andi %[k], %[coef_length], 1 \n\t" 116 "2: \n\t" 117 "lh %[tmp1], 0(%[p_data_in]) \n\t" 118 "lh %[tmp2], 0(%[p_coefs]) \n\t" 119 "lh %[tmp3], -2(%[p_data_in]) \n\t" 120 "lh %[tmp4], 2(%[p_coefs]) \n\t" 121 "mul %[tmp1], %[tmp1], %[tmp2] \n\t" 122 "addiu %[p_coefs], %[p_coefs], 4 \n\t" 123 "mul %[tmp3], %[tmp3], %[tmp4] \n\t" 124 "addiu %[j], %[j], -1 \n\t" 125 "addiu %[p_data_in], %[p_data_in], -4 \n\t" 126 "addu %[tmp1], %[tmp1], %[tmp3] \n\t" 127 "bgtz %[j], 2b \n\t" 128 " addu %[out_s32], %[out_s32], %[tmp1] \n\t" 129 "3: \n\t" 130 "beq %[k], $zero, 4f \n\t" 131 " nop \n\t" 132 "lh %[tmp1], 0(%[p_data_in]) \n\t" 133 "lh %[tmp2], 0(%[p_coefs]) \n\t" 134 "mul %[tmp1], %[tmp1], %[tmp2] \n\t" 135 "addu %[out_s32], %[out_s32], %[tmp1] \n\t" 136 "4: \n\t" 137 "sra %[out_s32], %[out_s32], 12 \n\t" 138 "addu %[p_data_in_0], %[p_data_in_0], %[factor_2] \n\t" 139 #if defined(MIPS_DSP_R1_LE) 140 "shll_s.w %[out_s32], %[out_s32], 16 \n\t" 141 "sra %[out_s32], %[out_s32], 16 \n\t" 142 #else // #if defined(MIPS_DSP_R1_LE) 143 "slt %[tmp1], %[max_16], %[out_s32] \n\t" 144 "movn %[out_s32], %[max_16], %[tmp1] \n\t" 145 "slt %[tmp1], %[out_s32], %[min_16] \n\t" 146 "movn %[out_s32], %[min_16], %[tmp1] \n\t" 147 #endif // #if defined(MIPS_DSP_R1_LE) 148 "subu %[i], %[i], %[factor] \n\t" 149 "sh %[out_s32], 0(%[data_out]) \n\t" 150 "bgtz %[i], 1b \n\t" 151 " addiu %[data_out], %[data_out], 2 \n\t" 152 ".set pop \n\t" 153 : [tmp1] "=&r"(tmp1), [tmp2] "=&r"(tmp2), [tmp3] "=&r"(tmp3), 154 [tmp4] "=&r"(tmp4), [p_data_in] "=&r"(p_data_in), [k] "=&r"(k), 155 [p_data_in_0] "+r"(p_data_in_0), [p_coefs] "=&r"(p_coefficients), 156 [j] "=&r"(j), [out_s32] "=&r"(out_s32), [factor_2] "=&r"(factor_2), 157 [i] "=&r"(i) 158 : [coef_length] "r"(coefficients_length), [data_out] "r"(data_out), 159 [p_coefs_0] "r"(p_coefficients_0), [endpos] "r"(endpos), 160 #if !defined(MIPS_DSP_R1_LE) 161 [max_16] "r"(max_16), [min_16] "r"(min_16), 162 #endif // #if !defined(MIPS_DSP_R1_LE) 163 [delay] "r"(delay), [factor] "r"(factor) 164 : "memory", "hi", "lo"); 165 #endif // #if defined(MIPS_DSP_R2_LE) 166 return 0; 167 }