From Stack to Span<T> — A Recap into C#’s Memory

Article

Introduction

When I started diving into performance in C#, I realised there’s a hidden layer behind every line of code, particularly in terms of memory usage. The way we store, access, and pass data around can mean the difference between a smooth app and a laggy mess.

So, let’s go on a recap starting from the basics and building up to some of the coolest features in modern C#, like Span<T>.

The Foundation — Stack and Heap

Before we can talk about Span<T> or even “contiguous memory,” we need to know where our data lives.

Stack

Fast, short-lived storage
Holds local variables, method parameters, and most value types
Works like a stack of plates (Last In, First Out)
Gets cleared automatically when a method ends

Heap

Slower but flexible storage
Holds reference types (objects, arrays, strings)
Managed by the garbage collector (GC)
Data here can live far longer than the method that created it

Think of the stack as a desk in front of you — quick access but limited space. The heap is a big warehouse, but you need a forklift (GC) to manage it.

struct — The Lightweight Data Container

Now that we know about the stack, let’s meet its close friend: struct.

A struct is,

A value type
Usually stored on the stack
Passed by copy, not by reference
Great for small, lightweight data

public struct Point(int x, int y)
{
    public int X = x, Y = y;
}

Point p1 = new(3, 4);
Point p2 = p1; // Copy — p2 is independent
p2.X = 10;
Console.WriteLine($"p1: ({p1.X}, {p1.Y})");
Console.WriteLine($"p2: ({p2.X}, {p2.Y})");
///////////////////////////////////////////////////////
Output:
p1: (3, 4)
p2: (10, 4)

This clearly shows that p1 stays unchanged after modifying p2, proving that structs are copied by value.

Contiguous Memory — The Secret to Speed

Some data structures store items in contiguous memory, meaning all elements sit right next to each other in one block.

[ 10 ][ 20 ][ 30 ][ 40 ]

Why is this good?

The CPU can prefetch data efficiently
Slicing (taking part of the data) is faster
Fewer memory lookups

Why List<T> Is Contiguous… But Not Quite?

At first glance, List<T> seems contiguous; it wraps an internal array.

But there’s a catch.

The List object itself lives on the heap
It holds a reference to an internal array (which is contiguous)
When the List grows, it allocates a new array and copies elements over

So, List<T> can’t directly be treated as a block of memory the way arrays or Span<T> can.

Span<T> — Memory Manipulation Without the Cost

Span<T> lets you work with slices of contiguous memory without creating new arrays or copying data.

Traditionally, in C#,

If you wanted part of an array, you’d copy it to a new array.
That meant extra allocations → more work for the GC.

Span<T> changes this.

It represents a contiguous region of memory (stack, heap, or unmanaged memory).

You can create a slice without copying.

int[] numbers = { 10, 20, 30, 40, 50 };
Span<int> part = numbers.AsSpan(1, 3); // points to 20, 30, 40

Changing part[0] also changes numbers[1] — they’re literally pointing to the same memory.
No new array, no GC hit, no wasted CPU cycles.

Why Span<T> Is a ref struct?

Unlike normal structs, Span<T> is a ref struct, which means.

It must live on the stack
Can’t be stored in the heap
Can’t be used in async methods or lambdas

Why?

Because Span<T> points directly into memory, letting it move to the heap would risk dangling pointers and unsafe memory access.

Final Takeaways

By understanding,

Stack vs Heap: where data lives
Structs: lightweight value containers
Contiguous Memory: Why Sequential Storage Is Fast
List<T> limitations: when “contiguous” isn’t quite
Span<T>: how to slice memory without allocations

Thank You, and Stay Tuned for More!