Skip to content
Assembly and Fortran

Adversarial review: "Retro Game Dev Advisor: Assembly vs. Fortran" (HTML)

Metadata
Field Value
target ../segments/retro_game_dev_language_advisor.html
derived_from ../segments/Assembly-Versus-Fortran-Comparison.md (see review 02)
segments ../segments/retro-game-dev-language-advisor/
pass 1 (adversarial, desk review)
date 2026-09-27

This page is a summary of report 02. Thus all of review 02 applies. Review 02 covers these items. FORTRAN 77 has no array syntax or SIMD. Review 02 also covers the Mode 13h 60 fps and tearing claims. Developers wrote Elite/Frontier/M.U.L.E. in assembly. The page does not mention C. This review covers only what the page adds or changes.

1. Goals and objectives

Section titled “1. Goals and objectives”
  1. Recommend a language from the game concept of the user through a 4-question diagnostic.
  2. Give a 12-week roadmap and a 90-day deliverable per track.
  3. Show how each language meets modern player expectations, genre fit, and CPU budget.
  4. Point to a working retro toolchain.

2. Verdict (pass 1)

Section titled “2. Verdict (pass 1)”
Objective Met? Why
1. Concept diagnostic Partly It is better than the quiz on page 03. The reason is that it asks about the game, not the language. But it keeps the A=Assembly/B=Fortran bias and the one-click verdict (C1). Its “Assembly is mandatory” wording is absolute.
2. Roadmap No The Fortran track contradicts itself on dialect (C2). The Assembly deliverable promises an impossible 60 fps V-Sync lock (C3).
3. UX / genre / CPU analytics No The charts use invented numbers, and history contradicts them (C4, C5).
4. Toolchain Partly Most tools are correct, but some labels are wrong (M2, M3).

3. Findings

Section titled “3. Findings”

Critical

Section titled “Critical”
# Location Problem Tag
C1 Quiz logic L499–541 The quiz has the same defects as page 03. The code does not use answeredCount, thus one click gives a 100% verdict. Option A is always Assembly. The Assembly result says “Assembly is mandatory for your retro action/arcade title” (L530). This is false. Developers wrote plenty of retro action games in C with small amounts of assembly. [DOC] + [KNOW]
C2 L552 vs track title The heading of the Fortran track is “OpenWatcom FORTRAN 77”. But week 1 teaches “FORTRAN 77 / Modern Fortran free-form syntax”. FORTRAN 77 is fixed-form. Free-form syntax started in Fortran 90. Week 3 teaches “array non-aliasing execution” and loop-unrolling optimization. These do not matter on DOS targets without SIMD. [KNOW]
C3 L353, L577, summary card The page says “Locked 60/70 FPS”, “guaranteeing locked 60 or 70 FPS without tearing”, and “smooth 60 FPS DOS executable”. VGA Mode 13h refreshes at 70 Hz. Thus V-Sync locking gives 70 or 35 fps. On a 70 Hz display, 60 fps judders. Mode 13h has no page flipping. Thus the “guaranteed” tear-free claim is wrong (review 02, C4). [KNOW]
C4 Genre radar L630–656 The page invents the scores. Assembly gets Orbital Flight Sim 40 and Procedural Universe 45. Fortran gets 95/95. History says otherwise. Developers wrote Elite and Frontier: Elite II in assembly. These games are the landmark procedural universe and orbital flight sims. Fortran gets 20 to 30 for arcade and raycaster. But the tie-break on the same page suggests Fortran plus assembly rendering. That combination would score higher. [KNOW], [VERIFY] per-title language
C5 CPU cycle chart L676–715 The chart label is “Estimated frame budget”. The page invents these percentages and gives no measurement. The chart also confuses genre workload with language. An “Assembly action loop” and a “Fortran simulation loop” are different games. Thus the chart cannot compare the languages. [DOC]

Major

Section titled “Major”
# Location Problem Tag
M1 L125 The page says “4.77–33 MHz CPUs … direct VGA framebuffer (0xA000)”. VGA and Mode 13h on a 4.77 MHz 8088 was a rare combination. The upper bound also excludes most of the 1990s. [KNOW]
M2 L432 The table row says “NASM / WASM — Netwide Assembler…”. NASM and WASM (the Watcom assembler) are different tools. The description covers only NASM. [KNOW]
M3 L445 The page says “DOSBox-X / 86Box — Cycle-exact”. This is not true for either tool in the strict sense. DOSBox-X is clearly approximate. [KNOW]
M4 UX cards The page shows “Fortran: OS/Library Dependent” for V-Sync as a weakness. But polling port 0x3DA needs only one I/O routine. This routine is trivial to link. The page overstates how hard hybrid development is. But its own tie-break recommends a hybrid. [DOC]
M5 whole page The page has no citations. It also does not mention C, the historically dominant choice (review 02, §5). [DOC]

Minor / technical

Section titled “Minor / technical”

The problems with the Tailwind Play CDN and the unpinned Chart.js also apply here. The same is true for the implicit-event usage (L493). See review 03, m1 to m3.

4. Suggested fixes

Section titled “4. Suggested fixes”
  1. Fix the Fortran dialect story. Use FORTRAN 77, fixed-form, with no array syntax. Or use Modern Fortran on a modern host. Do not use both.
  2. Correct the frame-rate language: “70 Hz Mode 13h; tear-free needs Mode X page flipping”.
  3. Replace the invented charts with a sourced table of real retro games and their implementation languages.
  4. Add a third quiz outcome, “C with assembly inner loops”. This is the historically correct answer for many concepts.