Fixes Gameboy output

* Fixes Gameboy output
* Documentation changes for Gameboy
* Gameboy example now works

4 files changed, 280 insertions, 60 deletions

diff --git a/doc/casm.html b/doc/casm.html
index 245a1ee..dcd21bd 100644
--- a/doc/casm.html
+++ b/doc/casm.html
@@ -140,8 +140,23 @@ along with this program.  If not, see
 http://www.gnu.org/licenses/gpl-3.0.html)</a></p>
 </div>
 
+<h1>Index</h1>
+
+<p>
+<a href="#casm">CASM</a> - General usage
+</p>
+<p>
+<a href="#z80">Z80</a> - Z80 processor support.
+</p>
+<p>
+<a href="#6502">6502</a> - 6502 processor support.
+</p>
+<p>
+<a href="#gbcpu">Gameboy Z80</a> - The Gameboy Z80 derivative processor.
+</p>
 
-<h1>CASM</h1>
+
+<h1 id="casm">CASM</h1>
 
 <h2>Usage</h2>
 
@@ -885,6 +900,13 @@ A ZX81 emulator P file.
 A Commodore 64 T64 tape file.
 </td></tr>
 
+<tr><td class="cmd">
+<a href="#gameboyout">gameboy</a>
+</td>
+<td class="def">
+A Nintendo Gameboy ROM file.
+</td></tr>
+
 </table>
 
 </td></tr>
@@ -1012,6 +1034,75 @@ to start the machine code, e.g.
 <p>Any remaining blocks will be stored as-is without any basic loader.</p>
 
 
+<h3 id="gameboyout">Nintendo Gameboy ROM File Output Format</h3>
+
+<p>
+Generates a ROM file for a Gameboy emulator or hardware.  Note that large ROM
+sizes have not been extensively checked and verified.
+</p>
+
+<p>If a single bank was used during the assembly then a simple 32K ROM is
+assumed, and an error will be shown if the addresses used fall outside the range
+0x150 to 0x7fff. </p>
+
+<p>Similarly if multiple banks are used then it is assumed that the first bank
+is only used in the range 0x150 to 0x3fff, and subsequent banks in the range
+0x4000 to 0x7fff.  This is to fit in with the method the Gameboy uses to page
+memory banks into the upper 16K portion of the normal 32K ROM address space.
+</p>
+
+<p>By default the output driver will try and fill in the ROM size and type in
+the header properly, but these can be overridden using settings.</p>
+
+<h4>Gameboy ROM Output Format options</h4>
+
+<p>The Gameboy output driver supports the following settings that can be set via
+an <b>option</b> command.
+</p>
+
+<table>
+
+<thead><tr><td class="head">Option</td>
+<td class="head">Description</td></tr></thead>
+
+<tr><td class="cmd">
+option gameboy-colour, &lt;on|off&gt;
+</td>
+<td class="def">
+Whether this is a Gameboy Colour cartridge.  Defaults to <i>off</i>.
+Note that <b>gameboy-color</b> can be used as a different spelling for this
+setting.
+</td></tr>
+
+<tr><td class="cmd">
+option gameboy-super, &lt;on|off&gt;
+</td>
+<td class="def">
+Whether this is a Gameboy Super extended cartridge.  Defaults to <i>off</i>.
+</td></tr>
+
+<tr><td class="cmd">
+option gameboy-cart-type, &lt;on|off&gt;
+</td>
+<td class="def">
+Specifies the cartridge type.  Defaults to -1, which means the output driver
+will pick the appropriate type.
+</td></tr>
+
+<tr><td class="cmd">
+option gameboy-irq, <i>irq</i>, <i>address</i>;
+</td>
+<td class="def">
+Specifies an address where an IRQ routine is stored, and requests that the
+IRQ be transferred to that address when it happens.  The IRQ routine must end
+with a <b>reti</b> opcode.<br><br>
+<i>irq</i> can be either <b>vbl</b>, <b>lcd</b>, <b>timer</b>, <b>serial</b>
+or <b>joypad</b>.  If left at the default value of -1 then an IRQ handler is
+installed with just a <b>reti</b> instruction.
+</td></tr>
+
+</table>
+
 <h2>Listing</h2>
 
 <p>
@@ -1138,7 +1229,7 @@ down to a single blank line in the listing.
 </table>
 
 
-<h1>Z80 CPU</h1>
+<h1 id="z80">Z80 CPU</h1>
 
 <h2>Opcodes</h2>
 
@@ -1239,7 +1330,7 @@ can be any value at all:
 
 The Z80 assembler has no options.
 
