title |
---|
Octo Assembly Language |
Octo programs are a series of tokens separated by whitespace. Some tokens represent Chip-8 instructions and some tokens are directives which instruct Octo to do some special action as the program is compiled. The :
directive, followed by a name (which cannot contain spaces) defines a label. A label represents a memory address- a location in your program. You must define at least one label called main
which serves as the entrypoint to your program.
Using a label by itself will perform a subroutine call to the address the label represents. Alternatively, you can be more explicit by using :call
followed by an address or name. A semicolon (;
) is another way to write return
, which returns from a subroutine. The #
directive is a single-line comment; it ignores the rest of the current line. Numbers can be written using 0x
or 0b
prefixes to indicate hexadecimal or binary encodings, respectively.
Numeric constants can be defined with the :const
directive followed by a name and then a value, which may be a number, another constant or a (non forward-declared) label. Registers may be given named aliases with :alias
followed by a name and then a register or a constant expression 0-15 enclosed in curly braces ({ ... }
). The i
register may not be given an alias, but v registers can be given as many aliases as desired. Here are some examples of constants and aliases:
:alias x v0
:alias CARRY_FLAG vF
:alias NTH { 3 + CALLS }
:const iterations 16
:const sprite-height 9
Chip-8 has 16 general-purpose 8-bit registers named v0
to vF
. vF
is the "flag" register", and some operations will modify it as a side effect. i
is the memory index register and is used when reading and writing memory via load
, save
and bcd
, and also provides the address of the graphics data drawn by sprite
. Sprites are drawn by XORing their pixels with the contents of the screen. The screen is 64 pixels wide and 32 pixels tall, and sprites drawn off the edge of the display will wrap around. Drawing a sprite sets vF
to 0, or to 1 if drawing the sprite toggles any pixels which were previously on. For a more detailed description of the behavior of Chip-8 instructions consult The Chip-8 Technical Reference.
In the following descriptions, vx
and vy
refer to some register name (v0-vF) and n
refers to some number. The Chip-8 keypad is represented on your keyboard as follows:
Chip-8 Key Keyboard
---------- ---------
1 2 3 C 1 2 3 4
4 5 6 D q w e r
7 8 9 E a s d f
A 0 B F z x c v
For convenience, Octo predefines constants beginning with OCTO_KEY_
for each keyboard key with the value of the corresponding Chip-8 key. For example, OCTO_KEY_W
has a value of 5
.
return
return from the current subroutine. (alias for ;)clear
clear the screen.bcd vx
decode vx into BCD at i, i+1, i+2.save vx
save registers v0-vx to i and advance i. (see note below.)load vx
load registers v0-vx from i and advance i. (see note below.)sprite vx vy n
draw a sprite at x/y position, n rows tall.jump n
jump to address.jump0 n
jump to address n + v0.
note: The load
and save
instructions postincrement i
by x
+1. For example, load v3
will add 4 to i
after loading 4 bytes of memory into the first 4 v
registers.
The various Chip-8 copy/fetch/arithmetic opcodes have been abstracted to mostly fit into a consistent <dest-reg> <operator> <source>
format. For some instructions, <source>
can have several forms.
delay := vx
set delay timer to register value.buzzer := vx
set sound timer to register value.i := n
set i to constant.i := hex vx
set i to hex char corresponding to register value.i += vx
increment i by a register value.vx := vy
copy register to register.vx := n
set register to constant.vx := random n
set register to random number AND n.vx := delay
set register to delay timer.vx := key
block for a keypress and then store code in register.vx += n
add constant to register.vx -= n
subtract constant from register. (see note below.)vx += vy
add register to register. (set vF to 1 if result overflows, else 0)vx -= vy
subtract y from x, store in x (set vF to 0 if result underflows, else 1)vx =- vy
subtract x from y, store in x (set vF to 0 if result underflows, else 1)vx |= vy
bitwise OR register with register.vx &= vy
bitwise AND register with register.vx ^= vy
bitwise XOR register with register.vx >>= vy
shift vy right by 1 and store result in vx. (set vF to LSB of vy)vx <<= vy
shift vy left by 1 and store result in vx. (set vF to MSB of vy)
note: vx -= n
is syntactic sugar for vx += -n
, (or in other words, adding the two's complement of n) and does not represent a distinct instruction.
