#+TITLE Project Specification

* Binary interface

The VM does not use floating point numbers, it instead uses fixed point numbers.

This is for portability reasons as some devices might not have a FPU in them

especially microcontrollers and some retro game systems like the PS1.

** Numbers
| type | size (bytes) | description                           |
|------+--------------+---------------------------------------|
| u8   |            1 | unsigned 8bit, alias =char= and =byte=    |
| bool |            1 | unsigned 8bit, =false= or =true=          |
| i8   |            1 | signed 8bit for interop               |
| u16  |            2 | unsigned 16bit for interop            |
| i16  |            2 | signed 16bit for interop              |
| u32  |            4 | unsigned 32bit, alias =nat=             |
| i32  |            4 | signed 32bit, alias =int=               |
| f32  |            4 | signed 32bit fixed number, alias =real= |

* Memory

Uses a harvard style archecture, meaning the code and ram memory
are split up into two seperate blocks.

In the C version you can see these are two seperate arrays 'code' and 'mem'.

During compilation constants and local variables are put onto 'mem'

* Opcodes

| 2^n | count |
|----+-------|
| 2^3 |     8 |
| 2^4 |    16 |
| 2^5 |    32 |
| 2^6 |    64 |
| 2^8 |   128 |

** could be encoded
- op type [this is to maximize jump immidate and load immidate size]
- memory location
- local value / register
- local value type

*** Simplest
[opcode][dest][src1][src2]
[8][8][8][8]

*** Maximize inline jump and load-immidate
[0 0 0 0 0 0 0 0][0 0 0 0 0 0 0 0][0 0 0 0 0 0 0 0][0 0 0 0 0 0 0 0]

[0 0 0 | 0 0 0 0 0][0 0 0 | 0 0 0 0 0][0 0 0 | 0 0 0 0 0][0 0 0 | 0 0 0 | 0 0 ] noop
[0 0 0 | 0 0 0 0 0][0 0 0 | 0 0 0 0 0][0 0 0 | 0 0 0 0 0][0 0 0 | 0 0 1 | 0 0 ] call
[0 0 0 | 0 0 0 0 0][0 0 0 | 0 0 0 0 0][0 0 0 | 0 0 0 0 0][0 0 0 | 0 1 0 | 0 0 ] return
[0 0 0 | 0 0 0 0 0][0 0 0 | 0 0 0 0 0][0 0 0 | 0 0 0 0 0][0 0 0 | 0 1 1 | 0 0 ] syscall? 
[0 0 0 | 0 0 0 0 0][0 0 0 | 0 0 0 0 0][0 0 0 | 0 0 0 0 1][0 0 0 | 1 0 0 | 0 0 ] exit

[0 0 0 0 0 0 0 0][0 0 0 0 0 0 0 0][0 0 0 0 0 0 0 0][0 0 0 0 0 0 | 0 1 ] jump ~1GB range
[0 0 0 0 0 0 0 0][0 0 0 0 0 0 0 0][0 0 0 0 0 0 0 0][0 0 0 0 0 0 | 1 0 ] load-immidate 2^30 max 

*** multibyte ops
- ones that would be easier if they were multibyte
  - jump
  - load immidate
  - syscall
  - call


0 0 - system, lowest because lower opcodes are faster
0 1 - memory 
1 0 - math
1 1 - jump

[0 0][0 0 0 0 0 0] = [system][no op]
[0 0][1 0 0 0 0 0] = [system][loadimm]
[0 0][0 0 0 0 0 0] = [system][return]


[0 0 0 | r r r r r][0 0 0 | r r r r r][0 0 0 | r r r r r][b b | t | o o o | 1 1] [math][add][f][8]
[0 0 0 | 0 0 0 0 0][0 0 0 | 0 0 0 0 0][0 0 0 | 0 0 0 0 0][0 0 | 1 | 0 0 1 | 1 1] [math][add][f][8]

[0 1][0 0 1][0][0 0] = [math][sub][f][16]
[0 1][0 1 0][0][0 0] = [math][mul][f][32]
[0 1][0 1 1][0][0 0] = [math][div][f][?]
[0 1][1 0 0][0][0 0] = [math][and][f][?]

[1 1][0][0 0][0 0] = [jmp][u][eq]
[1 1][0][0 0][0 0] = [jmp][u][ne]
[1 1][0][0 0][0 0] = [jmp][u][lt]
[1 1][0][0 0][0 0] = [jmp][u][gt]

[1 1][1][0 0 0][0 0] = [jmp][s][le]
[1 1][1][0 0 0][0 0] = [jmp][s][ge]
[1 1][1][0 0 0][0 0] = [jmp][s][]
[1 1][1][0 0 0][0 0] = [jmp][s][]




3     2 2  1
[...] i 32 0

3     1 3  1
[...] u rl j
[...] u ab j
[...] u eq j
[...] u nq j
[...] u lt j
[...] u le j
[...] u gt j
[...] u ge j

[jmp][dest][18]
[lli][dest][2]



int 3

add 
sub
mul
div

jeq
jne

jlt
jle
jgt
jge