1 /* Copyright (c) 2015 Xiph.Org Foundation 2 Written by Viswanath Puttagunta */ 3 /** 4 @file celt_ne10_fft.c 5 @brief ARM Neon optimizations for fft using NE10 library 6 */ 7 8 /* 9 Redistribution and use in source and binary forms, with or without 10 modification, are permitted provided that the following conditions 11 are met: 12 13 - Redistributions of source code must retain the above copyright 14 notice, this list of conditions and the following disclaimer. 15 16 - Redistributions in binary form must reproduce the above copyright 17 notice, this list of conditions and the following disclaimer in the 18 documentation and/or other materials provided with the distribution. 19 20 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 21 ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 22 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 23 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER 24 OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 25 EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 26 PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 27 PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 28 LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 29 NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 30 SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33 #ifndef SKIP_CONFIG_H 34 #ifdef HAVE_CONFIG_H 35 #include "config.h" 36 #endif 37 #endif 38 39 #include <NE10_dsp.h> 40 #include "os_support.h" 41 #include "kiss_fft.h" 42 #include "stack_alloc.h" 43 44 #if !defined(FIXED_POINT) 45 # define NE10_FFT_ALLOC_C2C_TYPE_NEON ne10_fft_alloc_c2c_float32_neon 46 # define NE10_FFT_CFG_TYPE_T ne10_fft_cfg_float32_t 47 # define NE10_FFT_STATE_TYPE_T ne10_fft_state_float32_t 48 # define NE10_FFT_DESTROY_C2C_TYPE ne10_fft_destroy_c2c_float32 49 # define NE10_FFT_CPX_TYPE_T ne10_fft_cpx_float32_t 50 # define NE10_FFT_C2C_1D_TYPE_NEON ne10_fft_c2c_1d_float32_neon 51 #else 52 # define NE10_FFT_ALLOC_C2C_TYPE_NEON(nfft) ne10_fft_alloc_c2c_int32_neon(nfft) 53 # define NE10_FFT_CFG_TYPE_T ne10_fft_cfg_int32_t 54 # define NE10_FFT_STATE_TYPE_T ne10_fft_state_int32_t 55 # define NE10_FFT_DESTROY_C2C_TYPE ne10_fft_destroy_c2c_int32 56 # define NE10_FFT_DESTROY_C2C_TYPE ne10_fft_destroy_c2c_int32 57 # define NE10_FFT_CPX_TYPE_T ne10_fft_cpx_int32_t 58 # define NE10_FFT_C2C_1D_TYPE_NEON ne10_fft_c2c_1d_int32_neon 59 #endif 60 61 #if defined(CUSTOM_MODES) 62 63 /* nfft lengths in NE10 that support scaled fft */ 64 # define NE10_FFTSCALED_SUPPORT_MAX 4 65 static const int ne10_fft_scaled_support[NE10_FFTSCALED_SUPPORT_MAX] = { 66 480, 240, 120, 60 67 }; 68 69 int opus_fft_alloc_arm_neon(kiss_fft_state *st) 70 { 71 int i; 72 size_t memneeded = sizeof(struct arch_fft_state); 73 74 st->arch_fft = (arch_fft_state *)opus_alloc(memneeded); 75 if (!st->arch_fft) 76 return -1; 77 78 for (i = 0; i < NE10_FFTSCALED_SUPPORT_MAX; i++) { 79 if(st->nfft == ne10_fft_scaled_support[i]) 80 break; 81 } 82 if (i == NE10_FFTSCALED_SUPPORT_MAX) { 83 /* This nfft length (scaled fft) is not supported in NE10 */ 84 st->arch_fft->is_supported = 0; 85 st->arch_fft->priv = NULL; 86 } 87 else { 88 st->arch_fft->is_supported = 1; 89 st->arch_fft->priv = (void *)NE10_FFT_ALLOC_C2C_TYPE_NEON(st->nfft); 90 if (st->arch_fft->priv == NULL) { 91 return -1; 92 } 93 } 94 return 0; 95 } 96 97 void opus_fft_free_arm_neon(kiss_fft_state *st) 98 { 99 NE10_FFT_CFG_TYPE_T cfg; 100 101 if (!st->arch_fft) 102 return; 103 104 cfg = (NE10_FFT_CFG_TYPE_T)st->arch_fft->priv; 105 if (cfg) 106 NE10_FFT_DESTROY_C2C_TYPE(cfg); 107 opus_free(st->arch_fft); 108 } 109 #endif 110 111 void opus_fft_neon(const kiss_fft_state *st, 112 const kiss_fft_cpx *fin, 113 kiss_fft_cpx *fout) 114 { 115 NE10_FFT_STATE_TYPE_T state; 116 NE10_FFT_CFG_TYPE_T cfg = &state; 117 VARDECL(NE10_FFT_CPX_TYPE_T, buffer); 118 SAVE_STACK; 119 ALLOC(buffer, st->nfft, NE10_FFT_CPX_TYPE_T); 120 121 if (!st->arch_fft->is_supported) { 122 /* This nfft length (scaled fft) not supported in NE10 */ 123 opus_fft_c(st, fin, fout); 124 } 125 else { 126 memcpy((void *)cfg, st->arch_fft->priv, sizeof(NE10_FFT_STATE_TYPE_T)); 127 state.buffer = (NE10_FFT_CPX_TYPE_T *)&buffer[0]; 128 #if !defined(FIXED_POINT) 129 state.is_forward_scaled = 1; 130 131 NE10_FFT_C2C_1D_TYPE_NEON((NE10_FFT_CPX_TYPE_T *)fout, 132 (NE10_FFT_CPX_TYPE_T *)fin, 133 cfg, 0); 134 #else 135 NE10_FFT_C2C_1D_TYPE_NEON((NE10_FFT_CPX_TYPE_T *)fout, 136 (NE10_FFT_CPX_TYPE_T *)fin, 137 cfg, 0, 1); 138 #endif 139 } 140 RESTORE_STACK; 141 } 142 143 void opus_ifft_neon(const kiss_fft_state *st, 144 const kiss_fft_cpx *fin, 145 kiss_fft_cpx *fout) 146 { 147 NE10_FFT_STATE_TYPE_T state; 148 NE10_FFT_CFG_TYPE_T cfg = &state; 149 VARDECL(NE10_FFT_CPX_TYPE_T, buffer); 150 SAVE_STACK; 151 ALLOC(buffer, st->nfft, NE10_FFT_CPX_TYPE_T); 152 153 if (!st->arch_fft->is_supported) { 154 /* This nfft length (scaled fft) not supported in NE10 */ 155 opus_ifft_c(st, fin, fout); 156 } 157 else { 158 memcpy((void *)cfg, st->arch_fft->priv, sizeof(NE10_FFT_STATE_TYPE_T)); 159 state.buffer = (NE10_FFT_CPX_TYPE_T *)&buffer[0]; 160 #if !defined(FIXED_POINT) 161 state.is_backward_scaled = 0; 162 163 NE10_FFT_C2C_1D_TYPE_NEON((NE10_FFT_CPX_TYPE_T *)fout, 164 (NE10_FFT_CPX_TYPE_T *)fin, 165 cfg, 1); 166 #else 167 NE10_FFT_C2C_1D_TYPE_NEON((NE10_FFT_CPX_TYPE_T *)fout, 168 (NE10_FFT_CPX_TYPE_T *)fin, 169 cfg, 1, 0); 170 #endif 171 } 172 RESTORE_STACK; 173 } 174
