Files
wassembly/README.md

173 lines
6.1 KiB
Markdown

# About
This is an exploratory project into virtual machines and assembly language. By
no means is this ready for production use or particularly well maintained. The
language is inspired by x86 and ARM assembly and does very little hand holding.
Checkout the `bin/example.wasm` example source to get an overview of the
language, or keep on reading!
# Design
## From Text To Runtime Behaviour
In order to turn the source text into executable code we use 3 passes:
- Pass 1: tokenization (syntax check) and preprocessing (substitution)
- Pass 2: interpretation (semantics check)
- Pass 3: execution (runtime check)
After pass 2 ties to the source code are lost, meaning that any error occurring
afterwards can be a bit cryptic as to where it originated.
## Notation
- `[operation][number type]`, e.g. `divi` for divide (div) integer
- `%[register]` for addressing registers, e.g. `%A`
- `$[value]` for using immediate (literal) integer values, e.g. `$38`
- `'[character]'` for using immediate character values, e.g. `'r'`
- `;` for end of statement (mandatory)
- `[label]:` for labels, e.g. `loop:`
- `#[text]` for comments: any text is ignored till a newline (`\n`) is found
- `[[%register|$value]]` for accessing memory, e.g. `[$104]`
- Elements must be separated by whitespace character(s)
- Good: `add $2 $5 %A;`
- Bad: `add $2$5%A;`
## Examples
Divide register A by 5 and store the result in register A:
`divi %A $5 %A;`
Increment B until it is 10:
```
# Set B to zero
addi $0 $0 %B;
loop:
addi $1 %B %B;
lti %B $10;
jmp loop;
```
Read the integer at memory location `1024` into register A:
```
seti %A [$1024];
```
Remember not to use spaces inside the `[` brackets.
## Reserved Symbols
The following whitespace characters are used to separate symbols:
- space (` `)
- tab (`\t`)
- return carriage (`\r`)
- newline (`\n`)
The following characters are used as identifiers:
- dollar (`$`) for immediate (literal) integer values
- single quote (`'`) for immediate character values
- percentage (`%`) for register identifiers
- colon (`:`) for jump labels
- semicolon (`;`) for statement termination
- hash (`#`) for comments
- square brackets (`[` and `]`) for addressing memory
## Memory Model
The stack, with which you interact through pop/push operations, grows from
memory location 0 to the end of the memory. There is no strict checking on
whether your own memory operations through `[]` affect the stack: this is a
feature, not a bug. Keep in mind that the stack can underflow and overflow
and that the memory uses byte units (8 bits), whereas the registers are all
32 bits wide. This means that reading from location `$900` overlaps with 3
bytes when reading from location `$901` (the first byte of `$901` is the
second byte of location `$900`).
## Symbols
All symbols are reserved keywords and can therefore NOT be used as labels.
There is currently no strict checking, so be careful.
### Preprocessor
All preprocessor directives are prefixed by a `#`. Ill formed preprocessor
directives do not halt compilation, they are merely ignored. All preprocessing
is done in a single pass. Recursion or definition of a directive by another
directive is not supported therefore.
- `DEFINE <x> [y]` replaces any occurrence of the first argument (`x`) by the
second optional argument (`y`). The second argument can be empty, effectively
deleting all occurrences of `x`. Quotes are currently not supported and
arguments are separated by whitespace. If multiple defines exist the later
declarations will overwrite the previous.
### Registers
All registers are 32 bits wide. The following 4 general purpose registers
currently exist:
- A
- B
- C
- D
### Immediates
An immediate integer value for 42 is for examle `$42`. Negative values are
allowed, for example `$-42`. Notation must be in decimal, hexadecimal and octals
are **not supported**.
The immediate character value for the letter g is `'g'`. Character values must
be a single character, escaped or multi byte characters are **not supported**.
### Operands
- `addi` add the first to the second argument and store the result in the third
argument
- `subi` subtract the first from the second argument and store the result in
the third argument
- `divi` divide the first by the second argument and store the result in the
third argument
- `muli` multiply the first by the second argument and store the result in the
third argument
- `shli` shift left the first argument by the number of positions given by the
second argument and store the result in the third argument
- `shri` shift right the first argument by the number of positions given by the
second argument and store the result in the third argument
- `seti` set the first register argument to the second argument
- `int` calls the interrupt specified by the first (integer) argument
### Control Flow
- `jmp` jump to the label given by the first argument
- `call` put the next statement to execute on the stack and jump to the label
given by the first argument
- `ret` pop the the next statement to execute off the stack, e.g. returning to
the next execution statement before calling `call`
- `lti` execute next statement if argument 1 is less than argument 2 else skip
the next statement
- `gti` execute next statement if argument 1 is greater than argument 2 else
skip the next statement
- `eqi` execute the next statement if argument 1 is equal to argument 2 else
skip the next statement
## Memory
- `popi` pops the first value on the stack into the register specified as the
first argument
- `pushi` pushes the value on the stack from the register or immediate value as
the first argument
## Interupts
- [0..9] Output to STDOUT
- `0` put value of register A as ASCII character on stdout
- `1` put value of register A as decimal integer on stdout
- `2` put value of register A as hexadecimal integer on stdout
- `3` put the string pointed at by register A for the amount of characters
defined by register B on stdout
- [10..19] Input from STDIN
- `10` get a single ASCII character from STDIN and store it in register A
- `11` get a string of a maximum length determined by register B and store it
in the address specified by register A. After execution register B will
contain the number of characters actually read.