foreach[_tiled] seems to work now

examples_cuda/mandelbrot_tasks3d/mandel_cu.cpp

@@ -142,7 +142,7 @@ int main(int argc, char *argv[]) {
   deviceFree(d_buf);
 
   printf("[mandelbrot ispc+tasks]:\t[%.3f] million cycles\n", minISPC);
-  writePPM(buf, width, height, "mandelbrot-ispc.ppm");
+  writePPM(buf, width, height, "mandelbrot-cuda.ppm");
 
 
   // 

examples_cuda/mandelbrot_tasks3d/mandelbrot_tasks3d.ispc

@@ -73,7 +73,8 @@ mandelbrot_scanline(uniform float x0, uniform float dx,
 
     const uniform int ystart = taskIndex1 * yspan;
     const uniform int yend   = min(ystart  + yspan, height);
-    
+   
+#if 0 
   for (uniform int yi = ystart; yi < yend; yi++)
     for (uniform int xi = xstart; xi < xend; xi += programCount)
     {
@@ -85,6 +86,17 @@ mandelbrot_scanline(uniform float x0, uniform float dx,
       if (xi + programIndex < xend)
         output[index] = res;
     }
+#else
+    foreach (yi = ystart ... yend, xi = xstart ... xend)
+    {
+      const float x = x0 + xi * dx;
+      const float y = y0 + yi * dy;
+
+      const int res = mandel(x,y,maxIterations);
+      const int index = yi * width + xi;
+      output[index] = res;
+    }
+#endif
     
 }