forked from pi7mcrg2k/opcomplex
Merge pull request '把数据扩大，并做了一些优化' (#2) from p8sljnpht/opcomplex:main into main
commit
783f0bf04d
@ -0,0 +1,33 @@
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <time.h>
|
||||
#include <string.h>
|
||||
#include "render.h"
|
||||
|
||||
void applyGaussianBlur(float src[][MAT_SIZE], float dst[][MAT_SIZE], float kernel[3][3]);
|
||||
int main() {
|
||||
float inputImage[MAT_SIZE][MAT_SIZE];
|
||||
Render(inputImage);
|
||||
float kernel[3][3] = {
|
||||
{1.0f/16, 2.0f/16, 1.0f/16},
|
||||
{2.0f/16, 4.0f/16, 2.0f/16},
|
||||
{1.0f/16, 2.0f/16, 1.0f/16}
|
||||
};
|
||||
float outputImage[MAT_SIZE][MAT_SIZE]={0};
|
||||
clock_t start = clock();
|
||||
applyGaussianBlur(inputImage, outputImage, kernel);
|
||||
clock_t end = clock();
|
||||
printf("Time: %lf s\n", (double)(end-start) / CLOCKS_PER_SEC);
|
||||
Print(outputImage);
|
||||
}
|
||||
|
||||
void applyGaussianBlur(float src[][MAT_SIZE], float dst[][MAT_SIZE], float kernel[3][3]) {
|
||||
int i, j;
|
||||
for(i=1; i<MAT_SIZE-1; i++)
|
||||
for(j=1; j<MAT_SIZE-1; j++) {
|
||||
dst[i][j] =
|
||||
src[i-1][j-1]*kernel[0][0]+src[i-1][j ]*kernel[0][1]+src[i-1][j+1]*kernel[0][2]+
|
||||
src[i ][j-1]*kernel[1][0]+src[i ][j ]*kernel[1][1]+src[i ][j+1]*kernel[1][2]+
|
||||
src[i+1][j-1]*kernel[2][0]+src[i+1][j ]*kernel[2][1]+src[i+1][j+1]*kernel[2][2];
|
||||
}
|
||||
}
|
||||
@ -0,0 +1,38 @@
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <time.h>
|
||||
#include <string.h>
|
||||
#include "render.h"
|
||||
#define IDX(n) ((n) % 3)
|
||||
|
||||
void applySeparableGaussianBlur(float[][MAT_SIZE], float[][MAT_SIZE], float[3], float[3]);
|
||||
int main() {
|
||||
float inputImage[MAT_SIZE][MAT_SIZE];
|
||||
Render(inputImage);
|
||||
|
||||
float kx[3] = {1.0f/4, 1.0f/2, 1.0f/4};
|
||||
float ky[3] = {1.0f/4, 1.0f/2, 1.0f/4};
|
||||
float outputImage[MAT_SIZE][MAT_SIZE] = {0};
|
||||
clock_t start = clock();
|
||||
applySeparableGaussianBlur(inputImage, outputImage, kx, ky);
|
||||
clock_t end = clock();
|
||||
printf("Time: %lf s\n", (double)(end-start) / CLOCKS_PER_SEC);
|
||||
Print(outputImage);
|
||||
}
|
||||
|
||||
void applySeparableGaussianBlur(float src[][MAT_SIZE], float dst[][MAT_SIZE], float kx[3], float ky[3]) {
|
||||
float buf[3][MAT_SIZE+3];
|
||||
int i, j;
|
||||
// 计算前两行的行内卷积
|
||||
