Mode d’emploi
- Copiez la commande et examinez le paquet avant de l’installer dans votre environnement agent.
- Consultez le document original pour les outils, identifiants et environnements requis.
- Décrivez les entrées et le livrable souhaité ; comparez la sortie aux sources.
npx skills add https://github.com/ewrogers/sgdk-skills --skill sgdk-mathAvant de commencer
- L’installation n’inclut ni abonnement au modèle, ni quota API, ni accès aux fichiers.
- Étoiles et installations reflètent l’activité, pas un test indépendant de qualité ou de sécurité.
Rendu depuis le dépôt source en conservant titres, exemples, code, tableaux, liens et images.
SGDK Math
Choose the coordinate range, smallest useful fraction and largest intermediate before choosing a type. A wider storage type does not automatically mean a more accurate or safer multiply/divide implementation.
Choose the right reference
- Numbers and arithmetic: SGDK's actual formats,
conversions, rounding, overflow and multiply/divide limits.
- Motion and effects: degrees, vectors, aiming,
diagonal movement, squared distances, interpolation and bounded tables.
- Math Lab: a runnable perspective starfield and
orbit/figure-eight/spiral formations, with controls and a playtest checklist.
Use ewrogers/sgdk:latest and check its inc/maths.h, inc/types.h and src/maths.c before using an unfamiliar helper. These references were checked against SGDK 2.11. In particular, current F16_sin/F16_cos take degrees; deprecated sinFix16 uses a different angle convention.
Make the calculation safe
Keep world positions separate from screen coordinates. Prefer fix32 or integer world positions when a level exceeds fix16's range. Use small fix16 offsets and velocities where they fit. Convert at explicit boundaries; do not cast between fixed-point formats as though their scales were identical.
Widen before subtracting or multiplying. Establish a valid divisor and bound the quotient before division. Avoid signed overflow, shifting negative signed values in your own C and assuming that a wrapped result will be useful. SGDK's helpers have implementation-specific limits too; follow the reference rather than replacing them with floating point or an untested fast-math trick.
For a new effect, test known points, negative inputs, near-zero divisors, maximum supported coordinates and a full angle cycle. Compare with an independent higher-precision oracle within a stated tolerance. Execute the actual target SDK routines when host arithmetic cannot establish their behavior.
Keep repeated work small
Cache a shared sine or reciprocal table when the memory tradeoff is worthwhile. Bound the angle/depth index before looking it up. Prefer squared-distance tests to square roots when only a threshold matters. Do not replace an exact hit test with SGDK's approximate distance without allowing for its error.
Use a consistent simulation clock and cache input once per display frame. Separate simulation coordinates from rendering so a PAL display or a missed frame does not silently change the units of velocity. Measure optimized ROMs; a successful build or smooth host animation does not establish a 68000 budget.