-<h1>6502 CPU</h1>
+<h1 id="6502">6502 CPU</h1>
 
 <h2>Opcodes</h2>
 
@@ -1308,6 +1399,115 @@ e.g.
 </td></tr>
 </table>
 
+<h1 id="gbcpu">Gameboy Z80 derivative CPU</h1>
+
+<h2>Opcodes</h2>
+
+<p>
+The Gameboy assembler uses the standard Z80 opcodes where applicable.
+Note that the Gameboy processor has a reduced number of opcodes, flags
+and no index registers, though it has some additional instructions and
+addressing modes.
+</p>
+
+<p>
+For instructions were the Accumulator can be assumed it can be omitted, and
+EOR can be used the same as XOR:
+</p>
+
+<pre class="codeblock">
+        xor     a,a         ; These are equivalent
+        xor     a
+        eor     a,a
+
+        and     a,b         ; These are equivalent
+        and     b
+</pre>
+
+<p>
+The Gameboy CPU has a special addressing mode used for one opcode, where the
+referenced address is stored as a single byte, and used as an offset into the
+top page (0xff00).  This can be either triggered by using the special opcode, or
+will automatically used whenever an address is accessed in the range 0xff00 to
+0xffff:
+</p>
+
+<pre class="codeblock">
+        ; These all will use the special addressing mode opcode, accessing
+        ; memory location $ff34
+        ;
+label   equ     $ff34
+
+        ldh     a,($34)
+        ldh     a,($ff34)
+        ld      a,($ff34)
+        ld      a,(label)
+
+        ld      (label),a
+        ld      ($ff34),a
+        ldh     ($34),a
+        ldh     ($ff34),a
+</pre>
+
+<p>
+The Gameboy CPU also supports incrementing or decrementing the HL register when
+it is used as an address:
+</p>
+
+<pre class="codeblock">
+        ; All these decrement HL after the value has been used.
+        ;
+        ld      a,(hl-)
+        ld      a,(hld)
+        ldd     a,(hl)
+        ld      (hl-),a
+        ld      (hld),a
+        ldd     (hl),a
+
+        ; All these increment HL after the value has been used.
+        ;
+        ld      a,(hl+)
+        ld      a,(hli)
+        ldi     a,(hl)
+        ld      (hl+),a
+        ld      (hli),a
+        ldi     (hl),a
+</pre>
+
+<p>In addition the Gameboy CPU supports these extra instructions over the Z80:
+</p>
+
+<pre class="codeblock">
+        ; Actually loads using the address $ff00 + C
+        ;
+        ld      a,(c)
+        ld      (c),a
+
+        ; Put the Gameboy into a low-power mode till a control is pressed.
+        ; Note it is accepted practice to put a NOP afterwards.  This may be
+        ; due to the stop replacing DJNZ, which may still be wired to expect
+        ; an argument.  That is just a wild guess though.
+        ;
+        stop
+        nop
+
+        ; Swaps the low/high nibbles of the register
+        ;
+        swap    a
+        swap    b
+        swap    c
+        swap    d
+        swap    e
+        swap    h
+        swap    l
+        swap    (hl)
+</pre>
+
+
+<h2>Options</h2>
+
+The Gameboy CPU assembler has no options.
+
 <!--  vim: ai sw=4 ts=8 expandtab spell
 -->
 </body>
diff --git a/src/Makefile b/src/Makefile
index 6ac1bdf..baadcb4 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -79,14 +79,14 @@ clean:
 alias.o: alias.c global.h basetype.h util.h state.h alias.h
 casm.o: casm.c global.h basetype.h util.h state.h expr.h label.h macro.h \
  cmd.h parse.h codepage.h stack.h listing.h alias.h output.h rawout.h \
- specout.h t64out.h zx81out.h z80.h 6502.h gbcpu.h
+ specout.h t64out.h zx81out.h gbout.h z80.h 6502.h gbcpu.h
 codepage.o: codepage.c global.h basetype.h util.h state.h codepage.h \
  parse.h cmd.h
 expr.o: expr.c global.h basetype.h util.h state.h expr.h label.h
 gbcpu.o: gbcpu.c global.h basetype.h util.h state.h expr.h label.h \
  parse.h cmd.h codepage.h varchar.h gbcpu.h
-gbout.o: gbout.c global.h basetype.h util.h state.h codepage.h parse.h \
- cmd.h gbout.h
+gbout.o: gbout.c global.h basetype.h util.h state.h expr.h codepage.h \
+ parse.h cmd.h gbout.h
 label.o: label.c global.h basetype.h util.h state.h codepage.h parse.h \
  cmd.h stack.h label.h
 listing.o: listing.c global.h basetype.h util.h state.h label.h macro.h \
@@ -94,7 +94,7 @@ listing.o: listing.c global.h basetype.h util.h state.h label.h macro.h \
 macro.o: macro.c global.h basetype.h util.h state.h codepage.h parse.h \
  cmd.h varchar.h macro.h
 output.o: output.c global.h basetype.h util.h state.h output.h parse.h \
- cmd.h rawout.h specout.h t64out.h zx81out.h
+ cmd.h rawout.h specout.h t64out.h zx81out.h gbout.h
 parse.o: parse.c global.h basetype.h util.h state.h codepage.h parse.h \
  cmd.h
 rawout.o: rawout.c global.h basetype.h util.h state.h rawout.h parse.h \
diff --git a/src/example/gb.asm b/src/example/gb.asm
index 10381e9..2f52e0e 100644
--- a/src/example/gb.asm
+++ b/src/example/gb.asm
@@ -1,22 +1,26 @@
+	;
+	; This example is poor, and leaves lots of junk lymg in memory.
+	; Still, it works enough to test
+	;
 	
