How does Atomic differ from a regular variable assignment?

In Go, a simple assignment to a variable does not guarantee atomicity of operations, especially in a multithreaded environment. This means that if multiple goroutines read and write to the same variable simultaneously, a race condition may occur.

Atomic operations from the sync/atomic package provide atomic read, write, and modification of values without using mutexes. This means the operation is performed entirely and indivisibly, preventing race conditions.

Example of an atomic increment:

import (
    "fmt"
    "sync/atomic"
)

var counter int64

func increment() {
    atomic.AddInt64(&counter, 1)
}

func main() {
    increment()
    fmt.Println(atomic.LoadInt64(&counter)) // 1
}

Thus, atomic operations are more efficient than mutexes for simple variable operations, but they are not suitable for complex critical sections.