The Chip-8 conditional opcodes are all conditional skips, so Octo control structures have been designed to map cleanly to this approach. The following conditional expressions can be used with if
or while
:
vx == n
vx != n
vx == vy
vx != vy
vx key
(true if the key indicated by vx is pressed)vx -key
(true if the key indicated by vx is not pressed)
if...then
conditionally executes a single statement. For example,
if v0 != 5 then v1 += 2
Octo also provides pseudo-ops for using <
, >
, <=
and >=
to compare two registers or a register with a constant:
if v1 > v2 then v3 := 5
if v1 <= 0xA then v3 := 7
These are implemented by using the subtraction instructions -=
and =-
and querying vf
. Note that these pseudo-ops produce 3 Chip-8 instructions each and should be avoided when the simpler direct comparisons are suitable.
If you wish to conditionally execute a group of statements, you can use if...begin...end
instead of if...then
. Optionally you may include an else
clause.
if v0 > 5 begin
v1 := random 0xFF
if v1 == 5 begin
v2 := v1
v3 := v1
else
delay := v1
end
end
if...begin...else...end
will not always be the fastest or most compact way to express your desired conditions. Consider rearranging your logic, using jump tables with jump0
or factoring the bodies of conditional clauses into subroutines if they are reused elsewhere. if...begin...end
requires more instructions than a plain if...then
, so prefer the latter when practical.
loop...again
is an unconditional infinite loop. loop
marks the address of the start of the loop and produces no code, while again
compiles a jump instruction based on the address provided by loop
. Since again
is itself a statement, we can use an if...then
at the end of a loop to skip over the backwards jump and efficiently break out of the loop. The following loop will execute 5 times:
v0 := 0
loop
# do something...
v0 += 1
if v0 != 5 then
again
The other way to break out of a loop is while
. while
creates a conditional skip around a forward jump. These forward jumps are resolved by again
to point to the address immediately outside their loop. while
will thus exit the current loop if its condition is not true. You can have as many while
statements in a loop as you want. Here is an example of while
which is similar to the previous except for when the condition is checked.
v0 := 0
loop
v0 += 1
while v0 != 5
# do something...
again
loop...again
constructs may be nested as desired and will behave as expected, but note that simply chaining together if...then
statements (as in if v0 == 0 then if v1 == 1 then v2 := 4
) does not elicit useful behavior.
Sometimes you may wish to have the 12-bit address represented by a label available in v
registers. Octo provides a command called :unpack
for this purpose which expands into a pair of register assignment opcodes. It takes a nybble (0-15 numeric literal or constant) followed by a label as arguments. The lower 8 bits of the address will be stored in v1
and the upper 4 bits of the address will be bitwise ORed with the specified nybble shifted left 4 bits and stored in v0
. If the label cucumber
represented the address 0x582
, the following sets of statements would be identical in meaning:
v0 := 0xA5
v1 := 0x82
:unpack 0xA cucumber
This operation makes it possible to write self-modifying code without hardcoding addresses as numeric literals. If you wish to unpack addresses into registers other than v0
and v1
you can define aliases called unpack-hi
or unpack-lo
, respectively. If you want a 16-bit address, you can use the keyword long
in place of the high nybble:
:unpack long cucumber
Another type of self-modifying code that comes up frequently is overwriting the second half of an instruction, particularly instructions like vX := NN
whose second byte is an immediate operand. This requires a label at the second byte of an instruction, which can be achieved with :next
:
: init :next target va := 2 ;
...
i := target
v0 := 5
save v0
init # va will be set to 5 instead of 2.
The behavior of :unpack
and :next
could also be obtained in a more general case using macros- see below. The main motivation for using these over macros with :calc
is that with :unpack
you may forward-reference labels which have not yet been defined.
You can also specify an address at which subsequent instructions should be compiled by using :org
followed by an address. The use of this directive is very brittle, so it should be avoided unless absolutely necessary.
Sometimes your code will contain repetitive patterns that don't make sense to break out into subroutines. Perhaps they differ by the registers they operate upon, or for performance reasons you need to avoid the overhead of a call and a return. The :macro
command is the solution. It takes a name, followed by names for 0 or more arguments, then a {
, a sequence of arbitrary Octo statements and finally a terminal }
. When you reference the name of a macro, you must provide tokens corresponding to each argument, and then Octo will inline the contents of the macro with any instances of the argument names substituted by the input tokens. Here's a trivial use and definition example:
:macro swap A B {
vf := A
A := B
B := vf
}
...
swap v0 v1
swap v2 v1
This generates code equivalent to the following:
vf := v0
v0 := v1
v1 := vf
vf := v2
v2 := v1
v1 := vf
Macros must be defined before expansion, and nesting macro definitions does not generally make sense, but macro invocations may appear within macro definitions. Unless it has been shadowed by a macro argument, the special name CALLS
will be substituted within a macro with a number corresponding to how many times this macro has been expanded, counting from 0.
