In the emulator

What has been checked and what has not

Let us be clear before anything else: the pictures on this site have not been compared byte by byte against openMSX's VRAM. They have been looked at, and what is on them matches what the code says - the stadium with its scoreboard, the title, the two leagues, the twenty-two poses - but that is not the same as a comparison.

What is checked is that the listing reproduces the ROM byte by byte (make verify) and that not one byte of the cartridge is left unassigned (make sanity). Both have been run after every working round.

How the pictures are drawn

There is not one capture. tools/pantallas.py runs in Python the same script interpreter the Z80 runs, over a 16 KB VRAM in memory, and then develops it as SCREEN 2:

python3 tools/pantallas.py baseball.rom 0x4000 docs/imagenes

Each scene reproduces the cartridge's own sequence, in the same order:

THE TITLE SCREEN, as 0x40cc builds it:
    0x40d1  script 0x47cf ....... patterns and colours
    0x40e2  script 0x4a96
    0x40d6  script 0x4a27 into VRAM 0x6680
    0x40e7  script 0x49d4 into VRAM 0x6780
    0x40f9  replica of VRAM 0x0000 -> 0x0800 (4095 bytes) ..... colours
    0x4102  replica of VRAM 0x2000 -> 0x2800 (4095 bytes) .... patterns

THE FIELD, as 0x4219 builds it:
    0x421e  clearing the name table
    0x4226  script 0x65ad .......... the stadium's patterns and colours
    0x4231  replica of VRAM 0x0000 -> 0x0800
    0x423a  replica of VRAM 0x2000 -> 0x2800
    0x4243  script 0x6a1c ................... the name table

If the picture comes out right it is because the interpreter is correct, not because a photo was looked at. And when the interpreter was wrong, it showed: the first passes at the player poses came out as noise, and that is where it turned out that what was being read as rows of tiles were in fact sprite patterns.

Two comments that lied, caught by drawing

Drawing things uncovers errors that reading the code does not catch. Two turned up in this cartridge:

• 0x44b7 was commented as "the two rows of numbered scoreboard cells". It is not: it is the game's title, 46 consecutive tiles from 0x40 to 0x6D in two rows of 23. It showed the moment the menu screen was drawn and "Konami's Baseball" appeared.
• The second screen was commented as "the team screen". It is not: it is the league screen, CENTRAL and PACIFIC, and the team one comes after it.

Running the cartridge

openmsx -machine Philips_NMS_8250 -carta baseball.rom

Dumping VRAM and comparing for real needs an emulator script, and this repository does not carry one. That comparison is still pending, and it is written down as such on the open questions page; the good places to stop are the ones just below.

What to know before measuring

• The interrupt is everything. Stop the emulator inside the handler at 0x4010 and half the state is half-written. The good places to stop are the wait loops: 0x4260, 0x4266, 0x42ec.
• Buffers are reused several times per frame. 0xE700 is a scratch slot used by at least four different routines; reading it outside the exact instant tells you nothing.
• The randomness is the R register. The cartridge uses it as a die in five places (0x4219, 0x4d69, 0x714b, 0x7196, 0x7d66), so two identical games do not exist even stopping the emulator at the same spot.