Vous pouvez compiler du code source intelligent dans des applications DOS .COM assez facilement (vous aurez également besoin de NASM ou de code supplémentaire pour émettre des codes d'instruction et calculer des sauts). Ci-dessous un interprète bf légèrement modifié, transformé en un compilateur de toutes sortes:
// file: bfcompil.c
#include <stddef.h>
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#define MAX_CODE_SIZE 30000
char code[MAX_CODE_SIZE];
char* pc = &code[0];
char* pcEnd = &code[0];
#define MAX_DATA_SIZE 30000
char data[MAX_DATA_SIZE] = { 0 };
char* pd = &data[0];
// Structures for quick bracket matching
unsigned brStack[MAX_CODE_SIZE];
unsigned brSptr = 0;
unsigned brMatch[MAX_CODE_SIZE];
int main(int argc, char** argv)
{
FILE* f = NULL;
int ch;
if (argc != 2)
{
fprintf(stderr, "usage:\n bfcompil <brainfuck-source-code-file>\n"
"bfcompil will output NASM-compilable source code for"
"a DOS program\n");
return EXIT_FAILURE;
}
if ((f = fopen(argv[1], "rb")) == NULL)
{
fprintf(stderr, "can't open file \"%s\" for reading\n", argv[1]);
return EXIT_FAILURE;
}
while ((ch = getc(f)) != EOF)
{
if (strchr(" \t\r\n", ch) != NULL) // skip white space
{
continue;
}
else if (strchr("><+-.,[]", ch) != NULL) // store valid commands
{
if (pcEnd >= &code[sizeof(code)])
{
fprintf(stderr, "too many commands in file \"%s\", expected at most "
"%u commands\n", argv[1], (unsigned)sizeof(code));
fclose(f);
return EXIT_FAILURE;
}
if (ch == '[')
{
brStack[brSptr++] = (unsigned)(pcEnd - &code[0]);
}
else if (ch == ']')
{
if (brSptr == 0)
{
fprintf(stderr, "unmatched ']' in file \"%s\"\n", argv[1]);
fclose(f);
return EXIT_FAILURE;
}
brSptr--;
brMatch[brStack[brSptr]] = (unsigned)(pcEnd - &code[0]);
brMatch[pcEnd - &code[0]] = brStack[brSptr];
}
*pcEnd++ = ch;
}
else // fail on invalid commands
{
fprintf(stderr, "unexpected character '%c' in file \"%s\", valid command "
"set is: \"><+-.,[]\"\n", ch, argv[1]);
fclose(f);
return EXIT_FAILURE;
}
}
fclose(f);
if (brSptr != 0)
{
fprintf(stderr, "unmatched '[' in file \"%s\"\n", argv[1]);
return EXIT_FAILURE;
}
if (pcEnd == &code[0])
{
fprintf(stderr, "no commands found in file \"%s\"\n", argv[1]);
return EXIT_FAILURE;
}
printf("; how to compile: nasm -f bin <input file with this code.asm> -o "
"<output executable.com>\n\n"
"org 0x100\n"
"bits 16\n\n"
" mov bx, data\n"
" mov di, bx\n"
" mov cx, 30000\n"
" xor al, al\n"
" cld\n"
" rep stosb\n\n"
" jmp code\n\n"
"print:\n"
" mov ah, 2\n"
" cmp byte [bx], 10\n"
" jne lprint1\n"
" mov dl, 13\n"
" int 0x21\n"
"lprint1:\n"
" mov dl, [bx]\n"
" int 0x21\n"
" ret\n\n"
#if 01
// buffered input
"input:\n"
" cmp byte [kbdbuf+1], 0\n"
" jne linput1\n"
" mov ah, 0xa\n"
" mov dx, kbdbuf\n"
" int 0x21\n"
" inc byte [kbdbuf+1]\n"
"linput1:\n"
" mov al, [kbdbuf+2]\n"
" cmp al, 13\n"
" jne linput4\n"
" mov al, 10\n"
"linput4:\n"
" mov [bx], al\n"
" mov si, kbdbuf+3\n"
" mov di, kbdbuf+2\n"
" xor cx, cx\n"
" dec byte [kbdbuf+1]\n"
" mov cl, [kbdbuf+1]\n"
" jz linput3\n"
"linput2:\n"
" lodsb\n"
" stosb\n"
" loop linput2\n"
"linput3:\n"
" ret\n\n"
#else
// unbuffered input
"input:\n"
" mov ah, 1\n"
" int 0x21\n"
" cmp al, 13\n"
" jne linput\n"
" mov al, 10\n"
"linput:\n"
" mov [bx], al\n"
" ret\n\n"
#endif
"code:\n\n");
for (pc = &code[0]; pc < pcEnd; pc++)
{
switch (*pc)
{
case '>':
printf(" inc bx\n");
break;
case '<':
printf(" dec bx\n");
break;
case '+':
printf(" inc byte [bx]\n");
break;
case '-':
printf(" dec byte [bx]\n");
break;
case '.':
printf(" call print\n");
break;
case ',':
printf(" call input\n");
break;
case '[':
printf("label%u:\n", (unsigned)(pc - &code[0]));
printf(" cmp byte [bx], 0\n");
printf(" je label%u\n", (unsigned)brMatch[pc - &code[0]]);
break;
case ']':
printf(" jmp label%u\n", brMatch[pc - &code[0]]);
printf("label%u:\n", (unsigned)(pc - &code[0]));
break;
}
}
printf("\n ret\n\n");
printf("kbdbuf:\n"
" db 254\n"
" db 0\n"
" times 256 db 0\n\n");
printf("data:\n");
return EXIT_SUCCESS;
}
Si vous alimentez le programme Bonjour tout le monde:
++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>.