for(i=0; i<2; i++)
|
||||
for(j=1; j<MAT_SIZE-1; j++) {
|
||||
buf[i][j] = src[i][j-1]*kx[0]+src[i][j]*kx[1]+src[i][j+1]*kx[2];
|
||||
}
|
||||
// 利用buf进行累计
|
||||
for(i=1; i<MAT_SIZE-1; i++) {
|
||||
for(j=1; j<MAT_SIZE-1; j++)
|
||||
buf[IDX(i+1)][j] = src[i+1][j-1]*kx[0]+src[i+1][j]*kx[1]+src[i+1][j+1]*kx[2];
|
||||
for(j=1; j<MAT_SIZE-1; j++)
|
||||
dst[i][j] = buf[IDX(i-1)][j]*ky[0]+buf[IDX(i)][j]*ky[1]+buf[IDX(i+1)][j]*ky[2];
|
||||
}
|
||||
}
|
||||
@ -0,0 +1,61 @@
|
||||
#include <stdio.h>
|
||||
#include <stdlib.h>
|
||||
#include <time.h>
|
||||
#include <string.h>
|
||||
#include <arm_neon.h>
|
||||
#include "render.h"
|
||||
#define IDX(n) ((n) % 3)
|
||||
|
||||
void applySeparableGaussianBlur(float src[][MAT_SIZE], float dst[][MAT_SIZE],
|
||||
float kx[], float ky[]);
|
||||
int main() {
|
||||
float inputImage[MAT_SIZE][MAT_SIZE];
|
||||
Render(inputImage);
|
||||
float kernel[3][3] = {
|
||||
{1.0f/16, 2.0f/16, 1.0f/16},
|
||||
{2.0f/16, 4.0f/16, 2.0f/16},
|
||||
{1.0f/16, 2.0f/16, 1.0f/16}
|
||||
};
|
||||
float kx[4] = {1.0f/4, 1.0f/2, 1.0f/4, 0.0f}; // 防止越界多定义一个
|
||||
float ky[4] = {1.0f/4, 1.0f/2, 1.0f/4, 0.0f};
|
||||
float outputImage[MAT_SIZE][MAT_SIZE] = {0};
|
||||
clock_t start = clock();
|
||||
applySeparableGaussianBlur(inputImage, outputImage, kx, ky);
|
||||
clock_t end = clock();
|
||||
printf("Time: %lf s\n", (double)(end-start) / CLOCKS_PER_SEC);
|
||||
Print(outputImage);
|
||||
}
|
||||
|
||||
void applySeparableGaussianBlur(float src[][MAT_SIZE], float dst[][MAT_SIZE],
|
||||
float kx[], float ky[]) {
|
||||
int i, j;
|
||||
float buf[3][MAT_SIZE];
|
||||
float32x4_t kx_vec = vld1q_f32(kx);
|
||||
float32x4_t ky_vec = vld1q_f32(ky);
|
||||
// 计算前两行的行内卷积
|
||||
float32x4_t* left, *mid, *right, *result;
|
||||
for(i=0; i<2; i++)
|
||||
for(j=1; j<MAT_SIZE-1; j+=4) {
|
||||
left = (float32x4_t*)&src[i][j-1];
|
||||
mid = (float32x4_t*)&src[i][j];
|
||||
right = (float32x4_t*)&src[i][j+1];
|
||||
result = (float32x4_t*)&buf[i][j];
|
||||
*result = vmulq_lane_f32(*left, vget_low_f32(kx_vec), 0);
|
||||
*result += vmulq_lane_f32(*mid, vget_low_f32(kx_vec), 1);
|
||||
*result += vmulq_lane_f32(*right, vget_high_f32(kx_vec), 0);
|
||||
}
|
||||
// 利用buf进行累计
|
||||
for(i=1; i<MAT_SIZE-1; i++) {
|
||||
for(j=1; j<MAT_SIZE-1; j+=4) {
|
||||
left = (float32x4_t*)&src[i+1][j-1];
