From 4369c76fdf7bbc9ccb7c534141c50beafccc9231 Mon Sep 17 00:00:00 2001
From: Suryansh Gupta <suryangupta@microsoft.com>
Date: Thu, 30 Jan 2025 17:10:44 +0530
Subject: [PATCH] Add single threaded benchmarking script

---
 apps/CMakeLists.txt                      |  3 +
 apps/benchmark_reads.cpp                 |  6 +-
 apps/benchmark_reads_single_threaded.cpp | 73 ++++++++++++++++++++++++
 3 files changed, 79 insertions(+), 3 deletions(-)
 create mode 100644 apps/benchmark_reads_single_threaded.cpp

diff --git a/apps/CMakeLists.txt b/apps/CMakeLists.txt
index 5d2c8faaa..736460246 100644
--- a/apps/CMakeLists.txt
+++ b/apps/CMakeLists.txt
@@ -31,6 +31,9 @@ target_link_libraries(test_insert_deletes_consolidate ${PROJECT_NAME} ${DISKANN_
 add_executable(benchmark_reads benchmark_reads.cpp)
 target_link_libraries(benchmark_reads ${PROJECT_NAME} ${DISKANN_TOOLS_TCMALLOC_LINK_OPTIONS} Boost::program_options)
 
+add_executable(benchmark_reads_single_threaded benchmark_reads_single_threaded.cpp)
+target_link_libraries(benchmark_reads_single_threaded ${PROJECT_NAME} ${DISKANN_TOOLS_TCMALLOC_LINK_OPTIONS} Boost::program_options)
+
 if (NOT MSVC)
     install(TARGETS build_memory_index
             build_stitched_index
diff --git a/apps/benchmark_reads.cpp b/apps/benchmark_reads.cpp
index 79c755dd3..1b05c2b47 100644
--- a/apps/benchmark_reads.cpp
+++ b/apps/benchmark_reads.cpp
@@ -43,9 +43,9 @@ void do_multiple_reads_with_threads(int thread_count)
 
     vector<char *> buffers(thread_count);
 
-    // omp_set_num_threads(thread_count);
+    omp_set_num_threads(thread_count);
 
-// #pragma omp parallel for num_threads((int)thread_count)
+#pragma omp parallel for num_threads((int)thread_count)
     for (int i = 0; i < thread_count; i++)
     {
         char *buf = nullptr;
@@ -56,7 +56,7 @@ void do_multiple_reads_with_threads(int thread_count)
 
     int no_of_reads = total_reads / batches_of;
     Timer timer;
-// #pragma omp parallel for schedule(dynamic, 1)
+#pragma omp parallel for schedule(dynamic, 1)
     for (int i = 0; i < no_of_reads; i++)
     {
         char *buf = buffers[omp_get_thread_num()];
diff --git a/apps/benchmark_reads_single_threaded.cpp b/apps/benchmark_reads_single_threaded.cpp
new file mode 100644
index 000000000..9c1b9e70f
--- /dev/null
+++ b/apps/benchmark_reads_single_threaded.cpp
@@ -0,0 +1,73 @@
+#include <iostream>
+#include <cstdlib>
+#include "windows_aligned_file_reader.h"
+#include "aligned_file_reader.h"
+#include "utils.h"
+#include "timer.h"
+#include <omp.h>
+
+using namespace std;
+using namespace diskann;
+
+#define SECTOR_LEN 4096
+#define TOTAL_READS 1000000
+
+void do_reads(WindowsAlignedFileReader *reader, char *buf, int batches_of)
+{
+    auto ctx = reader->get_ctx();
+
+    std::vector<AlignedRead> read_reqs;
+    read_reqs.reserve(batches_of);
+
+    // create read requests
+    for (size_t i = 0; i < batches_of; ++i)
+    {
+        AlignedRead read;
+        read.len = SECTOR_LEN;
+        read.buf = buf + i * SECTOR_LEN;
+        auto sector_id = (rand() % 1650000);
+        read.offset = sector_id * SECTOR_LEN;
+        read_reqs.push_back(read);
+    }
+
+    reader->read(read_reqs, ctx, false);
+}
+
+void do_reads_in_batches_of(int batches_of)
+{
+    string file_name = "F:\\indices\\turing_10m\\disk_index_disk.index";
+    auto reader = new WindowsAlignedFileReader();
+    reader->open(file_name.c_str());
+    char *buf = nullptr;
+    alloc_aligned((void **)&buf, batches_of * SECTOR_LEN, SECTOR_LEN);
+    reader->register_thread();
+
+    int no_of_reads = TOTAL_READS / batches_of;
+    Timer timer;
+    for (int i = 0; i < no_of_reads; i++)
+    {
+        do_reads(reader, buf, batches_of);
+    }
+    // cout << "Time taken to read in microseconds: " << timer.elapsed() << endl;
+    cout<<timer.elapsed()<<endl;
+
+    reader->close();
+}
+
+int main(int argc, char *argv[])
+{
+    int val = 10;
+    if (argc >= 2)
+    {
+        std::istringstream iss(argv[1]);
+
+        if (iss >> val)
+        {
+            // cout << "Got cmd argument" << endl;
+        }
+    }
+    cout << "Using batches of " << val << endl;
+
+    // cout << "Hello World" << endl;
+    do_reads_in_batches_of(val);
+}
\ No newline at end of file