Drive-by optimization of blur_surface()

clients/cairo-util.c

@@ -49,9 +49,11 @@ blur_surface(cairo_surface_t *surface, int margin)
 	dst = malloc(height * stride);
 
 	half = size / 2;
+	a = 0;
 	for (i = 0; i < size; i++) {
 		f = (i - half);
 		kernel[i] = exp(- f * f / ARRAY_LENGTH(kernel)) * 10000;
+		a += kernel[i];
 	}
 
 	for (i = 0; i < height; i++) {
@@ -67,7 +69,6 @@ blur_surface(cairo_surface_t *surface, int margin)
 			y = 0;
 			z = 0;
 			w = 0;
-			a = 0;
 			for (k = 0; k < size; k++) {
 				if (j - half + k < 0 || j - half + k >= width)
 					continue;
@@ -77,7 +78,6 @@ blur_surface(cairo_surface_t *surface, int margin)
 				y += ((p >> 16) & 0xff) * kernel[k];
 				z += ((p >> 8) & 0xff) * kernel[k];
 				w += (p & 0xff) * kernel[k];
-				a += kernel[k];
 			}
 			d[j] = (x / a << 24) | (y / a << 16) | (z / a << 8) | w / a;
 		}
@@ -96,7 +96,6 @@ blur_surface(cairo_surface_t *surface, int margin)
 			y = 0;
 			z = 0;
 			w = 0;
-			a = 0;
 			for (k = 0; k < size; k++) {
 				if (i - half + k < 0 || i - half + k >= height)
 					continue;
@@ -107,7 +106,6 @@ blur_surface(cairo_surface_t *surface, int margin)
 				y += ((p >> 16) & 0xff) * kernel[k];
 				z += ((p >> 8) & 0xff) * kernel[k];
 				w += (p & 0xff) * kernel[k];
-				a += kernel[k];
 			}
 			d[j] = (x / a << 24) | (y / a << 16) | (z / a << 8) | w / a;
 		}