nikola/GRA-Project
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Restructured versions and file names - Changed main implementation

Browse Source
This commit is contained in:
Yashar Alan Zenker
2026-07-04 20:13:20 +02:00
parent 9c28454b7e
commit b14cf7a79c
11 changed files with 384 additions and 380 deletions
Download Patch File Download Diff File Expand all files Collapse all files
Implementierung/tricorn_V1.c
+30 -95
View File
@@ -1,11 +1,10 @@
#include "tricorn_V1.h"
#include <immintrin.h>
static unsigned converge(float c_rea, float c_imag, unsigned n);
static const float convergenceLimit = 2.0f; // Mandelbrot set limit
static unsigned converge(float complex position,unsigned n);
/**
* @brief Generates a Tricorn fractal image using highly optimized SSE/SIMD instructions, however the algoriithm is not fully optimized.
* @brief Generates a Tricorn fractal image using the standard, unoptimized scalar implementation for benchmarking comparison (Version 2).
* @param start The starting complex coordinate (bottom-left or top-left depending on height sign).
* @param width The width of the target image in pixels.
* @param height The height of the target image. A negative height reverses the Y-axis calculation.
@@ -15,112 +14,48 @@ static unsigned converge(float c_rea, float c_imag, unsigned n);
* @param img Pointer to the pre-allocated memory buffer where the bitmap data will be stored.
*/
void tricorn_V1(float complex start, size_t width, ssize_t height, float res, unsigned n, _Bool color, unsigned char* img){
int bytesPerColor = color ? 3 : 1;
// Absolute height for iteration purposes
ssize_t actual_height = height < 0 ? (ssize_t)(-height) : height;
ssize_t imag_step = height < 0 ? -1:1;
size_t simd_width = (width / 4) * 4;
float imag_direction = height < 0 ? -1.0f : 1.0f;
unsigned bytesPerColor = color ? 3 : 1;
height = height < 0? -height : height;
float c_start_real = creal(start);
float c_start_imag = cimag(start);
// Simple iteration over every pixel in the picture
for(size_t i = 0; i < width; i++){
for(ssize_t j = 0; j < height; j++){
__m128 four = _mm_set1_ps(4.0f);
__m128 minus_two = _mm_set1_ps(-2.0f);
unsigned value = converge(start + (i*res - I* imag_direction*j*res),n);
for(ssize_t y = 0; y < actual_height;y++) {
float c_imag = c_start_imag - (y*imag_step) * res;
__m128 c_imag_vector = _mm_set1_ps(c_imag);
// Pixel value position in the output array
size_t position = (j*width +i) * bytesPerColor;
size_t x = 0;
for(; x < simd_width;x += 4) { //SIMD Loop
float c_real = c_start_real + (x*res);
__m128 c_real_vector = _mm_setr_ps(c_real,c_real+1.0f*res,c_real+2.0f*res,c_real+3.0f*res);
__m128 real = _mm_setzero_ps();
__m128 imag = _mm_setzero_ps();
__m128i iterations = _mm_setzero_si128();
__m128i active_mask = _mm_set1_epi32(-1);
for(unsigned i = 0; i < n;i++) {
__m128 real2 = _mm_mul_ps(real,real);
__m128 imag2 = _mm_mul_ps(imag,imag);
__m128 unbounded = _mm_cmple_ps(_mm_add_ps(real2,imag2),four);
__m128i unbounded_i = _mm_castps_si128(unbounded);
active_mask = _mm_and_si128(active_mask,unbounded_i);
if(_mm_test_all_zeros(active_mask,active_mask)) {
break;
}
iterations = _mm_sub_epi32(iterations,active_mask);
imag = _mm_add_ps(c_imag_vector,_mm_mul_ps(minus_two, _mm_mul_ps(real,imag)));
real = _mm_add_ps(_mm_sub_ps(real2,imag2),c_real_vector);
if(!color){ // Grayscale colors as specified by the assignment r = g = b
img[position] = value;
}
__m128i mod64 = _mm_and_si128(iterations, _mm_set1_epi32(63));
__m128i times4 = _mm_slli_epi32(mod64, 2);
__m128i values = _mm_sub_epi32(_mm_set1_epi32(255), times4);
__m128i max_iter = _mm_set1_epi32(n);
__m128i is_max = _mm_cmpeq_epi32(iterations, max_iter);
values = _mm_andnot_si128(is_max, values);
unsigned int value_array[4];
_mm_storeu_si128((__m128i*)value_array,values);
for(unsigned i = 0 ; i < 4;i++) {
unsigned int value = value_array[i];
size_t position = (y*width + (x+i)) * bytesPerColor;
if(!color) {
img[position] = value;
}
else {
img[position] = ((value*3) % 255);
img[position + 1] = ((value*2) % 255);
img[position + 2] = ((value*5) % 255);
}
}
}
for(;x < width; x++) {
float c_real = c_start_real + (x*res);
unsigned int value = converge(c_real,c_imag,n);
size_t position = (y*width + x) * bytesPerColor;
if(!color) {
img[position] = value;
}
else {
else{ // Temporary colors with random multipliers
img[position] = ((value*3) % 255);
img[position + 1] = ((value*2) % 255);
img[position + 1] = ((value*2) % 255) ;
img[position + 2] = ((value*5) % 255);
}
}
}
}
}
static unsigned converge(float c_real, float c_imag, unsigned n) {
float real = 0.0f;
float imag = 0.0f;
// Iterative application of the equation given by the assignment
static unsigned converge(float complex position,unsigned n){
for (unsigned i = 0; i < n; i++) {
float real2 = real * real;
float imag2 = imag * imag;
complex float result = 0;
for(unsigned i = 0; i < n; i++){
// Detection of unbounded behaviour
// If the magnitude of the result vector is greater than the convergence limit, we know that the function will grow unbounded
// Current convergence limit taken from mandelbrot set
if(sqrtf(powf (crealf(result),2)+powf(cimagf(result),2)) > convergenceLimit)
return 255-4*(i % 64);
if (real2 + imag2 > 4.0f) {
return 255 - 4 * (i % 64);
}
imag = -2.0f * real * imag + c_imag;
real = real2 - imag2 + c_real;
// Application of the given complex equation with m = 2
result = cpowf(conjf(result),2) + position;
}
return 0;
return 0;
}