Optimized SIMD implementation - V0

Shifted V1,V2 to V2,V3 and made V1 be the old not fully optimized SIMD implementation

Implementierung/tricorn_V2.c

@@ -1,10 +1,10 @@
 #include "tricorn_V2.h"
 
-static const float convergenceLimit = 2.0f; // Mandelbrot set limit
-static unsigned converge(float complex position,unsigned n);
+static unsigned converge(float c_real, float c_imag, unsigned n);
+
 
 /**
- * @brief Generates a Tricorn fractal image using the standard, unoptimized scalar implementation for benchmarking comparison (Version 2).
+ * @brief Generates a Tricorn fractal image using an optimized SISD implementation.
  * @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.
@@ -14,20 +14,30 @@ static unsigned converge(float complex position,unsigned n);
  * @param img Pointer to the pre-allocated memory buffer where the bitmap data will be stored.
  */
 void tricorn_V2(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
-    height = height < 0 ?  height*-1 : height;
-    unsigned bytesPerColor = color ? 3 : 1;
+    ssize_t actual_height = height < 0 ?  (ssize_t)(-height) : height;
+    ssize_t imag_step = height < 0 ? -1:1;
+
+    float c_start_real = creal(start);
+    float c_start_imag = cimag(start);
+
+    float c_imag = c_start_imag;
+    size_t position = 0;
+
 
     // Simple iteration over every pixel in the picture
-    for(size_t i = 0; i < width; i++){
-        for(ssize_t j = 0; j < height; j++){
+    for(ssize_t y = 0; y < actual_height; y++){
 
-            unsigned value = converge(start + (i*res - I*j*res),n);
+        // Calculate imaginary part on each row as it stays the same
+
+        float c_real = c_start_real;
+
+        for(size_t x = 0; x < width; x++){
+            unsigned value = converge(c_real,c_imag,n);
 
             // Pixel value position in the output array
-            size_t position = (j*width +i) * bytesPerColor;
-
 
             if(!color){ // Grayscale colors as specified by the assignment r = g = b
                 img[position] =  value;
@@ -38,24 +48,29 @@ void tricorn_V2(float complex start, size_t width, ssize_t height, float res, un
                 img[position + 2] = ((value*5) % 255);
             }
 
+            position += bytesPerColor;
+            c_real += res;
+
         }
+        c_imag -= imag_step * res;
     }
 }
 
 // Iterative application of the equation given by the assignment
-static unsigned converge(float complex position,unsigned n){
+static inline unsigned converge(float c_real, float c_imag, unsigned n) {
+    float real = 0.0f;
+    float imag = 0.0f;
 
-    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);
+    for (unsigned i = 0; i < n; i++) {
+        float real2 = real * real;
+        float imag2 = imag * imag;
 
-        // Application of the given complex equation with m = 2
-        result = cpowf(conjf(result),2) + position;
-       
+        if (real2 + imag2 > 4.0f) {
+            return 255 - 4 * (i % 64);
+        }
+
+        imag = -2.0f * real * imag + c_imag;
+        real = real2 - imag2 + c_real;
     }
     return 0;
 }