From 8b685b21282847358c28a0d51b0be7a89b48681c Mon Sep 17 00:00:00 2001
From: teridax <sven.vogel123@web.de>
Date: Sun, 30 Apr 2023 17:34:23 +0200
Subject: [PATCH] fixed memory benchmarking

---
 sparse_vector/src/main.rs | 54 ++++++++++++++++++++++++++-------------
 1 file changed, 36 insertions(+), 18 deletions(-)

diff --git a/sparse_vector/src/main.rs b/sparse_vector/src/main.rs
index 0e214ff..948087c 100644
--- a/sparse_vector/src/main.rs
+++ b/sparse_vector/src/main.rs
@@ -32,15 +32,9 @@ impl SparseVec {
     pub fn new(elements: usize, non_null: f64) -> Self {
         let non_zero_elements = (elements as f64 * non_null) as usize;
 
-        let heap_element_size = std::mem::size_of::<f64>() + std::mem::size_of::<usize>();
-
-        println!("Estimated size on heap: {}", ByteSize::b((non_zero_elements * heap_element_size) as u64));
-
-        println!("allocating...");
         let mut values = Vec::with_capacity(non_zero_elements);
         let mut indices = Vec::with_capacity(non_zero_elements);
 
-        println!("generating some data...");
         let mut rng = rand::thread_rng();
 
         for i in 0..non_zero_elements {
@@ -57,8 +51,8 @@ impl SparseVec {
     }
 }
 
+#[inline]
 fn binary_search(target: usize, indices: &[usize], values: &[f64]) -> f64 {
-
     let mut range = 0..indices.len();
     loop {
         let mut median = (range.end - range.start) >> 1;
@@ -69,7 +63,9 @@ fn binary_search(target: usize, indices: &[usize], values: &[f64]) -> f64 {
 
         if indices[median] == target {
             return values[median];
-        } else if indices[median] > target {
+        }
+
+        if indices[median] > target {
             range.end = median;
         } else {
             range.start = median;
@@ -79,20 +75,42 @@ fn binary_search(target: usize, indices: &[usize], values: &[f64]) -> f64 {
     0.0
 }
 
-fn main() {
-    let now = Instant::now();
-    // generate a sparse vector with 10^10 random elements
-    // but only with 2% of them being non-null
-    let vec = SparseVec::new(10_usize.pow(10), 0.02);
-    println!("Created sparse vector took: {}s", Instant::now().sub(now).as_secs_f32());
+macro_rules! time {
+    ($name:literal, $block:expr) => {{
+        let start = Instant::now();
+        $block;
+        println!("{} took {}s", $name, start.elapsed().as_secs_f64());
+    }}
+}
 
-    println!("Sparse vector stack bytes: {} B", std::mem::size_of_val(&vec));
+fn main() {
+
+    /// Theoretical size of the vector in elements
+    /// This would mean the we would require 10 GBs of memory to store a single vector
+    const VECTOR_SIZE: usize = 10_000_000_000;
+    /// Ratio between null and non null elements per vector
+    /// this means that only `NULL_NON_NULL_RATIO * 100%` elements of every vector will
+    /// be non-null and thus actually stored
+    const NULL_NON_NULL_RATIO: f64 = 0.02;
+
+    let non_zero_elements = (VECTOR_SIZE as f64 * NULL_NON_NULL_RATIO) as usize;
+    let heap_element_size = std::mem::size_of::<f64>() + std::mem::size_of::<usize>();
+
+    println!("Estimated size on heap: {}", ByteSize::b((non_zero_elements * heap_element_size) as u64));
+    println!("Size on stack: {} B", std::mem::size_of::<SparseVec>());
+
+    let mut vec: SparseVec;
+
+    time!("Sparse vector creation", {
+        // generate a vector
+        vec = SparseVec::new(VECTOR_SIZE, NULL_NON_NULL_RATIO);
+    });
 
     // many statistics are cached and only updated when the epoch is advanced.
     epoch::advance().unwrap();
     println!("Heap allocated bytes (total): {}", ByteSize::b(stats::allocated::read().unwrap() as u64));
 
-    let now = Instant::now();
-    vec.dot(&vec);
-    println!("Dot product took: {}s", Instant::now().sub(now).as_secs_f32());
+    time!("Sparse vector dot product", {
+        vec.dot(&vec);
+    });
 }
\ No newline at end of file