RightSignal

Operating systems · 1 of 91 / 3

Processes and threads

The concept

How your laptop runs Spotify, Chrome and Zoom at once

A process is a running program. The operating system (OS) gives it its own private memory, a process ID number and its own list of open files. One process can't read another's memory, so Spotify can't see or overwrite what Zoom is doing.

A thread is one line of work inside a process. Each has its own stack (scratch space for function calls), registers and place in the code, but all threads in a process share its memory and files. Every process starts with one.

On a machine with 8 cores, up to 8 threads can run at the same moment. If there are more threads than cores, the OS swaps them on and off the cores many times a second. That's fast enough that everything looks like it's running at once.

Operating systems · 1 of 92 / 3

Processes and threads

Real-life use case

One Chrome tab crashes with "Aw, Snap!" and the others keep working

Chrome runs separate websites in separate processes. If one page's process crashes, you see the "Aw, Snap!" page in that tab only. The other tabs live in other processes with their own memory, so they keep running.

Inside each process, threads split the work. In a page's renderer process, the main thread runs JavaScript while a separate compositor thread helps keep scrolling smooth.

Open Chrome's own Task Manager (Shift+Esc on Windows) to see the list: tabs, extensions and a GPU process, each one a separate process.

Operating systems · 1 of 93 / 3

Processes and threads

How it helps

Threads share one memory, so splitting the work costs almost nothing

Threads share their process's memory, so splitting work is cheap. Four threads resizing one photo can each take a quarter of the pixels in place, copying nothing between them.

Creating or switching threads is also cheaper than doing it with processes, because the OS doesn't have to set up or swap a whole private memory space.

The catch: only the part that can be split speeds up. Steps that must run in order still run at one core's speed and cap the gain. And one buggy thread can corrupt shared data or crash the whole process.

What it’s forEstimates how much faster a job gets when you split it across threads on several cores (Amdahl's law).

speedup = 1 / ((1 − p) + p / n)

p: the fraction of the work that can run in parallel, e.g. 0.9
1 − p: the part that must run in order, on one core
n: the number of cores running threads, e.g. 4

What it givesHow many times faster the job finishes compared with running it on one thread.

p = 0.9, n = 4: 1 / (0.1 + 0.9/4) = 1 / (0.1 + 0.225)
= 1 / 0.325 ≈ 3.1× faster, not 4×
Even with endless cores: 1 / 0.1 = 10× at most

Your RightSignal

Topics in News

How your laptop runs Spotify, Chrome and Zoom at once

One Chrome tab crashes with "Aw, Snap!" and the others keep working

Threads share one memory, so splitting the work costs almost nothing

Got it?