fix erroneous correction

csrc/kernels.cu

@@ -2259,7 +2259,7 @@ template<typename T, int THREADS, int ITEMS_PER_THREAD, int TILE_ROWS, int TILE_
   }
 
   // 4. store data via atomicMax
-  // to store col data efficiency we need to rewrite the smem blocked data [0, 1, 2, 3...] for t0
+  // to store col data efficiently we need to rewrite the smem blocked data [0, 1, 2, 3...] for t0
   // into a striped arrangement: [0, 8, 16, 24, ..] for t0
   __syncthreads();
   BlockExchange(temp_storage.exchange).BlockedToStriped(local_col_absmax_values);