 	cpu	gameboy
 
 	option	output-file,gb.gb
 	option	output-format,gameboy
 
-	;option	gameboy-irq,vbl,vbl_code
+	option	gameboy-irq,vbl,vbl_code
 
 VRAM	equ	$8000
 SCRN	equ	$9800
 OAM	equ	$fe00
 LCDC	equ	$ff40
 STAT	equ	$ff41
+ISWITCH	equ	$ffff
 BGRDPAL	equ	$ff47
 OBJ0PAL	equ	$ff48
 OBJ1PAL	equ	$ff49
 CURLINE	equ	$ff44
 
-READY	equ	$ff81
 XPOS	equ	$ff82
 
 VBLANK	macro
@@ -39,31 +43,29 @@ VBLANK	macro
 	org	$150
 
 	di
-	xor	a
-	ldh	(READY),a
 	ld	sp,$fffe
 
-	; Set LCD so only sprites show
+	; Switch of display during setup
 	;
 	VBLANK
+	xor	a
+	ld	(LCDC),a
+
+	ld	(XPOS),a
 
-	ld	a,$81
-	ldh	(LCDC),a
-	ld	a,$10
-	ldh	(STAT),a
 	ld	a,$e4
-	ldh	(BGRDPAL),a
 	ldh	(OBJ0PAL),a
 	ldh	(OBJ1PAL),a
 
-	; Copy to VRAM
+	swap	a
+	ldh	(BGRDPAL),a
+
+	; Copy to VRAM and set up
 	;
 	ld	hl,VRAM
 	ld	de,sprite
 	ld	c,16
 
-	VBLANK
-
 .copy
 	ld	a,(de)
 	ld	(hl+),a
@@ -71,12 +73,6 @@ VBLANK	macro
 	dec	c
 	jr	nz,copy
 
-	ld	a,1
-	ldh	(READY),a
-
-	ei
-
-	VBLANK
 
 	; Set sprite numbers
 	;
@@ -92,9 +88,24 @@ VBLANK	macro
 	ld	(OAM+7),a
 	ld	(OAM+11),a
 
+	; Set LCD back on
+	;
+	ld	a,$83
+	ldh	(LCDC),a
+
+	; Activate VBL
+	;
+	ld	a,1
+	ldh	(ISWITCH),a
+
+	ei
+
 .idle
-	VBLANK
+	halt
+	nop
+	jr	idle
 
