diff --git a/0215_kth_largest_element_in_an_array/kth_elem.c b/0215_kth_largest_element_in_an_array/kth_elem.c
index bf9ea50..b14d278 100644
--- a/0215_kth_largest_element_in_an_array/kth_elem.c
+++ b/0215_kth_largest_element_in_an_array/kth_elem.c
@@ -2,6 +2,15 @@
 #include <stdlib.h>
 
 
+static void show(int *nums, int lo, int hi)
+{
+    int i;
+    for (i = lo; i <= hi; i++) {
+        printf("%d ", nums[i]);
+    }
+    printf("\n");
+}
+
 static inline void swap(int *a, int *b) 
 {
     int t = *a;
@@ -38,41 +47,43 @@ static void build_max_heap(int *nums, int size)
     }
 }
 
-static int quick_select(int *nums, int lo, int hi, int k)
+static void quick_select(int *nums, int lo, int hi, int k)
 {
     if (lo >= hi) {
-        return hi;
+        return;
     }
 
     int i = lo - 1;
-    int j = hi + 1;
-    int pivot = nums[lo];
+    int j = hi;
+    int pivot = nums[hi];
+
     while (i < j) {
         /* For case of large amounts of consecutive duplicate elements, we
          * shall make the partition in the middle of the array as far as
          * possible. If the partition is located in the head or tail, the
          * performance might well be very bad for it.
          */
-        while (nums[++i] > pivot) {}
-        while (nums[--j] < pivot) {}
+        while (i < hi && nums[++i] > pivot) {}
+        while (j > lo && nums[--j] < pivot) {}
         if (i < j) {
             swap(&nums[i], &nums[j]);
         }
     }
 
     /* invariant: i == j + 1 or i == j */
-    if (j >= k - 1) {
-        return quick_select(nums, lo, j, k);
+    swap(&nums[i], &nums[hi]);
+    if (i + 1 >= k) {
+        quick_select(nums, lo, i - 1, k);
     } else {
-        return quick_select(nums, j + 1, hi, k);
+        quick_select(nums, i + 1, hi, k);
     }
 }
 
 int findKthLargest(int* nums, int numsSize, int k)
 {
-#if 0
-    int i = quick_select(nums, 0, numsSize - 1, k);
-    return nums[i];
+#if 1
+    quick_select(nums, 0, numsSize - 1, k);
+    return nums[k - 1];
 #else
     int i;