Sometimes there is an arithmetic relationship between constants in your program. Rather than computing them by hand, the :calc
command allows you to perform calculations at compile time. It takes a name, followed by a {
, a sequence of numbers, constant references, binary operators, unary operators or parentheses, and finally a terminal }
. The name is assigned to the result of evaluating the expression within curly braces. The following operators are available:
unary: - ~ ! sin cos tan exp log abs sqrt sign ceil floor @ strlen
binary: - + * / % & | ^ << >> pow min max < <= == != >= >
The unary operator @
looks up an address in the compiled ROM at the time of evaluation. Logical operators return 0
or 1
to indicate false or true, respectively. Additionally, the mathematical constants E
and PI
are usable, and the constant HERE
indicates the address immediately following the end of the compiled ROM at the time of evaluation. The strlen
operator is special; it expects to be immediately followed by a string literal (or a symbol which expands into one) and expands into the length of said literal in characters.
Note that as with all Octo commands, the tokens of a :calc
expression must be separated by whitespace. Bitwise operations are performed as if arguments were 32-bit signed integers, and otherwise they are treated as floating-point. When referenced, calculated constants are truncated to integegral values as appropriate. Order of evaluation is strictly right-to-left unless overridden by parentheses. The following expressions are equivalent:
:calc foo { 2 * 3 + baz }
:calc foo { 2 * ( 3 + baz ) }
When using :calc
and :macro
together, it is often useful to write the contents of some constant to the ROM; this can be done with :byte
:
:macro with-complement X {
:calc Y { 0xFF & ~ X }
:byte X
:byte Y
}
XO-Chip programming often requires working with 16-bit pointers. If forward-referencing of labels is required, use :pointer
. This directive assembles as two sequential bytes:
: label-a
...
:pointer label-a
:pointer label-b
...
: label-b
For convenience and brevity, if :byte
is immediately followed by {
the expression is computed as with :calc
and compiled as a byte, without defining an intermediate constant. The :org
, :pointer
, and :call
directives can also accept a constant expression, truncated to a 16-bit or 12-bit address, as appropriate.
Many Chip-8 programs do not require strings or text at all. When text is desired, representing it using ASCII is often very inconvenient. For example, fonts may not include all displayable ASCII characters, or a text rendering routine might want to work with glyph offsets. For this reason, Octo has an unusually flexible approach to working with string literals.
A string literal is enclosed in double-quotes ("
), and may contain C-style backslash (\
) escape sequences. The escape characters tnrv0\"
are supported. Apart from their ability to contain whitespace and some special characters, string literals are totally interchangeable with ordinary tokens.
The :stringmode
directive is used to define an approach for converting a string literal (or any other non-numeric token) into data in your program. The directive takes a name for the mode, followed by an alphabet, and then a macro body enclosed in curly braces ({ ... }
). The name of the mode will then behave like a macro which consumes one token (a data string) and invokes the mode's macro body for each character in the string. The macro body is provided with three bound arguments for each character: CHAR
, the ASCII value of the character, INDEX
, the index of the character in the data string, and VALUE
, the index of the character in the mode's alphabet. Multiple :stringmode
directives may be issued for a given mode name, to provide a different macro body for different ranges of characters. The alphabets of these definitions should not overlap.
Here is a simple example which generates premultiplied offsets to a font where each character is 8 pixels tall:
:stringmode text8 "ABCDEFGHIJKLMNOPQRSTUVWXYZ !" {
:byte { 8 * VALUE }
}
: message
text8 "GAME OVER!"
This assembles as 10 bytes equivalent to:
0x30 0x00 0x60 0x20 0xD0 0x70 0xA8 0x20 0x88 0xD8
SuperChip or SCHIP is a set of extended Chip-8 instructions. Octo can emulate these instructions and will indicate if any such instructions are used in an assembled program. The SuperChip instructions are as follows:
hires
Switch to a 128x64 pixel high resolution display mode.lores
Switch to the normal 64x32 pixel low resolution display mode.scroll-down n
Scroll the contents of the display down by 0-15 pixels.scroll-left
Scroll the contents of the display left by 4 pixels.scroll-right
Scroll the contents of the display right by 4 pixels.i := bighex vx
Set I to a large 8x10 hex char corresponding to register value.exit
Halt the program and quit the emulator.saveflags vx
Save v0-vn to flag registers. (Originally, HP-48 flag registers.)loadflags vx
Restore v0-vn from flag registers.