+vbl_code:
 	ldh	a,(XPOS)
 	inc	a
 	ldh	(XPOS),a
@@ -110,23 +121,6 @@ VBLANK	macro
 	ld	(OAM+8),a
 	ld	(OAM+9),a
 
-	jr	idle
-
-vbl_code:
-	ldh	a,(READY)
-	or	a
-	jr	z,finish
-
-	ldh	a,(XPOS)
-	inc	a
-	ldh	(XPOS),a
-	ld	(OAM),a
-	ld	(OAM+1),a
-	xor	a
-	ld	(OAM+2),a
-	ld	(OAM+3),a
-
-.finish
 	reti
 
 sprite:
@@ -141,3 +135,21 @@ sprite:
 	defb	$ff,$ff
 	defb	$00,00
 
+	;
+	; Assembly checks
+	;
+	stop
+	swap	a
+	swap	b
+	swap	c
+	swap	d
+	swap	e
+	swap	h
+	swap	l
+	swap	(hl)
+	ld	a,(hl+)
+	ld	a,(hli)
+	ldi	a,(hl)
+	ld	a,(hl-)
+	ld	a,(hld)
+	ldd	a,(hl)
diff --git a/src/gbout.c b/src/gbout.c
index 24e365a..71547f9 100644
--- a/src/gbout.c
+++ b/src/gbout.c
@@ -41,7 +41,6 @@ enum option_t
     OPT_SUPER,
     OPT_CART_RAM,
     OPT_CART_TYPE,
-    OPT_RST,
     OPT_IRQ
 };
 
@@ -51,7 +50,7 @@ static const ValueTable option_set[] =
     {"gameboy-color",           OPT_COLOUR},
     {"gameboy-super",           OPT_SUPER},
     {"gameboy-cart-ram",        OPT_CART_RAM},
-    {"gameboy-cart-type",       OPT_CART_RAM},
+    {"gameboy-cart-type",       OPT_CART_TYPE},
     {"gameboy-irq",             OPT_IRQ},
     {NULL}
 };
@@ -193,8 +192,8 @@ int GBOutput(const char *filename, const char *filename_bank,
     int f;
     int offset;
     int rom_size;
-    unsigned global_csum = 0;
-    unsigned hdr_csum = 0;
+    Word global_csum = 0;
+    Byte hdr_csum = 0;
     Byte *mem;
 
     if (!fp)
@@ -374,35 +373,43 @@ int GBOutput(const char *filename, const char *filename_bank,
 
     /* Header checksum
     */
+    hdr_csum = 0;
+
     for(f = 0x134 ; f < 0x14d; f++)
     {
-        hdr_csum -= mem[f] - 1;
+        hdr_csum = hdr_csum - mem[f] - 1u;
     }
 
     PokeB(mem, 0x14d, hdr_csum);
 
     /* Global checksum
     */
+    global_csum = 0;
+
     if (count == 1)
     {
-        for(f = 0; f < 0x8000; f++)
+        for(f = 0; f < 0x14e; f++)
         {
-            if (f < 0x14e || f > 0x14f)
-            {
-                global_csum += mem[f];
-            }
+            global_csum += mem[f];
+        }
+
+        for(f = 0x150; f < 0x8000; f++)
+        {
+            global_csum += mem[f];
         }
     }
     else
     {
         int r;
 
-        for(f = 0; f < 0x4000; f++)
+        for(f = 0; f < 0x14e; f++)
         {
-            if (f < 0x14e || f > 0x14f)
-            {
-                global_csum += mem[f];
-            }
+            global_csum += mem[f];
+        }
+
+        for(f = 0x150; f < 0x4000; f++)
+        {
+            global_csum += mem[f];
         }
 
         for(r = 1; r < count; r++)
@@ -414,7 +421,8 @@ int GBOutput(const char *filename, const char *filename_bank,
         }
     }
 
-    PokeW(mem, 0x14e, global_csum);
+    PokeB(mem, 0x14e, global_csum >> 8);
+    PokeB(mem, 0x14f, global_csum);
 
     /* Output the ROM contents
     */