diff --git a/ksw_ext_avx2_aligned.c b/ksw_ext_avx2_aligned.c
index e708ea7..9951ee9 100644
--- a/ksw_ext_avx2_aligned.c
+++ b/ksw_ext_avx2_aligned.c
@@ -389,6 +389,7 @@ int ksw_extend_avx2_aligned(thread_mem_t *tmem,
             hA2[end + 1] = fA2[end + 1];
             mA2[end + 1] = 0;
         }
+
         SIMD_FIND_MAX;
 
         // 注意最后跳出循环j的值
diff --git a/ksw_ext_normal.c b/ksw_ext_normal.c
index 3492ce5..d8f3469 100644
--- a/ksw_ext_normal.c
+++ b/ksw_ext_normal.c
@@ -25,11 +25,11 @@ int ksw_extend_normal(thread_mem_t *tmem, int qlen, const uint8_t *query, int tl
 	int8_t *qp; // query profile
 	int i, j, k, oe_del = o_del + e_del, oe_ins = o_ins + e_ins, beg, end, max, max_i, max_j, max_ins, max_del, max_ie, gscore, max_off;
 	assert(h0 > 0);
-	qp = malloc(qlen * m);
-	eh = calloc(qlen + 1, 8);
-	// qp = thread_mem_request(tmem, qlen * m);
-	// eh = thread_mem_request_and_clean(tmem, (qlen + 1) * 9);
-	//  generate the query profile
+	// qp = malloc(qlen * m);
+	// eh = calloc(qlen + 1, 8);
+	qp = thread_mem_request(tmem, qlen * m);
+	eh = thread_mem_request_and_clean(tmem, (qlen + 1) * 8);
+	//   generate the query profile
 	for (k = i = 0; k < m; ++k)
 	{
 		const int8_t *p = &mat[k * m];
@@ -173,9 +173,9 @@ int ksw_extend_normal(thread_mem_t *tmem, int qlen, const uint8_t *query, int tl
 										   //   fprintf(stderr, "\n");
 										   //   fprintf(stderr, "%d\n", end);
 	}
-	free(eh);
-	free(qp);
-	// thread_mem_release(tmem, qlen * m + (qlen + 1) * 9);
+	// free(eh);
+	// free(qp);
+	thread_mem_release(tmem, qlen * m + (qlen + 1) * 8);
 	if (_qle)
 		*_qle = max_j + 1;
 	if (_tle)
diff --git a/main.c b/main.c
index f5adbcc..ccb4a69 100644
--- a/main.c
+++ b/main.c
@@ -141,7 +141,7 @@ int main(int argc, char *argv[])
     }
     // 记录计算出的分数
     int score[KERNEL_NUM] = {0};
-    int score_total[KERNEL_NUM] = {0};
+    int64_t score_total[KERNEL_NUM] = {0};
 
     // 读取测试数据
     char *query_arr = (char *)malloc(SEQ_BUF_SIZE);
@@ -242,18 +242,18 @@ int main(int argc, char *argv[])
         PERFORMANCE_TEST_NORMAL(0, ksw_extend_normal);
 
         // avx2
-        // PERFORMANCE_TEST_AVX2(1, ksw_extend_avx2);
+        PERFORMANCE_TEST_AVX2(1, ksw_extend_avx2);
         // avx2 heuristics
-        // PERFORMANCE_TEST_AVX2(2, ksw_extend_avx2_heuristics);
+        PERFORMANCE_TEST_AVX2(2, ksw_extend_avx2_heuristics);
         // avx2 mem aligned
-        // PERFORMANCE_TEST_AVX2(3, ksw_extend_avx2_aligned);
+        PERFORMANCE_TEST_AVX2(3, ksw_extend_avx2_aligned);
 
         // avx2 u8
-        // PERFORMANCE_TEST_AVX2(4, ksw_extend_avx2_u8);
+        PERFORMANCE_TEST_AVX2(4, ksw_extend_avx2_u8);
         // avx2 u8 heuristics
-        // PERFORMANCE_TEST_AVX2(5, ksw_extend_avx2_u8_heuristics);
+        PERFORMANCE_TEST_AVX2(5, ksw_extend_avx2_u8_heuristics);
         // avx2 u8 mem aligned
-        //        PERFORMANCE_TEST_AVX2(6, ksw_extend_avx2_u8_aligned);
+        PERFORMANCE_TEST_AVX2(6, ksw_extend_avx2_u8_aligned);
     }
 
 #ifdef SHOW_PERF
@@ -268,7 +268,7 @@ int main(int argc, char *argv[])
 
     for (i = 0; i < KERNEL_NUM; ++i)
     {
-        fprintf(stderr, "[%18s] time: %9.6f s; score: %d\n", kernel_names[i], time_sw[i] / DIVIDE_BY, score_total[i]);
+        fprintf(stderr, "[%18s] time: %9.6f s; score: %ld\n", kernel_names[i], time_sw[i] / DIVIDE_BY, score_total[i]);
     }
 #endif