Why do many developers create their own implementations of the standard library?

Main reasons for creating custom implementations of the standard library:

• Performance: Some standard implementations may be suboptimal for specific tasks or hardware platforms. A custom implementation can be tailored to achieve maximum performance.
• Code size: Embedded systems or resource-constrained projects may require minimizing the size of the final executable. The standard library may contain unnecessary functionality, and static linking increases the size.
• Portability: Cross-platform development or working on new/non-standard architectures may require an implementation adapted to the specific environment.
• Control over behavior: Developers may need precise control over the behavior of certain functions, for example, for error handling, memory management, or security.
• Resource management: Embedded systems often have strict limitations on memory usage and other resources. A custom implementation can be designed considering these constraints.
• Lack of standard implementation: In some specialized or outdated environments, the standard library may be unavailable or incomplete.
• Educational purposes / Research: Creating a custom implementation helps to better understand the principles of how the standard library works.

Example: A custom memcpy implementation optimized for a specific architecture using SIMD instructions.

// Optimized memcpy for x86-64 using AVX
void* custom_memcpy(void* restrict dest, const void* restrict src, size_t n) {
    char* d = dest;
    const char* s = src;

    // Alignment and initial byte handling
    while (n-- > 0) {
        *d++ = *s++;
    }

    // Further implementation could include SIMD for larger blocks
    // ... AVX/SSE instructions ...

    return dest;
}