Repository navigation
Releases
Releases · laoo/JagMCP
JagMCP v0.1.6
- Added: recording a video from the window: File > Record video, or
jagmcp game.j64 --record game.mp4(needs ffmpeg). The window now warns that it does not write--frames,--wavor--log;jagmcp-cliwrites them. - Added:
cpu.set_registers(MCP:cpu_set_registers) sets the GPU's and the DSP's registers (r0–r31, flags) and the 68000's (D0–D7, A0–A7, PC, SR, USP, ISP) (#3).cpu.get_stategives the RISC's bank in use. - Fixed: a GPU's phrase store to Tom's 16-bit space wrote only its first word, and its stores through the 32-bit window onto the line buffers ($F09800) were lost (#5, by @42Bastian in #6). Cannon Fodder's "Licensed By Atari Corporation" screen now shows its text whole and fades in and out.
JagMCP v0.1.5
- Added: recording of the picture and the sound to an MP4 through ffmpeg (video.record_start, video.record_stop), sharp 2×2 pixels by default.
JagMCP v0.1.4
- Fixed: the blitter can now read Tom's registers, CLUT and line buffers (e.g. the CLUT as a span buffer), as on the console.
JagMCP v0.1.3
Fixed
- COFF programs froze at start. Before it starts a cartridge, the BIOS leaves a STOP
object at $1000 and sets OLP to point to it. A.cofprogram is loaded after the BIOS
has run, so it finds this list. The emulator did not set it up and left OLP at zero.
A program that turned on the video before it set OLP made the object processor read
RAM from $0 as an object list. That list might never end, so the object processor kept
the bus and the 68000 waited forever. Now RAM holds the STOP object before the program
is loaded, and OLP is set as the BIOS sets it. A program loaded over $1000 replaces the
object, as on the console (decision 0028). Thanks to the forum user DrTypo who found and
reported it.
JagMCP v0.1.2
JagMCP v0.1.2
Fixed
- GPU hang after MMULT. If MMULT came right after loads from main RAM, its matrix reads
were counted as loads from main RAM too. The GPU then waited forever for data that never came.
The hardware does not do this. One homebrew 3D game showed only its background and no 3D
scene; it now draws correctly.
Added
debug.singlestep(control protocol and MCP tooldebug_singlestep):cpu: run one instruction of the 68000, the GPU or the DSP. The other processors run
the same cycles beside it, as on the console. None is held (decision 0030).frame: true: run to the end of the next video frame.- A processor that cannot run (GO clear, a halted 68000, a STOP that no interrupt ends)
answershaltedat once. - A step stops with
reason: "step". The window shows "Step" in the same way. - A run with steps gives the same frames and RAM as a run without them.
Changed
- CI runs only for a pushed
v*tag or when started by hand. It no longer runs on every push
(decision 0029).
Full diff: v0.1.1...v0.1.2
JagMCP v0.1.1
- COFF files (.cof, from rln/aln) run on the 68000 without a BIOS, with
the video set as the BIOS leaves it (decision 0028). - A run ends where it was asked to, whatever the program does: the
window and the servers no longer freeze; emu.status gives bus_wait
(decision 0027). - A read of a Tom address that no register answers gives MEMCON1, as on
the chip, and debug.diagnostics reports it; polygon.j64 now boots
(decision 0026). - cpu.disassemble gives each instruction's bytes (opcode_hex), and
jagmcp.py prints a listing; emu.load and emu.status give the BIOS."