fix: type corrections and proper copying of result data from device to host

Author: madhav-madhusoodanan

samples/introduction/matmul/src/main.rs

@@ -47,8 +47,8 @@ fn matrix_multiply(block_size: usize, dimsA: (usize, usize, usize), dimsB: (usiz
             d_c.as_device_ptr(),
             d_a.as_device_ptr(),
             d_b.as_device_ptr(),
-            dimsA.0 as u32,
-            dimsB.0 as u32
+            dimsA.0 as usize,
+            dimsB.0 as usize
         ))?;
     }
 
@@ -67,8 +67,8 @@ fn matrix_multiply(block_size: usize, dimsA: (usize, usize, usize), dimsB: (usiz
                 d_c.as_device_ptr(),
                 d_a.as_device_ptr(),
                 d_b.as_device_ptr(),
-                dimsA.0 as u32,
-                dimsB.0 as u32,
+                dimsA.0 as usize,
+                dimsB.0 as usize,
             ))?;
         }
     }
@@ -90,6 +90,12 @@ fn matrix_multiply(block_size: usize, dimsA: (usize, usize, usize), dimsB: (usiz
     let gigaFlops = (flopsPerMatrixMul / (avg_time)) / 1000.0;
     println!("Performance = {} GFlop/s", gigaFlops);
 
+    unsafe{ 
+        d_c
+            .async_copy_to(&mut h_c, &stream)
+            .expect("Could not copy from device to host!");
+    }
+
     // checking computed result
     // test relative error by the formula
     // |<x, y>_cpu - <x, y>_gpu| / |<x, y>_cpu|