diff --git a/examples/stencil/stencil.cpp b/examples/stencil/stencil.cpp
index a3ce5712..98ff937a 100644
--- a/examples/stencil/stencil.cpp
+++ b/examples/stencil/stencil.cpp
@@ -116,20 +116,38 @@ int main() {
     InitData(Nx, Ny, Nz, Aispc, vsq);
 
     //
-    // Compute the image using the ispc implementation; report the minimum
-    // time of three runs.
+    // Compute the image using the ispc implementation on one core; report
+    // the minimum time of three runs.
     //
-    double minISPC = 1e30;
+    double minTimeISPC = 1e30;
     for (int i = 0; i < 3; ++i) {
         reset_and_start_timer();
         loop_stencil_ispc(0, 6, width, Nx - width, width, Ny - width,
                           width, Nz - width, Nx, Ny, Nz, coeff, vsq,
                           Aispc[0], Aispc[1]);
         double dt = get_elapsed_mcycles();
-        minISPC = std::min(minISPC, dt);
+        minTimeISPC = std::min(minTimeISPC, dt);
     }
 
-    printf("[stencil ispc]:\t\t\t[%.3f] million cycles\n", minISPC);
+    printf("[stencil ispc 1 core]:\t\t[%.3f] million cycles\n", minTimeISPC);
+
+    InitData(Nx, Ny, Nz, Aispc, vsq);
+
+    //
+    // Compute the image using the ispc implementation with tasks; report
+    // the minimum time of three runs.
+    //
+    double minTimeISPCTasks = 1e30;
+    for (int i = 0; i < 3; ++i) {
+        reset_and_start_timer();
+        loop_stencil_ispc_tasks(0, 6, width, Nx - width, width, Ny - width,
+                                width, Nz - width, Nx, Ny, Nz, coeff, vsq,
+                                Aispc[0], Aispc[1]);
+        double dt = get_elapsed_mcycles();
+        minTimeISPCTasks = std::min(minTimeISPCTasks, dt);
+    }
+
+    printf("[stencil ispc + tasks]:\t\t[%.3f] million cycles\n", minTimeISPCTasks);
 
     InitData(Nx, Ny, Nz, Aserial, vsq);
 
@@ -137,19 +155,20 @@ int main() {
     // And run the serial implementation 3 times, again reporting the
     // minimum time.
     //
-    double minSerial = 1e30;
+    double minTimeSerial = 1e30;
     for (int i = 0; i < 3; ++i) {
         reset_and_start_timer();
         loop_stencil_serial(0, 6, width, Nx-width, width, Ny - width,
                             width, Nz - width, Nx, Ny, Nz, coeff, vsq,
                             Aserial[0], Aserial[1]);
         double dt = get_elapsed_mcycles();
-        minSerial = std::min(minSerial, dt);
+        minTimeSerial = std::min(minTimeSerial, dt);
     }
 
-    printf("[stencil serial]:\t\t[%.3f] millon cycles\n", minSerial);
+    printf("[stencil serial]:\t\t[%.3f] millon cycles\n", minTimeSerial);
 
-    printf("\t\t\t\t(%.2fx speedup from ISPC)\n", minSerial/minISPC);
+    printf("\t\t\t\t(%.2fx speedup from ISPC, %.2f from ISPC + tasks)\n", 
+           minTimeSerial / minTimeISPC, minTimeSerial / minTimeISPCTasks);
 
     // Check for agreement
     int offset = 0;
diff --git a/examples/stencil/stencil.ispc b/examples/stencil/stencil.ispc
index bb618c2b..d707640c 100644
--- a/examples/stencil/stencil.ispc
+++ b/examples/stencil/stencil.ispc
@@ -32,7 +32,7 @@
 */
 
 
-static task void
+static void
 stencil_step(uniform int x0, uniform int x1,
              uniform int y0, uniform int y1,
              uniform int z0, uniform int z1,
@@ -67,14 +67,26 @@ stencil_step(uniform int x0, uniform int x1,
 }
 
 
-export void loop_stencil_ispc(uniform int t0, uniform int t1, 
-                              uniform int x0, uniform int x1,
-                              uniform int y0, uniform int y1,
-                              uniform int z0, uniform int z1,
-                              uniform int Nx, uniform int Ny, uniform int Nz,
-                              uniform const float coef[4], 
-                              uniform const float vsq[],
-                              uniform float Aeven[], uniform float Aodd[])
+static task void
+stencil_step_task(uniform int x0, uniform int x1,
+                  uniform int y0, uniform int y1,
+                  uniform int z0, uniform int z1,
+                  uniform int Nx, uniform int Ny, uniform int Nz,
+                  uniform const float coef[4], uniform const float vsq[],
+                  uniform const float Ain[], uniform float Aout[]) {
+    stencil_step(x0, x1, y0, y1, z0, z1, Nx, Ny, Nz, coef, vsq, Ain, Aout);
+}
+
+
+export void
+loop_stencil_ispc_tasks(uniform int t0, uniform int t1, 
+                        uniform int x0, uniform int x1,
+                        uniform int y0, uniform int y1,
+                        uniform int z0, uniform int z1,
+                        uniform int Nx, uniform int Ny, uniform int Nz,
+                        uniform const float coef[4], 
+                        uniform const float vsq[],
+                        uniform float Aeven[], uniform float Aodd[])
 {
     for (uniform int t = t0; t < t1; ++t) {
         // Parallelize across cores as well: each task will work on a slice
@@ -83,14 +95,35 @@ export void loop_stencil_ispc(uniform int t0, uniform int t1,
         uniform int dz = 1;
         for (uniform int z = z0; z < z1; z += dz) {
             if ((t & 1) == 0)
-                launch < stencil_step(x0, x1, y0, y1, z, z+dz, Nx, Ny, Nz, coef, vsq, 
-                                      Aeven, Aodd) >;
+                launch < stencil_step_task(x0, x1, y0, y1, z, z+dz, Nx, Ny, Nz, 
+                                           coef, vsq, Aeven, Aodd) >;
             else
-                launch < stencil_step(x0, x1, y0, y1, z, z+dz, Nx, Ny, Nz, coef, vsq, 
-                                      Aodd, Aeven) >;
+                launch < stencil_step_task(x0, x1, y0, y1, z, z+dz, Nx, Ny, Nz, 
+                                           coef, vsq, Aodd, Aeven) >;
         }
         // We need to wait for all of the launched tasks to finish before
         // starting the next iteration.
         sync;
     }
 }
+
+
+export void
+loop_stencil_ispc(uniform int t0, uniform int t1, 
+                  uniform int x0, uniform int x1,
+                  uniform int y0, uniform int y1,
+                  uniform int z0, uniform int z1,
+                  uniform int Nx, uniform int Ny, uniform int Nz,
+                  uniform const float coef[4], 
+                  uniform const float vsq[],
+                  uniform float Aeven[], uniform float Aodd[])
+{
+    for (uniform int t = t0; t < t1; ++t) {
+        if ((t & 1) == 0)
+            stencil_step(x0, x1, y0, y1, z0, z1, Nx, Ny, Nz, coef, vsq, 
+                         Aeven, Aodd);
+        else
+            stencil_step(x0, x1, y0, y1, z0, z1, Nx, Ny, Nz, coef, vsq, 
+                         Aodd, Aeven);
+    }
+}