Document LY reset behavior

[xenophile127]

From The Cycle-Accurate Game Boy Docs:

"Line 153 takes only a few clocks to complete (the exact timings are below). The rest of the clocks of line 153 are spent in line 0 in mode 1!"

I ran into this behavior debugging Final Fantasy Adventure's intro text scroll.

Quoting @ISSOtm: "LY is reset before VBlank is over". "It's documented neither in the “LCD Status Registers” nor “Rendering”... this is indeed a problem".

Preferably this would be documented visually somehow in PPU modes. While it sounds like the real behavior is that LY is wrong for part of line 153 I was looking at the diagram's representation of line 0.

[ISSOtm]

To be precise, line 153 lasts for its normal duration (in particular, STAT continues reporting Mode 1 accurately), but LY gets reset earlier than that. (Note also that, since LYC is compared to LY itself, a LY=LYC == 0 interrupt does get triggered early.)

I think hatching part of line 153 and adding a note below the diagram would work.

@LIJI32 pointed out that TCAGBD documents some “PPU odd mode” timings, so its single-speed measurements shouldn't all be trusted (especially around transitions). But this behaviour is correct.

[LIJI32]

They should be reliable for DMG (and CGB and DMG mode) since you can't get into odd mode in these cases.