Merge pull request #40 from mahmudhera/main

added sketch sizes in compare output (resolves #39)

Author: mahmudhera

README.md

@@ -18,14 +18,16 @@ All tool usages are available using `--help` flag.
 
 ## `compare` output format
 1. Index of query
-2. Name of query
-3. MD5 of query
-4. Index of match
-5. Name of match
-6. MD5 of match
-7. Jaccard
-8. Containment(query, target)
-9. Containment(target, query)
+1. Name of query
+1. MD5 of query
+1. Size of the query sketch
+1. Index of match
+1. Name of match
+1. MD5 of match
+1. Size of the matched reference sketch
+1. Jaccard
+1. Containment(query, target)
+1. Containment(target, query)
 
 
 # Testing

benchmark/plot_indexing_time_memory.py

@@ -12,7 +12,7 @@
 # col names: filesize,cpu_time,wall_clock_time,peak_memory_usage
 
 # plot wall-clock time against file size
-plt.plot(df_sourmash_index['filesize'], df_sourmash_index['wall_clock_time'], label='sourmash index', marker='o')
+plt.plot(df_sourmash_index['filesize'], df_sourmash_index['wall_clock_time'], label='sourmash index (1 thread)', marker='o')
 plt.plot(df_cpp_index_one_thread['filesize'], df_cpp_index_one_thread['wall_clock_time'], label='cpp index (1 thread)', marker='o')
 plt.plot(df_cpp_index_128_threads['filesize'], df_cpp_index_128_threads['wall_clock_time'], label='cpp index (128 threads)', marker='o')
 
@@ -26,7 +26,7 @@
 
 # plot cpu time against file size
 plt.clf()
-plt.plot(df_sourmash_index['filesize'], df_sourmash_index['cpu_time'], label='sourmash index', marker='o')
+plt.plot(df_sourmash_index['filesize'], df_sourmash_index['cpu_time'], label='sourmash index (1 thread)', marker='o')
 plt.plot(df_cpp_index_one_thread['filesize'], df_cpp_index_one_thread['cpu_time'], label='cpp index (1 thread)', marker='o')
 plt.plot(df_cpp_index_128_threads['filesize'], df_cpp_index_128_threads['cpu_time'], label='cpp index (128 threads)', marker='o')
 
@@ -40,7 +40,7 @@
 
 # plot peak memory usage against file size
 plt.clf()
-plt.plot(df_sourmash_index['filesize'], df_sourmash_index['peak_memory_usage'], label='sourmash index', marker='o')
+plt.plot(df_sourmash_index['filesize'], df_sourmash_index['peak_memory_usage'], label='sourmash index (1 thread)', marker='o')
 plt.plot(df_cpp_index_one_thread['filesize'], df_cpp_index_one_thread['peak_memory_usage'], label='cpp index (1 thread)', marker='o')
 plt.plot(df_cpp_index_128_threads['filesize'], df_cpp_index_128_threads['peak_memory_usage'], label='cpp index (128 threads)', marker='o')
 

benchmark/plot_many_v_many.py

@@ -21,8 +21,8 @@
 df_compare_without_index_30000 = df_compare_without_index[df_compare_without_index['query_filesize'] == 30000]
 df_multisearch_30000 = df_multisearch[df_multisearch['query_filesize'] == 30000]
 
-plt.plot(df_compare_30000['ref_filesize'], df_compare_30000['wall_clock_time'], label='compare (cpp)', marker='o')
-plt.plot(df_compare_without_index_30000['ref_filesize'], df_compare_without_index_30000['wall_clock_time'], label='compare (cpp) without index', marker='o')
+plt.plot(df_compare_30000['ref_filesize'], df_compare_30000['wall_clock_time'], label='compare (cpp) by building index', marker='o')
+plt.plot(df_compare_without_index_30000['ref_filesize'], df_compare_without_index_30000['wall_clock_time'], label='compare (cpp) by loading index', marker='o')
 plt.plot(df_multisearch_30000['ref_filesize'], df_multisearch_30000['wall_clock_time'], label='multisearch (sourmash)', marker='o')
 
