Skip to content
Eliott65wiki
Contents

Where the project stands

Eliott65 moves forward in small versions, each measured on the board before being delivered. Here is what works, what is under construction and what is waiting.

On this page

What works, validated on the board

AreaStatus
6502 busserved by PIO and DMA, with no ARM core; TURBO mode by default; at 270 MHz, 7.707 MHz measured by the 6502 with video on (v0.59.1)
System calls2-byte gate, ≈ 5 µs per call; 7 services, 47 functions
VideoHDMI 720 × 480 at 60 Hz, one signal for every mode (the default since v0.52; stable picture on the screen with v0.58); 40 × 30 text in 16 colours, read directly from the 6502’s memory
KeyboardUSB, read by an in-house driver, delivered to the 6502 without a system call; US and FR layouts
USB stickreading and writing, multi-block commands
FilesFAT12, FAT16, FAT32, long names; launching a program from the stick
Multitaskingup to 8 programs, cooperative and pre-emptive; shell with background tasks and Ctrl+Tab, commands in English (HELP, PS, KILL)
Hibernationthe whole machine in a .HIB file, exact resumption
Graphics320 × 240, two tile or bitmap layers, 128 sprites, 256-colour palette; core 0 draws, core 1 encodes; 13 measurement scenarios out of 16 fit in the time allowed
Soundthe board’s speaker: notes and melodies without blocking the program (SON service)
Emulatorthe same kernel on the PC, with test benches for the bus, video and USB; matches the board on screens and memory; debugger (trace, breakpoints, watches)

Under construction

  • The Hercules mode (720 × 350 monochrome, 80 × 25 text): the 720 × 480 signal became the default (v0.52), the graphics part exists (v0.53), so does the 80 × 25 text (v0.54), with a 9 × 14 font drawn for the project (v0.58), and the shell can switch to 80 columns (v0.57). Exact on the bench, but too slow on the board: measured in v0.59, they overrun the time to prepare a line. v0.60 speeds them up, not yet checked on the board. See Video.
  • An open problem on the board: at 270 MHz, the GRAPHE demo stops calling the kernel after a few minutes. The bus model points to a plausible cause; the fix has been on the board since v0.59 (7.707 MHz measured), and a long watch by the wireless probe is under way. See The bus.
  • Sound over the HDMI socket: a test tone travels in the data islands and the bench decodes it correctly, but it has not yet been confirmed by ear; set aside at the user’s request. See Sound.

Waiting

  • USB floppy drive: the FAT12 format is ready and tested on 1.44 MB floppy disk images. A real drive still has to be plugged in to find out whether it speaks the Bulk-Only protocol (like sticks) or UFI/CBI, common among floppy drives.

Under study or in the backlog

  • a Forth ported to the Eliott65 interface;
  • on-demand virtual memory, so that a program can exceed 64 KiB;
  • the mouse: it is already recognised on the hub, but not yet used.

Timeline

The project started on 13 October 2026 with a feasibility study. The main milestones:

VersionMilestone
0.1Feasibility study
0.2The 6502 runs for the first time on the Eliott65 bus — first time round
0.3Bus served by PIO + DMA; both ARM cores are freed
0.4In-house DVI video driver, text screen
0.5 – 0.6In-house USB driver: keyboard, then USB stick
0.7 – 0.11FAT32, launching programs, long names, writing, FAT12 and FAT16
0.12Hibernation
0.13 – 0.18Multitasking: two spaces, pre-emption by NMI, shell, 8 processes, sleep
0.19 – 0.26Host emulator, test benches for the bus, video and USB, SDL window
0.27 – 0.42Graphics: engine, services, display on the board, collaboration between the two cores, sprites in assembler
0.43 – 0.47Real HDMI, redesigned video DMA, test tone over HDMI; EDID turns out to be impossible on this board
0.48The board’s speaker plays: SON service
0.49 – 0.50Faster sprites (13 scenarios out of 16); 720 × 480 signal at 270 MHz as an option, first step of the Hercules mode
0.51 – 0.52HDMI in 720 × 480, then this signal as the default for every mode; candidate bus fix at 270 MHz
0.53Hercules mode, graphics part; the whole project under the EUPL 1.2
0.54Hercules mode, 80 × 25 text (provisional font)
0.55 – 0.564 KiB given back to the firmware (page of zeros in the pool); emulator debugger
0.57 – 0.58Shell in 80 × 25 and commands in English; Eliott65’s 9 × 14 font
0.59 – 0.60Tools through the wireless probe; v0.58 seen on screen, stable picture; 7.707 MHz measured; faster Hercules modes

The details of each version are recorded in the repository’s development log; this wiki covers the essentials by topic.

Sources in the repository : docs/agile/BACKLOG.md, docs/journal/README.md.