This project implements code translation to uneven bytecode, disassembly tool for generated bytecode, and bytecode executor (Soft Processor Unit).

In addition, I implemented code analyser that checks for syntax mistakes. Analyser outputs info in clang format, so most IDEs highlight wrong expressions and provide descriptions of errors as well as possible solutions when possible.

During the translation, .lst file is generated. It provides information about bytecode generation and what command at what offset is located.

Documentation

Enter project's directory and execute:

mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=Release ..
cmake --build .

Three files will be created:

• SPU - Soft Processor Unit, executes created programs
• SPUAsm - translates source file (.spus) to binary file (.spub) that can be executed by SPU
• SPUDisAsm - disassembly for generated binary files

SPUAsm

$ SPUAsm <source file>
        --input <filename.spus>     - source file (can be specified without --input)
        --output <filename.spub>    - binary assembled file
        -E                          - generate preprocessed file
        --verbose                   - output assembly debug information
        --lst <filename.lst>        - listing file of generated code (assembly.lst by default) 
        -h, --help                  - show help message

SPUDisAsm

$ SPUDisAsm <binary file>
        --input <filename.spub>     - binary assembled file
        --output <filename.spus>    - source file (if not specified, stdout selected)
        --verbose                   - output assembly debug information
        -h, --help                  - show help message

SPU

$ SPU <binary file>
        --input <filename.spub>     - binary assembled file
        --output <filename.spus>    - source file (if not specified, stdout selected)
        --verbose                   - output assembly debug information
        --vsync                     - render vram after every command
        -h, --help                  - show help message

All examples are available in this dir.

;
; Squares
;

in rbx  ; rbx as counter

push 0
pop rax

loop:
dec rbx  ; decrease counter

; Calculating square
push rax
push rax
mul
out

; Prepare for the next loop
pop
inc rax

jm mondayOnly ; exit if it is monday

; Insert conditional jump operands
push 0
push rbx

jne loop

mondayOnly:

; height 32
; width  64

push 32
push 16
push 10
push 35
call drawCircle

clrscr

push 32
push 16
push 4
push 42
call drawCircle

clrscr

push 16
push 8
push 2
push 42
call drawCircle

clrscr
rend

push 40
push 25
push 4
push 48
call drawCircle

clrscr
rend

push 60
push 13
push 8
push 48
call drawCircle


hlt

drawCircle: ;(x0, y0, r, symbol)
    pop [2]
    pop rcx  ; r
    pop rbx  ; y0
    pop rax  ; x0

    mov rdx 0   ; initial angle
    loop:
        push rbx
        push rdx
        sin
        push rcx
        mul
        add ; y0 + r*sin(alpha)

        pop [0] ; y coordinate

        push rax
        push rdx
        cos
        push rcx
        push 2
        mul
        mul
        add ; x0 + r*cos(alpha)

        pop [1] ; x coordinate

        pixset [1] [0] [2]
        
        rend
        slp 10000
    ; loop params
    mov rdx rdx+0.1
    push rdx
    push 3.1415926535897
    push 2
    mul
    jbe loop
    ret

;
; Data input
; (a, b, c)
;
in      rax
in      rbx
in      rcx

;
; D calculation
;
push    rbx
push    2
pow

push    4
push    rax
mul
push    rcx
mul
sub
sqrt
pop     rdx


;
; First x
;
clear
push    -1
push    rbx
mul
push    rdx
sub
push    2
div
push    rax
div
pop     rcx

;
; Second x
;
clear
push    -1
push    rbx
mul
push    rdx
add
push    2
div
push    rax
div
pop     rax

;
; Output
;
out     rax
out     rcx

Binary generation can be splitted into several parts:

• Cleanup
• Analysis
• Final cleanup
• Translation
• Final checks

If process fails on any part, the next parts are not executed.

Just double whitespaces, trailing/leading whitespaces removed. Operations that does not change code structure, lines number etc.

Analyses code for syntax errors, builds temporary code to calculate labels offsets. Prints errors descriptions in clang format.

All unnecessary indent symbols removed (comments, blank lines, unneded whitespaces). Code structure will be altered, line numbers are not preserved.

Code is translated using labels table generated on analysis step. At this point, all errors must be catched by analysis, but still some can make thir way to this point. In this case, generation will fail, dumping wrong instruction and its number.

Check that all operations completed correctly; labels table is complete and not redundant.

Operators with arguments accept complex value argument. Argument can be assignable or not assignable

Complex value examples:

• [rax+5] - assignable, RAM on adress rax+5 with double adressation
• 4 - not assignable, immediate value 4
• [5] - assignable, RAM on adress 5 with double adressation
• [rcx] - assignable, RAM on adress rcx with double adressation
• rcx - assignable, register rcx
• (rcx) - assignable, RAM on adress rcx with char adressation
• (rcx+10) - assignable, RAM on adress rcx+10 with char adressation
• (10) - assignable, RAM on adress 10 with char adressation

Notice: whitespaces inside complex value are not allowed and considered as syntax error.

Push value to the stack

Pop value from the stack

• no args - just delete
• cvalue - pop and save to assignable cvalue

Request value from the console

• no args - push value to the stack
• cvalue - save to assignable cvalue

Prints value to the console

• no args - last stack value
• cvalue - cvalue value

destination -> source

Sets the first argument value to the secon argument's value

Dump stack information

Clear stack

Force render vram

Sleep for ... nanoseconds

Finish the program

x y value

Set vram pixel (x, y) to value

Fill vram with spaces

Increments the value

• no args - last stack value
• cvalue - cvalue value

Decrements the value

• no args - last stack value
• cvalue - cvalue value

Add two last stack values and push to the stack

Substract two last stack values and push to the stack

Multiply two last stack values and push to the stack

Divide two last stack values and push to the stack

Sin of the last stack value

Cos of the last stack value

Abs of the last stack value

Square root of the last stack value

The pre-last element of the stack to the power of the last one

Gives execution to the label block. Can be returned to the call instruction using ret

Returnes to the last call position. May fail if there were no calls.

Jump to the label

Jump if last element is bigger than pre-last. Removes both operands from the stack

Jump if last element is bigger-or-equal than pre-last. Removes both operands from the stack

Jump if last element is equal to the pre-last one. Removes both operands from the stack

Jump if last element is not equal to the pre-last one. Removes both operands from the stack

Jump if last element is lower than pre-last. Removes both operands from the stack

Jump if last element is lower-or-equal than pre-last. Removes both operands from the stack

Jump if it's Monday.

Creates label

• SPUAsm code
• SPUDisAsm code
• SPU code
• Examples