 plt.xlabel('Reference list size')
@@ -36,8 +36,8 @@
 
 # plot CPU time for query file size = 30000, use different reference file sizes in x-axis
 plt.clf()
-plt.plot(df_compare_30000['ref_filesize'], df_compare_30000['cpu_time'], label='compare (cpp)', marker='o')
-plt.plot(df_compare_without_index_30000['ref_filesize'], df_compare_without_index_30000['cpu_time'], label='compare (cpp) without index', marker='o')
+plt.plot(df_compare_30000['ref_filesize'], df_compare_30000['cpu_time'], label='compare (cpp) by building index', marker='o')
+plt.plot(df_compare_without_index_30000['ref_filesize'], df_compare_without_index_30000['cpu_time'], label='compare (cpp) by loading index', marker='o')
 plt.plot(df_multisearch_30000['ref_filesize'], df_multisearch_30000['cpu_time'], label='multisearch (sourmash)', marker='o')
 
 plt.xlabel('Reference list size')
@@ -51,8 +51,8 @@
 
 # plot peak memory usage for query file size = 30000, use different reference file sizes in x-axis
 plt.clf()
-plt.plot(df_compare_30000['ref_filesize'], df_compare_30000['peak_memory_usage'], label='compare (cpp)', marker='o')
-plt.plot(df_compare_without_index_30000['ref_filesize'], df_compare_without_index_30000['peak_memory_usage'], label='compare (cpp) without index', marker='o')
+plt.plot(df_compare_30000['ref_filesize'], df_compare_30000['peak_memory_usage'], label='compare (cpp) by building index', marker='o')
+plt.plot(df_compare_without_index_30000['ref_filesize'], df_compare_without_index_30000['peak_memory_usage'], label='compare (cpp) by loading index', marker='o')
 plt.plot(df_multisearch_30000['ref_filesize'], df_multisearch_30000['peak_memory_usage'], label='multisearch (sourmash)', marker='o')
 
 plt.xlabel('Reference list size')
@@ -70,8 +70,8 @@
 df_compare_without_index_30000 = df_compare_without_index[df_compare_without_index['ref_filesize'] == 30000]
 df_multisearch_30000 = df_multisearch[df_multisearch['ref_filesize'] == 30000]
 
-plt.plot(df_compare_30000['query_filesize'], df_compare_30000['wall_clock_time'], label='compare (cpp)', marker='o')
-plt.plot(df_compare_without_index_30000['query_filesize'], df_compare_without_index_30000['wall_clock_time'], label='compare (cpp) without index', marker='o')
+plt.plot(df_compare_30000['query_filesize'], df_compare_30000['wall_clock_time'], label='compare (cpp) by building index', marker='o')
+plt.plot(df_compare_without_index_30000['query_filesize'], df_compare_without_index_30000['wall_clock_time'], label='compare (cpp) by loading index', marker='o')
 plt.plot(df_multisearch_30000['query_filesize'], df_multisearch_30000['wall_clock_time'], label='multisearch (sourmash)', marker='o')
 
 plt.xlabel('Query list size')
@@ -85,8 +85,8 @@
 
 # plot peak memory usage for ref file size = 30000, use different query file sizes in x-axis
 plt.clf()
-plt.plot(df_compare_30000['query_filesize'], df_compare_30000['peak_memory_usage'], label='compare (cpp)', marker='o')
-plt.plot(df_compare_without_index_30000['query_filesize'], df_compare_without_index_30000['peak_memory_usage'], label='compare (cpp) without index', marker='o')
+plt.plot(df_compare_30000['query_filesize'], df_compare_30000['peak_memory_usage'], label='compare (cpp) by building index', marker='o')
+plt.plot(df_compare_without_index_30000['query_filesize'], df_compare_without_index_30000['peak_memory_usage'], label='compare (cpp) by loading index', marker='o')
 plt.plot(df_multisearch_30000['query_filesize'], df_multisearch_30000['peak_memory_usage'], label='multisearch (sourmash)', marker='o')
 