Flag registers are persisted using browser local storage, so provided no applications blow them away intentionally they can be used to store information between play sessions such as high score information or progress.
Finally, drawing a sprite with height 0 (which would otherwise do nothing) is used by the SuperChip to draw a large 16x16 sprite. The sprite data itself is stored as 16 pairs of bytes representing each row.
Beyond SuperChip, Octo provides a set of unique extended instructions called XO-Chip. These instructions provide a 4-color display, improved scrolling functionality, a flexible audio generator, expanded ram and instructions which make memory manipulation more convenient.
save vx - vy
save an inclusive range of registers to memory starting ati
.load vx - vy
load an inclusive range of registers from memory starting ati
.i := long NNNN
loadi
with a 16-bit address. (this instruction is 4 bytes long.)plane n
select zero or more drawing planes by bitmask (0 <= n <= 3).audio
store 16 bytes starting ati
in the audio pattern buffer.pitch := vx
set the audio pattern playback rate to4000*2^((vx-64)/48)
Hz. Initialized at 4000Hz.scroll-up n
scroll the contents of the display up by 0-15 pixels.
For more details, consult the XO-Chip specification in Octo's documentation directory. Octo is the main Chip-8 interpreter which supports these instructions, but a small but growing collection of third-party interpreters support them as well. Why not add them to yours?
Octo provides basic debugging facilities for Chip-8 programs. While a program is running, pressing the "i" key will interrupt execution and display the contents of the v
registers, i
and the program counter. Any register aliases and (guessed) labels will be indicated next to the raw register contents. You can click on registers in this view to cycle through displaying their contents in binary, decimal, or hexadecimal.
When interrupted, pressing "i" again or clicking the "continue" icon will resume execution, while pressing "o" will single-step through the program. The "u" key will attempt to step out (execute until the current subroutine returns) and the "l" key will attempt to step over (execute the contents of any subroutines until they return to the current level).
Pressing the "p" key will interrupt execution and display a profiler, indicating a best guess at the time spent in subroutines within your program so far. The profiler shows the top 20 results in a table, and you can also copy and paste a more detailed dump of profiling information for further analysis offline.
Breakpoints can also be placed in source code by using the command :breakpoint
followed by a name- the name will be shown when the breakpoint is encountered so that multiple breakpoints can be readily distinguished. :breakpoint
is an out-of-band debugging facility and inserting a breakpoint into your program will not add any code or modify any Chip-8 registers:
:breakpoint example-breakpoint
The command :monitor
is followed by a base address or register and then a length (in bytes) or format string. Format strings contain one or more formatting directive, which begins with a percent sign (%
), followed by an optional length in bytes (if omitted, 1 byte), and finally a character indicating how the source bytes should be interpreted: b
for binary digits, i
for an integer, x
for zero-padded hexadecimal digits, or c
for a 7-bit ASCII character. Newline characters (\n
) will appear as line breaks in the display. A few examples:
:monitor foobar 10 # 10 bytes of RAM, starting at a label "foobar"
:monitor v3 4 # registers v3-v6, shown as a block of hex digits
:monitor quux "x:%2i y:%2i z:%2i" # a 3-component vector, starting at label "quux", where each digit is 2 bytes
Pressing "m" will toggle the memory monitor on and off. While your program runs, monitors will be updated continuously to reflect the contents of memory. Like :breakpoint
, :monitor
is out-of-band and generates no instructions.
The command :assert
is followed by an optional message and a mandatory constant expression enclosed in curly braces (as with :calc
). If the expression evaluates to zero during compilation, compilation will halt and display the message. This can provide useful sanity-checks while modifying your programs, and is especially helpful when writing macros. For example:
:macro rol REG {
# note: v-register tokens are equivalent to their positions
# 0-15 when evaluating a constant expression:
:assert "cannot rotate-left through carry register" { REG != vF }
REG <<= REG
REG |= vF
}
: main
va := 0b10110111
rol va # modify this to use vF, and the macro will fail.
Assertions are strictly a compile-time facility and have no effect during the execution of a program.