.NET High-Performance I/O

.NET High-Performance I/O

A guide for APIs optimizing large-scale data input/output.

Quick Reference: See QUICKREF.md for essential patterns at a glance.

1. Core APIs

API	Purpose
Console.OpenStandardInput()	Buffered stream input
Console.OpenStandardOutput()	Buffered stream output
BufferedStream	Stream buffering
FileOptions.Asynchronous	Async file I/O

---

2. High-Speed Standard I/O

2.1 Basic Pattern

// Use buffer stream directly for large I/O
using var inputStream = Console.OpenStandardInput();
using var outputStream = Console.OpenStandardOutput();
using var reader = new StreamReader(inputStream, bufferSize: 65536);
using var writer = new StreamWriter(outputStream, bufferSize: 65536);

// Disable buffer flush for performance improvement
writer.AutoFlush = false;

string? line;
while ((line = reader.ReadLine()) is not null)
{
    writer.WriteLine(ProcessLine(line));
}

// Manual flush at the end
writer.Flush();

2.2 For Algorithm Problem Solving

using System.Text;

// High-speed input
using var reader = new StreamReader(
    Console.OpenStandardInput(),
    Encoding.ASCII,
    bufferSize: 65536);

// High-speed output
using var writer = new StreamWriter(
    Console.OpenStandardOutput(),
    Encoding.ASCII,
    bufferSize: 65536);

var sb = new StringBuilder();

// Collect large output in StringBuilder and write at once
for (int i = 0; i < 100000; i++)
{
    sb.AppendLine(i.ToString());
}

writer.Write(sb);
writer.Flush();

---

3. File I/O Optimization

3.1 Buffer Size Optimization

// Use larger buffer than default (4KB)
const int bufferSize = 64 * 1024; // 64KB

using var fileStream = new FileStream(
    path,
    FileMode.Open,
    FileAccess.Read,
    FileShare.Read,
    bufferSize: bufferSize);

3.2 Async File I/O

// Open file with async option
using var fileStream = new FileStream(
    path,
    FileMode.Open,
    FileAccess.Read,
    FileShare.Read,
    bufferSize: 4096,
    options: FileOptions.Asynchronous);

var buffer = new byte[4096];
int bytesRead = await fileStream.ReadAsync(buffer);

3.3 SequentialScan Hint

// Provide hint to OS for sequential reading
using var fileStream = new FileStream(
    path,
    FileMode.Open,
    FileAccess.Read,
    FileShare.Read,
    bufferSize: 64 * 1024,
    options: FileOptions.SequentialScan);

3.4 RandomAccess (.NET 6+)

// Direct offset access without file position management
using var handle = File.OpenHandle(path, FileMode.Open, FileAccess.Read);

var buffer = new byte[4096];
long offset = 1000;

int bytesRead = RandomAccess.Read(handle, buffer, offset);

// Async version
bytesRead = await RandomAccess.ReadAsync(handle, buffer, offset);

---

4. Large File Processing

4.1 Chunk-Based Reading

public async IAsyncEnumerable<byte[]> ReadChunksAsync(
    string path,
    int chunkSize = 64 * 1024,
    [EnumeratorCancellation] CancellationToken ct = default)
{
    using var stream = new FileStream(
        path,
        FileMode.Open,
        FileAccess.Read,
        FileShare.Read,
        bufferSize: chunkSize,
        options: FileOptions.Asynchronous | FileOptions.SequentialScan);

    var buffer = new byte[chunkSize];
    int bytesRead;

    while ((bytesRead = await stream.ReadAsync(buffer, ct)) > 0)
    {
        if (bytesRead == chunkSize)
        {
            yield return buffer;
            buffer = new byte[chunkSize];
        }
        else
        {
            yield return buffer[..bytesRead];
        }
    }
}

4.2 Memory-Mapped Files

using System.IO.MemoryMappedFiles;

// Map large file to memory
using var mmf = MemoryMappedFile.CreateFromFile(path, FileMode.Open);
using var accessor = mmf.CreateViewAccessor();

// Direct memory access
byte value = accessor.ReadByte(position);
accessor.Write(position, newValue);

---

5. Performance Comparison

Method	Relative Performance	Use Case
Console.ReadLine()	1x (baseline)	General
StreamReader (default buffer)	2x	Large data
StreamReader (64KB buffer)	3-5x	Large data
MemoryMappedFile	5-10x	Very large data

---

6. Important Notes

Buffer Size

• Too small increases system calls
• Too large wastes memory
• Recommended: 4KB ~ 64KB

Encoding Specification

// Read UTF-8 without BOM
using var reader = new StreamReader(
    stream,
    new UTF8Encoding(encoderShouldEmitUTF8Identifier: false));

Flush Timing

// Improve performance with AutoFlush = false
writer.AutoFlush = false;

// Manual flush after important data
writer.Flush();

---

7. References

• File and Stream I/O
• Memory-Mapped Files