diff --git a/main.c b/main.c
index a1cd6d4..9acb6a9 100644
--- a/main.c
+++ b/main.c
@@ -58,22 +58,25 @@ void drawTexInto(struct texture tex, struct texture buffer, struct rect texbound
    i32 stepy = 256 * (float)texheight / outheight;
    i32 texx = 256 * texbounds.x1;
    i32 texy = 256 * texbounds.y1;
+
+   i32 bufy = 3 * (bufbounds.x1 + bufbounds.y1 * buffer.width);
+   i32 bufyinc = 3 * buffer.width;
+   i32 bufi = bufy;
+
    //log("%f,%f step %f,%f", texx, texy, stepx, stepy);
    // Buffer is R G B for each pixel, then across then down. If you want X, Y,
    // it's: 3 * (x + y * width)
    for(int y = bufbounds.y1; y < bufbounds.y2; y++) {
       texx = texbounds.x1;
+      bufi = bufy;
       for(int x = bufbounds.x1; x < bufbounds.x2; x++) {
          i32 texi = 3 * ((texx >> 8) + tex.width * (texy >> 8));
-         i32 bufi = 3 * (x + buffer.width * y);
-         // You can get rid of this for loop for more performance, but probably
-         // not a big deal with -O3, it'll be unrolled
-         buffer.data[bufi] = tex.data[texi];
-         buffer.data[bufi + 1] = tex.data[texi + 1];
-         buffer.data[bufi + 2] = tex.data[texi + 2];
+         memcpy(buffer.data + bufi, tex.data + texi, 3);
          texx += stepx;
+         bufi += 3;
       }
       texy += stepy;
+      bufy += bufyinc;
    }
 }
 
@@ -98,10 +101,10 @@ int main() {
 
    // We actually stretch a texture here
    struct rect texrect = { 0, 0, 16, 16 };
-   struct rect bufrect = { 10, 10, 26, 26 };
+   struct rect bufrect = { 10, 10, 509, 499 };
 
    clock_t start = clock();
-   for(int i = 0; i < 50000; i++) { //overdraw factor
+   for(int i = 0; i < 50; i++) { //overdraw factor
       drawTexInto(tex, fb, texrect, bufrect);
    }
    clock_t end = clock();