|
||||
mid = (float32x4_t*)&src[i+1][j];
|
||||
right = (float32x4_t*)&src[i+1][j+1];
|
||||
result = (float32x4_t*)&buf[IDX(i+1)][j];
|
||||
*result = vmulq_lane_f32(*left, vget_low_f32(kx_vec), 0);
|
||||
*result += vmulq_lane_f32(*mid, vget_low_f32(kx_vec), 1);
|
||||
*result += vmulq_lane_f32(*right, vget_high_f32(kx_vec), 0);
|
||||
}
|
||||
for(j=1; j<MAT_SIZE-1; j++)
|
||||
dst[i][j] = buf[IDX(i-1)][j]*ky[0]+buf[IDX(i)][j]*ky[1]+buf[IDX(i+1)][j]*ky[2];
|
||||
}
|
||||
}
|
||||
@ -0,0 +1,40 @@
|
||||
#ifndef __RENDER_H
|
||||
#define __RENDER_H
|
||||
#pragma GCC optimize ("O1")
|
||||
#include <stdlib.h>
|
||||
#include <time.h>
|
||||
|
||||
#define MAT_SIZE 514
|
||||
#define RAND_SEED 114514
|
||||
void Render(float a[][MAT_SIZE]);
|
||||
void Print(float a[][MAT_SIZE]);
|
||||
|
||||
void Render(float a[][MAT_SIZE])
|
||||
{
|
||||
srand(RAND_SEED);
|
||||
int i, j;
|
||||
for(i=0; i<MAT_SIZE; i++)
|
||||
for(j=0; j<MAT_SIZE; j++) {
|
||||
a[i][j] = (float)rand() / (float)RAND_MAX;
|
||||
}
|
||||
}
|
||||
|
||||
void Print(float a[][MAT_SIZE])
|
||||
{
|
||||
printf("Matrix with Size=(%d,%d)\n", MAT_SIZE, MAT_SIZE);
|
||||
printf("%5.1f %5.1f %5.1f ... %5.1f %5.1f %5.1f\n", a[0][0], a[0][1], a[0][2],
|
||||
a[0][MAT_SIZE-3], a[0][MAT_SIZE-2], a[0][MAT_SIZE-1]);
|
||||
printf("%5.1f %5.1f %5.1f ... %5.1f %5.1f %5.1f\n", a[1][0], a[1][1], a[1][2],
|
||||
a[1][MAT_SIZE-3], a[1][MAT_SIZE-2], a[1][MAT_SIZE-1]);
|
||||
printf("%5.1f %5.1f %5.1f ... %5.1f %5.1f %5.1f\n", a[2][0], a[2][1], a[2][2],
|
||||
a[2][MAT_SIZE-3], a[2][MAT_SIZE-2], a[2][MAT_SIZE-1]);
|
||||
printf(" ... ... ... ... ... ... ... \n");
|
||||
printf("%5.1f %5.1f %5.1f ... %5.1f %5.1f %5.1f\n", a[MAT_SIZE-3][0], a[MAT_SIZE-3][1], a[MAT_SIZE-3][2],
|
||||
a[MAT_SIZE-3][MAT_SIZE-3], a[MAT_SIZE-3][MAT_SIZE-2], a[MAT_SIZE-3][MAT_SIZE-1]);
|
||||
printf("%5.1f %5.1f %5.1f ... %5.1f %5.1f %5.1f\n", a[MAT_SIZE-2][0], a[MAT_SIZE-2][1], a[MAT_SIZE-2][2],
|
||||
a[MAT_SIZE-2][MAT_SIZE-3], a[MAT_SIZE-2][MAT_SIZE-2], a[MAT_SIZE-2][MAT_SIZE-1]);
|
||||
printf("%5.1f %5.1f %5.1f ... %5.1f %5.1f %5.1f\n", a[MAT_SIZE-1][0], a[MAT_SIZE-1][1], a[MAT_SIZE-1][2],
|
||||
a[MAT_SIZE-1][MAT_SIZE-3], a[MAT_SIZE-1][MAT_SIZE-2], a[MAT_SIZE-1][MAT_SIZE-1]);
|
||||
}
|
||||
#endif
|
||||
|
||||
Loading…
Reference in new issue