Skip to content

roadmap.md

Latest commit

 

History

History
92 lines (81 loc) · 5.02 KB

roadmap.md

File metadata and controls

92 lines (81 loc) · 5.02 KB
Raw

Roadmap

This is the practical next-work list. It tracks what is visibly wrong or missing after the current documentation pass.

Immediate Visual Fixes

  • Keep checking HUD number and ammo-counter alignment against fresh screenshots whenever status-bar assets or weapon placement change.
  • Keep the current safer near-wall clamp unless the WAD render-line refinement proves a better close-wall strategy visually.
  • Recheck weapon vertical placement after the HUD number pass so the gun and status face do not visually fight each other.
  • Keep checking shotgun fire and pump frames against combat smoke captures; the weapon frame is centered, but palette and smoke timing changes can still affect how readable the fired frame feels.
  • Update screenshots whenever the HUD or weapon placement changes.

Gameplay And Assets

  • Keep hardware-checking the direct C+D-pad and diagonal weapon shortcuts with tools/smoke_weapon_shortcuts.sh, then tune the mapping if diagonals are awkward on the target controls.
  • Improve BFG/plasma fidelity under IWADs that actually provide those psprites: BFG trace behavior, projectile tuning, and visible pickup sprites when the source WAD provides them.
  • Complete enemy sprite coverage beyond the current A/B eight-way walk rotation groups plus partial rotated attack/pain coverage: full death rotations, registered/Doom II reaction rotations, and more faithful thing placement still need work.
  • Continue tuning line-of-sight, wall-depth fallback, and encounter placement now that monster tiles fit inside the visible C-ROM tile range and world sprites use a more stable floor-baseline anchor. Normal builds now preserve converted WAD monster placement; use DOOM_REVEAL_HIDDEN_MONSTERS only as an explicit debug aid.
  • Add exact sprites for any registered-only powerup frames that are not present in the shareware WAD, then consider HUD countdown/status feedback beyond the current timed-powerup palette tint.

Rendering Fidelity

  • Use the generated per-cell floor/ceiling heights for more than sprite seating. The current renderer now knows the heights, but it still does not draw true Doom floor/ceiling clipping or stacked multi-span sector geometry.
  • Continue tuning the cached perspective floor/ceiling pass. The stable default keeps one phase per direction because multi-phase floor banks exceeded the practical sprite tile index range; any future forward-motion cue needs a smaller tile layout, fewer plane rows/directions, or a non-tile-index-heavy approximation.
  • Extend the current WAD render-line refinement into a true higher-fidelity path using the generated grid/q8 BSP vertex/node data. The converter now emits and verifies that data; the runtime still needs the front-to-back visible-seg owner pass that feeds the existing sprite-strip buffers.
  • Extend the new one-span two-sided wall approximation into real multiple clipped spans for windows, ledges, and upper/lower sector transitions; these are still a main reason the native Doom E1M1 start view does not match exactly even though the same map and player start are loaded.
  • Keep checking native-vs-NeoGeo waypoints after converter changes. The centered 48x36 map, route-validated DOOM_MAP_DETAIL_CULL=0.5 collision grid, and DOOM_RENDER_DETAIL_CULL=1.5 visual-line pass deliberately reduce runtime map complexity while retaining larger Doom room-edge cues. Current E1M1/E1M2 start comparisons still show large visual differences from native Doom. The converter now tags lower/upper spans with sidedef ownership, so the next major renderer gap is drawing more than one visible sector/span in a column instead of replacing the far wall with one selected partial span.
  • Re-run native-vs-NeoGeo side-by-side captures after manual emulator testing is done. The current scripts kill existing ngdevkit-gngeo processes, so they should not be run while someone is interactively testing a ROM.
  • Experiment with diagonal wall or multi-span approximations within Neo Geo sprite limits.
  • Profile wall, plane, thing, and HUD update costs before increasing sprite or map complexity.

Map And Performance

  • Keep refining minimap responsiveness after the incremental open/close pass: redraw only changed cells where possible.
  • Keep make episode-route-check green as the Episode 1 conversion baseline: E1M1-E1M7 and E1M9 route through the generated grid, and E1M8 uses the boss-death completion path.
  • Reduce runtime work in monster selection/projection.
  • Add a repeatable screenshot smoke test to CI for the Linux ROM build.
  • Convert more maps once E1M1 visual and gameplay fundamentals are stable.

Audio And Flow

  • Add YM2610/Z80 sound effect playback.
  • Convert Doom sound lumps to Neo Geo-friendly sample data.
  • Decide whether music should be simplified YM2610 arrangements, sample-based approximations, or skipped until the visual/gameplay port is stronger.
  • Replace the current standalone-map progression prompt with a true multi-map episode/intermission flow once multiple generated maps can live in one cart.