arm_convolution_example_f32.c
Go to the documentation of this file.
00001 /* ---------------------------------------------------------------------- 00002 * Copyright (C) 2011 ARM Limited. All rights reserved. 00003 * 00004 * $Date: 15. December 2011 00005 * $Revision: V2.0.0 00006 * 00007 * Project: Cortex-R4 DSP Library 00008 * Title: arm_convolution_example_f32.c 00009 * 00010 * Description: Example code demonstrating Convolution of two input signals using fft. 00011 * 00012 * Target Processor: Cortex-R4 00013 * 00014 * Version 2.0.0 2011/12/15 00015 * Final release. 00016 * 00017 * ---------------------------------------------------------------------------- */ 00018 00083 #include "arm_math.h" 00084 #include "math_helper.h" 00085 00086 /* ---------------------------------------------------------------------- 00087 * Defines each of the tests performed 00088 * ------------------------------------------------------------------- */ 00089 #define MAX_BLOCKSIZE 128 00090 #define DELTA (0.000001f) 00091 #define SNR_THRESHOLD 90 00092 00093 /* ---------------------------------------------------------------------- 00094 * Declare I/O buffers 00095 * ------------------------------------------------------------------- */ 00096 float32_t Ak[MAX_BLOCKSIZE]; /* Input A */ 00097 float32_t Bk[MAX_BLOCKSIZE]; /* Input B */ 00098 float32_t AxB[MAX_BLOCKSIZE * 2]; /* Output */ 00099 00100 /* ---------------------------------------------------------------------- 00101 * Test input data for Floating point Convolution example for 32-blockSize 00102 * Generated by the MATLAB randn() function 00103 * ------------------------------------------------------------------- */ 00104 float32_t testInputA_f32[64] = 00105 { 00106 -0.808920, 1.357369, 1.180861, -0.504544, 1.762637, -0.703285, 00107 1.696966, 0.620571, -0.151093, -0.100235, -0.872382, -0.403579, 00108 -0.860749, -0.382648, -1.052338, 0.128113, -0.646269, 1.093377, 00109 -2.209198, 0.471706, 0.408901, 1.266242, 0.598252, 1.176827, 00110 -0.203421, 0.213596, -0.851964, -0.466958, 0.021841, -0.698938, 00111 -0.604107, 0.461778, -0.318219, 0.942520, 0.577585, 0.417619, 00112 0.614665, 0.563679, -1.295073, -0.764437, 0.952194, -0.859222, 00113 -0.618554, -2.268542, -1.210592, 1.655853, -2.627219, -0.994249, 00114 -1.374704, 0.343799, 0.025619, 1.227481, -0.708031, 0.069355, 00115 -1.845228, -1.570886, 1.010668, -1.802084, 1.630088, 1.286090, 00116 -0.161050, -0.940794, 0.367961, 0.291907 00117 00118 }; 00119 00120 float32_t testInputB_f32[64] = 00121 { 00122 0.933724, 0.046881, 1.316470, 0.438345, 0.332682, 2.094885, 00123 0.512081, 0.035546, 0.050894, -2.320371, 0.168711, -1.830493, 00124 -0.444834, -1.003242, -0.531494, -1.365600, -0.155420, -0.757692, 00125 -0.431880, -0.380021, 0.096243, -0.695835, 0.558850, -1.648962, 00126 0.020369, -0.363630, 0.887146, 0.845503, -0.252864, -0.330397, 00127 1.269131, -1.109295, -1.027876, 0.135940, 0.116721, -0.293399, 00128 -1.349799, 0.166078, -0.802201, 0.369367, -0.964568, -2.266011, 00129 0.465178, 0.651222, -0.325426, 0.320245, -0.784178, -0.579456, 00130 0.093374, 0.604778, -0.048225, 0.376297, -0.394412, 0.578182, 00131 -1.218141, -1.387326, 0.692462, -0.631297, 0.153137, -0.638952, 00132 0.635474, -0.970468, 1.334057, -0.111370 00133 }; 00134 00135 const float testRefOutput_f32[126] = 00136 { 00137 -0.818943, 1.229484, -0.533664, 1.016604, 0.341875, -1.963656, 00138 5.171476, 3.478033, 7.616361, 6.648384, 0.479069, 1.792012, 00139 -1.295591, -7.447818, 0.315830, -10.657445, -2.483469, -6.524236, 00140 -7.380591, -3.739005, -8.388957, 0.184147, -1.554888, 3.786508, 00141 -1.684421, 5.400610, -1.578126, 7.403361, 8.315999, 2.080267, 00142 11.077776, 2.749673, 7.138962, 2.748762, 0.660363, 0.981552, 00143 1.442275, 0.552721, -2.576892, 4.703989, 0.989156, 8.759344, 00144 -0.564825, -3.994680, 0.954710, -5.014144, 6.592329, 1.599488, 00145 -13.979146, -0.391891, -4.453369, -2.311242, -2.948764, 1.761415, 00146 -0.138322, 10.433007, -2.309103, 4.297153, 8.535523, 3.209462, 00147 8.695819, 5.569919, 2.514304, 5.582029, 2.060199, 0.642280, 00148 7.024616, 1.686615, -6.481756, 1.343084, -3.526451, 1.099073, 00149 -2.965764, -0.173723, -4.111484, 6.528384, -6.965658, 1.726291, 00150 1.535172, 11.023435, 2.338401, -4.690188, 1.298210, 3.943885, 00151 8.407885, 5.168365, 0.684131, 1.559181, 1.859998, 2.852417, 00152 8.574070, -6.369078, 6.023458, 11.837963, -6.027632, 4.469678, 00153 -6.799093, -2.674048, 6.250367, -6.809971, -3.459360, 9.112410, 00154 -2.711621, -1.336678, 1.564249, -1.564297, -1.296760, 8.904013, 00155 -3.230109, 6.878013, -7.819823, 3.369909, -1.657410, -2.007358, 00156 -4.112825, 1.370685, -3.420525, -6.276605, 3.244873, -3.352638, 00157 1.545372, 0.902211, 0.197489, -1.408732, 0.523390, 0.348440 00158 }; 00159 00160 00161 /* ---------------------------------------------------------------------- 00162 * Declare Global variables 00163 * ------------------------------------------------------------------- */ 00164 uint32_t srcALen = 64; /* Length of Input A */ 00165 uint32_t srcBLen = 64; /* Length of Input B */ 00166 uint32_t outLen; /* Length of convolution output */ 00167 float32_t snr; /* output SNR */ 00168 00169 int32_t main(void) 00170 { 00171 arm_status status; /* Status of the example */ 00172 arm_cfft_radix4_instance_f32 cfft_instance; /* CFFT Structure instance */ 00173 00174 /* CFFT Structure instance pointer */ 00175 arm_cfft_radix4_instance_f32 *cfft_instance_ptr = 00176 (arm_cfft_radix4_instance_f32*) &cfft_instance; 00177 00178 /* output length of convolution */ 00179 outLen = srcALen + srcBLen - 1; 00180 00181 /* Initialise the fft input buffers with all zeros */ 00182 arm_fill_f32(0.0, Ak, MAX_BLOCKSIZE); 00183 arm_fill_f32(0.0, Bk, MAX_BLOCKSIZE); 00184 00185 /* Copy the input values to the fft input buffers */ 00186 arm_copy_f32(testInputA_f32, Ak, MAX_BLOCKSIZE/2); 00187 arm_copy_f32(testInputB_f32, Bk, MAX_BLOCKSIZE/2); 00188 00189 /* Initialize the CFFT function to compute 64 point fft */ 00190 status = arm_cfft_radix4_init_f32(cfft_instance_ptr, 64, 0, 1); 00191 00192 /* Transform input a[n] from time domain to frequency domain A[k] */ 00193 arm_cfft_radix4_f32(cfft_instance_ptr, Ak); 00194 /* Transform input b[n] from time domain to frequency domain B[k] */ 00195 arm_cfft_radix4_f32(cfft_instance_ptr, Bk); 00196 00197 /* Complex Multiplication of the two input buffers in frequency domain */ 00198 arm_cmplx_mult_cmplx_f32(Ak, Bk, AxB, MAX_BLOCKSIZE/2); 00199 00200 /* Initialize the CIFFT function to compute 64 point ifft */ 00201 status = arm_cfft_radix4_init_f32(cfft_instance_ptr, 64, 1, 1); 00202 00203 /* Transform the multiplication output from frequency domain to time domain, 00204 that gives the convolved output */ 00205 arm_cfft_radix4_f32(cfft_instance_ptr, AxB); 00206 00207 /* SNR Calculation */ 00208 snr = arm_snr_f32((float32_t *)testRefOutput_f32, AxB, srcALen + srcBLen - 1); 00209 00210 /* Compare the SNR with threshold to test whether the 00211 computed output is matched with the reference output values. */ 00212 if( snr > SNR_THRESHOLD) 00213 { 00214 status = ARM_MATH_SUCCESS; 00215 } 00216 00217 if( status != ARM_MATH_SUCCESS) 00218 { 00219 while(1); 00220 } 00221 } 00222
