1 /* 2 * Copyright (C) 2010 The Android Open Source Project 3 * 4 * Licensed under the Apache License, Version 2.0 (the "License"); 5 * you may not use this file except in compliance with the License. 6 * You may obtain a copy of the License at 7 * 8 * http://www.apache.org/licenses/LICENSE-2.0 9 * 10 * Unless required by applicable law or agreed to in writing, software 11 * distributed under the License is distributed on an "AS IS" BASIS, 12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 13 * See the License for the specific language governing permissions and 14 * limitations under the License. 15 * 16 */ 17 #include <jni.h> 18 #include <time.h> 19 #include <stdio.h> 20 #include <stdlib.h> 21 #include <cpu-features.h> 22 #include "helloneon-intrinsics.h" 23 24 #define DEBUG 0 25 26 #if DEBUG 27 #include <android/log.h> 28 # define D(x...) __android_log_print(ANDROID_LOG_INFO,"helloneon",x) 29 #else 30 # define D(...) do {} while (0) 31 #endif 32 33 /* return current time in milliseconds */ 34 static double 35 now_ms(void) 36 { 37 struct timespec res; 38 clock_gettime(CLOCK_REALTIME, &res); 39 return 1000.0*res.tv_sec + (double)res.tv_nsec/1e6; 40 } 41 42 43 /* this is a FIR filter implemented in C */ 44 static void 45 fir_filter_c(short *output, const short* input, const short* kernel, int width, int kernelSize) 46 { 47 int offset = -kernelSize/2; 48 int nn; 49 for (nn = 0; nn < width; nn++) { 50 int sum = 0; 51 int mm; 52 for (mm = 0; mm < kernelSize; mm++) { 53 sum += kernel[mm]*input[nn+offset+mm]; 54 } 55 output[nn] = (short)((sum + 0x8000) >> 16); 56 } 57 } 58 59 #define FIR_KERNEL_SIZE 32 60 #define FIR_OUTPUT_SIZE 2560 61 #define FIR_INPUT_SIZE (FIR_OUTPUT_SIZE + FIR_KERNEL_SIZE) 62 #define FIR_ITERATIONS 600 63 64 static const short fir_kernel[FIR_KERNEL_SIZE] = { 65 0x10, 0x20, 0x40, 0x70, 0x8c, 0xa2, 0xce, 0xf0, 0xe9, 0xce, 0xa2, 0x8c, 070, 0x40, 0x20, 0x10, 66 0x10, 0x20, 0x40, 0x70, 0x8c, 0xa2, 0xce, 0xf0, 0xe9, 0xce, 0xa2, 0x8c, 070, 0x40, 0x20, 0x10 }; 67 68 static short fir_output[FIR_OUTPUT_SIZE]; 69 static short fir_input_0[FIR_INPUT_SIZE]; 70 static const short* fir_input = fir_input_0 + (FIR_KERNEL_SIZE/2); 71 static short fir_output_expected[FIR_OUTPUT_SIZE]; 72 73 /* This is a trivial JNI example where we use a native method 74 * to return a new VM String. See the corresponding Java source 75 * file located at: 76 * 77 * apps/samples/hello-neon/project/src/com/example/neon/HelloNeon.java 78 */ 79 jstring 80 Java_com_example_neon_HelloNeon_stringFromJNI( JNIEnv* env, 81 jobject thiz ) 82 { 83 char* str; 84 uint64_t features; 85 char buffer[512]; 86 char tryNeon = 0; 87 double t0, t1, time_c, time_neon; 88 89 /* setup FIR input - whatever */ 90 { 91 int nn; 92 for (nn = 0; nn < FIR_INPUT_SIZE; nn++) { 93 fir_input_0[nn] = (5*nn) & 255; 94 } 95 fir_filter_c(fir_output_expected, fir_input, fir_kernel, FIR_OUTPUT_SIZE, FIR_KERNEL_SIZE); 96 } 97 98 /* Benchmark small FIR filter loop - C version */ 99 t0 = now_ms(); 100 { 101 int count = FIR_ITERATIONS; 102 for (; count > 0; count--) { 103 fir_filter_c(fir_output, fir_input, fir_kernel, FIR_OUTPUT_SIZE, FIR_KERNEL_SIZE); 104 } 105 } 106 t1 = now_ms(); 107 time_c = t1 - t0; 108 109 asprintf(&str, "FIR Filter benchmark:\nC version : %g ms\n", time_c); 110 strlcpy(buffer, str, sizeof buffer); 111 free(str); 112 113 strlcat(buffer, "Neon version : ", sizeof buffer); 114 115 if (android_getCpuFamily() != ANDROID_CPU_FAMILY_ARM) { 116 strlcat(buffer, "Not an ARM CPU !\n", sizeof buffer); 117 goto EXIT; 118 } 119 120 features = android_getCpuFeatures(); 121 if ((features & ANDROID_CPU_ARM_FEATURE_ARMv7) == 0) { 122 strlcat(buffer, "Not an ARMv7 CPU !\n", sizeof buffer); 123 goto EXIT; 124 } 125 126 /* HAVE_NEON is defined in Android.mk ! */ 127 #ifdef HAVE_NEON 128 if ((features & ANDROID_CPU_ARM_FEATURE_NEON) == 0) { 129 strlcat(buffer, "CPU doesn't support NEON !\n", sizeof buffer); 130 goto EXIT; 131 } 132 133 /* Benchmark small FIR filter loop - Neon version */ 134 t0 = now_ms(); 135 { 136 int count = FIR_ITERATIONS; 137 for (; count > 0; count--) { 138 fir_filter_neon_intrinsics(fir_output, fir_input, fir_kernel, FIR_OUTPUT_SIZE, FIR_KERNEL_SIZE); 139 } 140 } 141 t1 = now_ms(); 142 time_neon = t1 - t0; 143 asprintf(&str, "%g ms (x%g faster)\n", time_neon, time_c / (time_neon < 1e-6 ? 1. : time_neon)); 144 strlcat(buffer, str, sizeof buffer); 145 free(str); 146 147 /* check the result, just in case */ 148 { 149 int nn, fails = 0; 150 for (nn = 0; nn < FIR_OUTPUT_SIZE; nn++) { 151 if (fir_output[nn] != fir_output_expected[nn]) { 152 if (++fails < 16) 153 D("neon[%d] = %d expected %d", nn, fir_output[nn], fir_output_expected[nn]); 154 } 155 } 156 D("%d fails\n", fails); 157 } 158 #else /* !HAVE_NEON */ 159 strlcat(buffer, "Program not compiled with ARMv7 support !\n", sizeof buffer); 160 #endif /* !HAVE_NEON */ 161 EXIT: 162 return (*env)->NewStringUTF(env, buffer); 163 } 164