Knight Lore: Routine at D7AC

D7AC: Jump into the row, then the shifted run

The JR's offset byte is patched by patch_sprite_rows: back into the aligned run for a byte-aligned sprite, into the shifted run below for any other. Each shifted unit is 16 bytes of code for one source byte. The left part of the mask and image, looked up on page H, finishes the byte in A, which is stored; the right part, on page H + 1, starts the next buffer byte, which is left in A for the next unit. The last spill is stored after the fifth unit.
sprite_row_jump	D7AC	JR sprite_aligned_tail	offset set by patch_sprite_rows
	D7AE	POP DE	the background byte in A, cleared under the mask shifted right and with the image shifted right put in; stored
	D7AF	LD L,E
	D7B0	AND (HL)
	D7B1	LD L,D
	D7B2	XOR (HL)
	D7B3	CPL
	D7B4	LD (BC),A
	D7B5	INC BC
	D7B6	INC H	the bits that fall out, into the next buffer byte, left in A
	D7B7	LD L,E
	D7B8	LD A,(BC)
	D7B9	AND (HL)
	D7BA	LD L,D
	D7BB	XOR (HL)
	D7BC	CPL
	D7BD	DEC H
	D7BE	POP DE	four more units the same
	D7BF	LD L,E
	D7C0	AND (HL)
	D7C1	LD L,D
	D7C2	XOR (HL)
	D7C3	CPL
	D7C4	LD (BC),A
	D7C5	INC BC
	D7C6	INC H
	D7C7	LD L,E
	D7C8	LD A,(BC)
	D7C9	AND (HL)
	D7CA	LD L,D
	D7CB	XOR (HL)
	D7CC	CPL
	D7CD	DEC H
	D7CE	POP DE
	D7CF	LD L,E
	D7D0	AND (HL)
	D7D1	LD L,D
	D7D2	XOR (HL)
	D7D3	CPL
	D7D4	LD (BC),A
	D7D5	INC BC
	D7D6	INC H
	D7D7	LD L,E
	D7D8	LD A,(BC)
	D7D9	AND (HL)
	D7DA	LD L,D
	D7DB	XOR (HL)
	D7DC	CPL
	D7DD	DEC H
	D7DE	POP DE
	D7DF	LD L,E
	D7E0	AND (HL)
	D7E1	LD L,D
	D7E2	XOR (HL)
	D7E3	CPL
	D7E4	LD (BC),A
	D7E5	INC BC
	D7E6	INC H
	D7E7	LD L,E
	D7E8	LD A,(BC)
	D7E9	AND (HL)
	D7EA	LD L,D
	D7EB	XOR (HL)
	D7EC	CPL
	D7ED	DEC H
	D7EE	POP DE
	D7EF	LD L,E
	D7F0	AND (HL)
	D7F1	LD L,D
	D7F2	XOR (HL)
	D7F3	CPL
	D7F4	LD (BC),A
	D7F5	INC BC
	D7F6	INC H
	D7F7	LD L,E
	D7F8	LD A,(BC)
	D7F9	AND (HL)
	D7FA	LD L,D
	D7FB	XOR (HL)
	D7FC	CPL
	D7FD	DEC H
	D7FE	LD (BC),A	store the last spilled byte