add timers to dyanmic apps

python/apps/in-mem-dynamic.py

@@ -7,7 +7,6 @@
 import numpy as np
 import utils
 
-
 def insert_and_search(
     dtype_str,
     indexdata_file,
@@ -44,25 +43,26 @@ def insert_and_search(
         raise ValueError("data_type must be float, int8 or uint8")
 
     tags = np.zeros(npts, dtype=int)
+    timer = utils.timer()
     for i in range(npts):
         tags[i] = i + 1
     index.batch_insert(data, tags, num_insert_threads)
-    print("batch_insert complete")
+    print('batch_insert complete in', timer.elapsed(), 's')
 
     delete_tags = np.random.choice(
         range(1, npts + 1, 1), size=int(0.5 * npts), replace=False
     )
     for tag in delete_tags:
         index.mark_deleted(tag)
-    print("mark deletion completed")
+    print('mark deletion completed in', timer.elapsed(), 's')
 
     index.consolidate_delete()
-    print("consolidation completed")
+    print('consolidation completed in', timer.elapsed(), 's')
 
     deleted_data = data[delete_tags - 1, :]
 
     index.batch_insert(deleted_data, delete_tags, num_insert_threads)
-    print("re-insertion completed")
+    print('re-insertion completed in', timer.elapsed(), 's')
 
     tags, dists = index.batch_search(queries, K, Ls, num_search_threads)
     res_ids = tags - 1

python/apps/insert-in-clustered-order.py

@@ -49,11 +49,13 @@ def insert_and_search(
     )
 
     i = 0
+    timer = utils.timer()
     for c in range(num_clusters):
         cluster_index_range = range(offsets[c], offsets[c + 1])
         cluster_indices = permutation[cluster_index_range]
         cluster_data = data[cluster_indices, :]
         index.insert(cluster_data, cluster_indices + 1, num_insert_threads)
+        print('Inserted cluster', c, ' in ', timer.elapsed(), 's')
     tags, dists = index.batch_search(queries, K, Ls, num_search_threads)
     res_ids = tags - 1
 
@@ -71,13 +73,13 @@ def insert_and_search(
     parser.add_argument("-d", "--data_type", required=True)
     parser.add_argument("-i", "--indexdata_file", required=True)
     parser.add_argument("-q", "--querydata_file", required=True)
-    parser.add_argument("-Lb", "--Lbuild", default=50)
-    parser.add_argument("-Ls", "--Lsearch", default=50)
-    parser.add_argument("-R", "--graph_degree", default=32)
-    parser.add_argument("-TI", "--num_insert_threads", default=8)
-    parser.add_argument("-TS", "--num_search_threads", default=8)
-    parser.add_argument("-C", "--num_clusters", default=32)
-    parser.add_argument("-K", default=10)
+    parser.add_argument("-Lb", "--Lbuild", default=50, type=int)
+    parser.add_argument("-Ls", "--Lsearch", default=50, type=int)
+    parser.add_argument("-R", "--graph_degree", default=32, type=int)
+    parser.add_argument("-TI", "--num_insert_threads", default=8, type=int)
+    parser.add_argument("-TS", "--num_search_threads", default=8, type=int)
+    parser.add_argument("-C", "--num_clusters", default=32, type=int)
+    parser.add_argument("-K", default=10, type=int)
     parser.add_argument("--gt_file", default="")
     args = parser.parse_args()
 

python/apps/utils.py

@@ -4,6 +4,7 @@
 import numpy as np
 from scipy.cluster.vq import vq, kmeans2
 from typing import Tuple
+from time import perf_counter
 
 
 def get_bin_metadata(bin_file) -> Tuple[int, int]:
@@ -15,6 +16,15 @@ def bin_to_numpy(dtype, bin_file) -> np.ndarray:
     npts, ndims = get_bin_metadata(bin_file)
     return np.fromfile(file=bin_file, dtype=dtype, offset=8).reshape(npts, ndims)
 
+class timer:
+    last = perf_counter()
+
+    def elapsed(self):
+        new = perf_counter()
+        elapsed_time = new - self.last
+        self.last = new
+        return elapsed_time
+
 
 def numpy_to_bin(array, out_file):
     shape = np.shape(array)

python/src/diskann_bindings.cpp

@@ -358,8 +358,8 @@ void build_in_memory_index(const diskann::Metric &metric, const std::string &vec
                                                            .build();
     size_t data_num, data_dim;
     diskann::get_bin_metadata(vector_bin_path, data_num, data_dim);
-    diskann::Index<T, TagT, LabelT> index(metric, data_dim, data_num, false, use_tags, false, use_pq_build, num_pq_bytes,
-                                          use_opq);
+    diskann::Index<T, TagT, LabelT> index(metric, data_dim, data_num, false, use_tags, false, use_pq_build,
+                                          num_pq_bytes, use_opq);
     if (label_file == "")
     {
         index.build(vector_bin_path.c_str(), data_num, index_build_params);