What is the AT&T assembly syntax and how does it differ from Intel syntax?

Question

Accepted Answer

AT&T (GCC/GAS default) writes op source, destination with % on registers and $ on immediates; Intel writes op destination, source with bare operands — the operand order is reversed.
The same machine instruction can be printed two ways, and mixing them up silently inverts your reading. The differences are mechanical but total:

Feature
AT&T
Intel

Operand order
src, dest
dest, src

Register prefix
%rax
rax

Immediate prefix
$42
42

Memory reference
(%rax)
[rax]

Size suffix
movl, movq
mov dword, mov qword

Displacement
8(%rbp)
[rbp+8]

# AT&T:  movl $42, %eax      # set eax to 42
# Intel: mov  eax, 42

You can make GCC emit Intel syntax with gcc -S -masm=intel program.c.
Tip: AT&T reads left-to-right as "move from A to B"; Intel reads as "set A to B." Decide which one you're looking at before trusting the direction of any instruction.
Go deeper:

x86 assembly language (Wikipedia) — a direct AT&T-vs-Intel comparison of operand order and sigils.

Feature	AT&T	Intel
Operand order	src, dest	dest, src
Register prefix	`%rax`	`rax`
Immediate prefix	`$42`	`42`
Memory reference	`(%rax)`	`[rax]`
Size suffix	`movl`, `movq`	`mov dword`, `mov qword`
Displacement	`8(%rbp)`	`[rbp+8]`