 plt.xlabel('Query list size')
@@ -100,8 +100,8 @@
 
 # plot CPU time for ref file size = 30000, use different query file sizes in x-axis
 plt.clf()
-plt.plot(df_compare_30000['query_filesize'], df_compare_30000['cpu_time'], label='compare (cpp)', marker='o')
-plt.plot(df_compare_without_index_30000['query_filesize'], df_compare_without_index_30000['cpu_time'], label='compare (cpp) without index', marker='o')
+plt.plot(df_compare_30000['query_filesize'], df_compare_30000['cpu_time'], label='compare (cpp) by building index', marker='o')
+plt.plot(df_compare_without_index_30000['query_filesize'], df_compare_without_index_30000['cpu_time'], label='compare (cpp) by loading index', marker='o')
 plt.plot(df_multisearch_30000['query_filesize'], df_multisearch_30000['cpu_time'], label='multisearch (sourmash)', marker='o')
 
 plt.xlabel('Query list size')

benchmark/plot_one_v_many.py

@@ -12,9 +12,9 @@
 # col names: ref_filesize,cpu_time,wall_clock_time,peak_memory_usage
 
 # plot wall-clock time against reference file size
-plt.plot(df_compare['ref_filesize'], df_compare['wall_clock_time'], label='compare (cpp)', marker='o')
-plt.plot(df_multisearch['ref_filesize'], df_multisearch['wall_clock_time'], label='multisearch (sourmash)', marker='o')
-plt.plot(df_prefetch_lm['ref_filesize'], df_prefetch_lm['wall_clock_time'], label='prefetch-lm (cpp)', marker='o')
+plt.plot(df_compare['ref_filesize'], df_compare['wall_clock_time'], label='compare (cpp) (1 thread)', marker='o')
+plt.plot(df_multisearch['ref_filesize'], df_multisearch['wall_clock_time'], label='multisearch (sourmash) (1 thread)', marker='o')
+plt.plot(df_prefetch_lm['ref_filesize'], df_prefetch_lm['wall_clock_time'], label='prefetch-lm (cpp) (1 thread)', marker='o')
 
 plt.xlabel('Reference list size')
 plt.ylabel('Wall-clock time (s)')
@@ -26,9 +26,9 @@
 
 # plot peak memory usage against reference file size
 plt.clf()
-plt.plot(df_compare['ref_filesize'], df_compare['peak_memory_usage'], label='compare (cpp)', marker='o')
-plt.plot(df_multisearch['ref_filesize'], df_multisearch['peak_memory_usage'], label='multisearch (sourmash)', marker='o')
-plt.plot(df_prefetch_lm['ref_filesize'], df_prefetch_lm['peak_memory_usage'], label='prefetch-lm (cpp)', marker='o')
+plt.plot(df_compare['ref_filesize'], df_compare['peak_memory_usage'], label='compare (cpp) (1 thread)', marker='o')
+plt.plot(df_multisearch['ref_filesize'], df_multisearch['peak_memory_usage'], label='multisearch (sourmash) (1 thread)', marker='o')
+plt.plot(df_prefetch_lm['ref_filesize'], df_prefetch_lm['peak_memory_usage'], label='prefetch-lm (cpp) (1 thread)', marker='o')
 
 plt.xlabel('Reference list size')
 plt.ylabel('Peak memory usage (GB)')

benchmark/plot_this_vs_sourmash_prefetch.py

@@ -11,8 +11,8 @@
 # col names: ref_filesize,cpu_time,wall_clock_time,peak_memory_usage
 
