What is the xor %eax, %eax idiom?

It zeroes %eax (anything XORed with itself is 0) — smaller and faster than mov $0, %eax, and it also clears the upper half of %rax.

xor %eax, %eax

Any value XORed with itself is zero. The CPU recognizes this special case and:

• Executes it with zero latency (register renaming)
• Clears dependency chains
• Is only 2 bytes (vs 5 bytes for mov $0, %eax)

Bonus: On x86_64, writing to %eax automatically zeros the upper 32 bits of %rax. So xor %eax, %eax also zeros %rax.

Where you'll see it: Function prologues, before a loop accumulator, before call to set %eax = 0 (indicates no floating-point varargs to printf/sscanf).

Go deeper:

• Intel XOR reference — DEST := DEST XOR SRC, so a register XORed with itself becomes zero
• Wikipedia: Bitwise operation — why x XOR x = 0