No one should use the AT&T syntax (2021)

Original Article ↗ Hacker News Discussion ↗

Overall context

  • Discussion centers on x86 assembly syntax styles: AT&T (GNU/GAS) vs Intel (MASM/NASM), assuming the reader already knows the article’s critique of AT&T.
  • Several commenters see this as a long-running “syntax flame war” that resurfaces periodically.

Preferences: Intel vs AT&T

  • Many professional assembly writers reportedly prefer Intel syntax; frequent readers of compiler output often like AT&T.
  • Some participants strongly dislike AT&T and avoid it; others say they actively prefer it or can live with both.
  • A few find the article’s complaints exaggerated or “childish,” arguing either syntax is mostly a matter of taste.

Operand order & semantics

  • Major contention: operand order (src,dst vs dst,src).
  • Intel fans value consistency with vendor manuals, common CPUs, and mathematical notation, especially for sub and cmp where AT&T’s order inverts the intuitive expression.
  • Some defend destination-first as emphasizing “what changes,” but others say any consistent convention works.
  • Condition codes and comparisons in AT&T are widely seen as confusing, especially when mapping to English like “jump if greater.”

Addressing syntax & type specification

  • Intel addressing ([base + index*scale + disp]) is widely considered clearer and more expressive.
  • AT&T’s disp(base,index,scale) is viewed by many as awkward and order-sensitive, though some praise it for exposing the underlying base/index/scale constraints and producing clearer error messages.
  • Suffixes (b/w/l/q) and register prefixes (%) divide opinion: helpful explicitness vs visual noise.

Learning, tooling, and other ISAs

  • Several note AT&T was never designed for humans first; it evolved around GCC/GAS needs.
  • Some learners found AT&T easier because it forces understanding of hardware addressing rules; others say Intel’s more “math-like” expressions are easier once you know the constraints.
  • Outside x86, participants report that Intel-like operand order has “won” on modern architectures (ARM, RISC-V, many DSPs), while historically PDP-11, m68k, etc. influenced AT&T-style destination-last formats.

Other discussion threads

  • Jokes about confusing LLMs with operand order, modem “AT&Tx” commands, ThinkPad over-tuning, and classic “neko”/“xneko” mouse-chasing cat animations on the article page.
  • Some find the animated cat charming; others find it distracting and discuss ways to disable it (adblock filter, reduced-motion settings).