 # plot wall-clock time against reference file size
-plt.plot(df_this_prefetch['ref_filesize'], df_this_prefetch['wall_clock_time'], label='cpp prefetch', marker='o')
-plt.plot(df_sourmash_prefetch['ref_filesize'], df_sourmash_prefetch['wall_clock_time'], label='sourmash prefetch', marker='o')
+plt.plot(df_this_prefetch['ref_filesize'], df_this_prefetch['wall_clock_time'], label='cpp prefetch (1 thread)', marker='o')
+plt.plot(df_sourmash_prefetch['ref_filesize'], df_sourmash_prefetch['wall_clock_time'], label='sourmash prefetch (1 thread)', marker='o')
 
 plt.xlabel('Reference list size')
 plt.ylabel('Wall-clock time (s)')
@@ -25,8 +25,8 @@
 
 # plot peak memory usage against reference file size
 plt.clf()
-plt.plot(df_this_prefetch['ref_filesize'], df_this_prefetch['peak_memory_usage'], label='cpp prefetch', marker='o')
-plt.plot(df_sourmash_prefetch['ref_filesize'], df_sourmash_prefetch['peak_memory_usage'], label='sourmash prefetch', marker='o')
+plt.plot(df_this_prefetch['ref_filesize'], df_this_prefetch['peak_memory_usage'], label='cpp prefetch (1 thread)', marker='o')
+plt.plot(df_sourmash_prefetch['ref_filesize'], df_sourmash_prefetch['peak_memory_usage'], label='sourmash prefetch (1 thread)', marker='o')
 
 plt.xlabel('Reference list size')
 plt.ylabel('Peak memory usage (GB)')

benchmark/plots/benchmark_results_compare_many_v_many_cpu_time.pdf

@@ -102,7 +102,7 @@ void do_compare(Arguments& args) {
 
     // write the header in the output file
     ofstream output_file(args.output_filename);
-    output_file << "query_id, query_name, query_md5, match_id, match_name, match_md5, jaccard, containment_query_in_match, containment_match_in_query, max_containment, max_containment_ani" << endl;
+    output_file << "query_id,query_name,query_md5,query_sketch_size,match_id,match_name,match_md5,match_sketch_size,jaccard,containment_query_in_match,containment_match_in_query,max_containment,max_containment_ani" << endl;
     output_file.close();
 
     // combining command: cat 

benchmark/plots/benchmark_results_compare_many_v_many_cpu_time_query.pdf

@@ -277,8 +277,13 @@ void compute_intersection_matrix_by_sketches(int query_sketch_start_index, int q
             double max_containment = std::max(containment_i_in_j, containment_j_in_i);
             double max_containment_ani = pow(max_containment, 1.0/ksize);
 
-            // write i, query_name, query_md5, j, ref_name, ref_md5, jaccard, containment_i_in_j, containment_j_in_i, max_containment, max_containment_ani
-            outfile << i << ",\"" << query_name << "\"," << query_md5 << "," << j << ",\"" << ref_name << "\"," << ref_md5 << "," << jaccard << "," << containment_i_in_j << "," << containment_j_in_i << "," << max_containment << "," << max_containment_ani << std::endl;
+            // write i, query_name, query_md5, quey_sketch_size, j, ref_name, ref_md5, ref_sketch_size, jaccard, containment_i_in_j, containment_j_in_i, max_containment, max_containment_ani
+            outfile << i << ",\"" << query_name << "\"," << query_md5 << "," << 
+                    sketches_query[i].size() << "," << j << ",\"" << ref_name << "\"," << 
+                    ref_md5 << "," << info_of_ref_sketches[j].size() << "," << jaccard << "," << 
+                    containment_i_in_j << "," << containment_j_in_i << "," << max_containment << "," << 
+                    max_containment_ani << std::endl;
+            //outfile << i << ",\"" << query_name << "\"," << query_md5 << "," << j << ",\"" << ref_name << "\"," << ref_md5 << "," << jaccard << "," << containment_i_in_j << "," << containment_j_in_i << "," << max_containment << "," << max_containment_ani << std::endl;
 
             similars[i].push_back(j);
         }