JIT bitmap text rendering for microcontrollers

A race-the-beam renderer for an RP2350 home-computer project needs to draw text within a tight scanline deadline while preserving SRAM for the application. Fixed-cell hardware text is fast but cannot provide proportional spacing or kerning; a conventional framebuffer offers flexibility but consumes scarce memory and processing time. The proposed alternative uses just-in-time compilation: when a text display object becomes active, its string is compiled into specialized ARM code. That code draws glyphs into a one-bit-per-pixel scratch buffer, which is then expanded to RGB565. The display list is double-buffered, and objects are activated and removed as their vertical bounds cross each scanline.

The approach relies on carefully packed fonts and instruction scheduling. Glyph ink shares a bitmap, with overlapping shapes reused; packing keeps each glyph’s row within a 32-bit word so it can be loaded efficiently. Glyph dimensions, bearings, and advance widths fit into compact metadata, supporting proportional text with little display-list overhead. Generated code extracts glyph bits and combines them into output words, while reordering instructions hides a load-to-use delay. The basic JIT loop costs 23 cycles per glyph, and the one-bit-to-color pass runs at 1.75 cycles per pixel - important against a budget of roughly six cycles per pixel. The technique aims to combine crisp bitmap typography and flexible spacing with the speed and low memory use needed for microcontroller video rendering.