The 2022 version worked but had some loose ends. Unicode support was inflexible–couldn’t mix ASCII with Katakana; Phosphor decay was stored in a separate array when it should’ve been packed with RGB; Code was harder to read than it needed to be.

Moved phosphor decay into the 4th byte of the RGB union–should’ve done this in 2022. What was I thinking.

Keeping the RGB union despite portability concerns. All my systems are little-endian and the code is cleaner this way.

Fixed Unicode by introducing a charset array. UNICODE(min, max) packs Unicode ranges into uint64–low four bytes for start, high four bytes for end. insert_code() unpacks a random block and picks a character from it:

static uint64_t glyphs[] = {
    UNICODE(0x0021, 0x007E), /* ASCII */
    UNICODE(0xFF65, 0xFF9F), /* Half-width Katakana */
};

Full-width Katakana breaks column alignment. Stick to half-width (U+FF61-U+FF9F) range. Compile with -DNOKANA to disable Katakana altogether.

blend() simulates phosphor decay by eroding RGB channels toward background color:

static inline void blend(matrix *mat, size_t row, size_t col)
{
    unsigned char *color = mat->rgb[index(mat, row, col)].color;
    color[R] = color[R] - (color[R] - RGB_BG_RED) / DECAY_MPLIER;
    color[G] = color[G] - (color[G] - RGB_BG_GRN) / DECAY_MPLIER;
    color[B] = color[B] - (color[B] - RGB_BG_BLU) / DECAY_MPLIER;
}

That’s still good. Leaving it alone.

Tossed license and automake cruft. Just cc -O3 main.c -o matrix now. Don’t need the ceremony.

Runs at 2-3% CPU on OpenBSD (T490). No regressions. Fans are quiet.

Commit: 03f8d87