Skip to content

preprocessor.md

Latest commit

 

History

History
94 lines (71 loc) · 3.71 KB

preprocessor.md

File metadata and controls

94 lines (71 loc) · 3.71 KB

BComp Assembler Preprocessor

Usage

cd src # enter code directory root
python3 ../preprocessor/main.py main.basm build/main.build.basm -t build/abi.build.basm
bcomp -c -r -C build/main.build.basm
  • first argument - path to the root code file, all include directive should be relative to the directory of that file
  • second argument - output file, in which will all included files be built
  • -t argument - if present, table of global functions with all addresses substituted would be written to that file, it could be later inserted as header file to applications to provide them with OS ABI.

Constructions available

String encoding

To encode strings to memory, the following format could be used: LABEL: WORD "text here" By default, preprocessor uses KOI8-R encoding. Data layout is 1 byte per char, 2 chars per word, with 0 at the end, in format (char[1] char[0]) ... (char[i+1]@1b char[i]@1b) (0x00, 0x00/char[2n-1]). For convinience, pointer to the starting address is created automatically at the end of string memory with original label and postfix _PTR.

File includes

The preprocessor provides instruction to include other asm files by directly inserting included program text in place of INCLUDE instruction. Syntax: INCLUDE file_path_here

Other preprocessor modules are run after joining all included files. No relocation is done, therefore use of ORG directive of assembler and INCLUDE of preprocesseor are allowed only at root file.

Function support

Preprocessor provides number of helpers for writting functions, however it is important to follow the template for module to be run.

The preprocessor enforces TODO calling convention.

Local varible names

Preprocessor provides labeling for stack memory cells. All variables must be declared in ; stack segment of function header as ; &: label_name_here. In function body *(only between function label and ; end func FUNCNAME) all occurences of &label_name got replaced with &stack_number.

As BComp hasn't got RBP, addressing is from SP. The first declared variable gets &0 and lines after get incremented numbers. Warning: When it is inevitable to use PUSH in body, with each incstruction SP moves so generated stack pointer shifts are moved also and will not point to original variable. Use &label_name+pushes_before_count to shift this label.

Prologue and epilogue generation

To reduce code size, macros APUSH and APOP provided for prologue and epilogue respectively. Only local variables declared in format described in previous paragraph are used.

The following template is used:

Prologue:
SPADD #%(1-CNT)%

Epilogue:
SPADD #%CNT%

Where CNT is the local stack variables count. All values are 1-byte signed integers.

(For info: SPADD x is SP + x -> SP instruction of bcomp)

Function template

All three NAMESPACE__FUNCNAME names should be same as function label and be present, other way function is ignored by preprocessor

; func NAMESPACE__FUNCNAME: (args, ...) description here
; 
; params: AC: arg1        <- just documentation about parameters (no preprocessing)
;         STACK: arg2
;         ...
; return: AC: ...         <- also only docs
; stack:                  <- used for local variables preprocessor
; &: var_name1            <- would be &0
; &: var_name2
; &: var_name3
; &: ret_addr             <- having variable that has `ret` in name is signing that all following declarations ignored in prologue
; &: func_stack_arg1 
; &: ...
; 
; begin globals
FUNCNAME__VARNAME: WORD ....
; end globals

; begin func
NAMESPACE__FUNCNAME:

; begin prologue
  APUSH
; end prologue

; code here

; begin epilogue
  APOP  
  RET
; end epilogue
; end func NAMESPACE__FUNCNAME