Il va produire un code d'assemblage compilable:
; how to compile: nasm -f bin <input file with this code.asm> -o <output executable.com>
org 0x100
bits 16
mov bx, data
mov di, bx
mov cx, 30000
xor al, al
cld
rep stosb
jmp code
print:
mov ah, 2
cmp byte [bx], 10
jne lprint1
mov dl, 13
int 0x21
lprint1:
mov dl, [bx]
int 0x21
ret
input:
cmp byte [kbdbuf+1], 0
jne linput1
mov ah, 0xa
mov dx, kbdbuf
int 0x21
inc byte [kbdbuf+1]
linput1:
mov al, [kbdbuf+2]
cmp al, 13
jne linput4
mov al, 10
linput4:
mov [bx], al
mov si, kbdbuf+3
mov di, kbdbuf+2
xor cx, cx
dec byte [kbdbuf+1]
mov cl, [kbdbuf+1]
jz linput3
linput2:
lodsb
stosb
loop linput2
linput3:
ret
code:
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
label10:
cmp byte [bx], 0
je label41
inc bx
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc bx
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc bx
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc bx
inc byte [bx]
dec bx
dec bx
dec bx
dec bx
dec byte [bx]
jmp label10
label41:
inc bx
inc byte [bx]
inc byte [bx]
call print
inc bx
inc byte [bx]
call print
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
call print
call print
inc byte [bx]
inc byte [bx]
inc byte [bx]
call print
inc bx
inc byte [bx]
inc byte [bx]
call print
dec bx
dec bx
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
inc byte [bx]
call print
inc bx
call print
inc byte [bx]
inc byte [bx]
inc byte [bx]
call print
dec byte [bx]
dec byte [bx]
dec byte [bx]
dec byte [bx]
dec byte [bx]
dec byte [bx]
call print
dec byte [bx]
dec byte [bx]
dec byte [bx]
dec byte [bx]
dec byte [bx]
dec byte [bx]
dec byte [bx]
dec byte [bx]
call print
inc bx
inc byte [bx]
call print
inc bx
call print
ret
kbdbuf:
db 254
db 0
times 256 db 0
data:
Si vous le compilez, vous pourrez l'exécuter sous DOS, Windows 9x/XP (probablement Vista-7 32 bits) et dans DosBox.
La sortie, sans surprise, est la suivante:
Hello World!
UPDATE: L'entrée DOS et des routines de sortie dans le code ci-dessus peuvent être remplacés par des accès directs à la mémoire tampon d'écran et les ports de clavier. Le code du clavier devra également gérer les interruptions du clavier. Ce n'est pas très difficile à faire sur le PC x86. Vous pouvez vraiment implémenter un compilateur pour qu'un langage s'exécute sur du matériel nu sans OS.
Vous devriez également jeter un oeil à Forth
car c'est exactement le genre de langage pour l'environnement donné. Et c'est facile à mettre en œuvre. Beaucoup plus facile que C. Harder que brainfuck, un peu comparable à l'assemblage.
MISE À JOUR 2: Voici un petit (~ 1 Ko en taille) interprète brainfuck qui est de ne pas utiliser toutes les fonctionnalités DOS ou BIOS:
; file: bfint.asm
; compile: nasm.exe -f bin bfint.asm -o bfint.com
; run in: DOS, DosBox or equivalent
bits 16
org 0x100
section .text
SCREEN_WIDTH equ 80
SCREEN_HEIGHT equ 25
SCAN_BUF_SIZE equ 256
MAX_CODE_SIZE equ 20000
MAX_DATA_SIZE equ 30000
cld
; set new keyboard (IRQ1) ISR
push byte 0
pop es
cli ; update ISR address w/ ints disabled
mov word [es:9*4], Irq1Isr
mov [es:9*4+2], cs
sti
push cs
pop es
Restart:
call ClearScreen
mov si, MsgHello
call PrintStr
mov word [CodeSize], 0
mov byte [EnterCount], 0
WaitForKey:
call GetKey
; Escape erases code
cmp ah, 1 ; Escape
je Restart
; Non-characters are ignored
cmp al, 0 ; non-character key
je WaitForKey
; Enter is "printed" but not stored, use for formatting
cmp al, 10 ; Enter
je KeyEnter
mov byte [EnterCount], 0
; Backspace deletes last character
cmp al, 8 ; Backspace
je KeyBackspace
; Space is printed but not stored, use for formatting
cmp al, " " ; Space
je PrintOnly
; 0 runs a test program
cmp al, "0"
je TestProgram
; Other chracters are stored as code
mov bx, [CodeSize]
cmp bx, MAX_CODE_SIZE
jae ErrCodeTooBig
mov [Code + bx], al
inc word [CodeSize]
PrintOnly:
call PrintChar
jmp WaitForKey
ErrCodeTooBig:
mov si, MsgCodeTooBig
call PrintStr
mov word [CodeSize], 0
jmp WaitForKey
KeyEnter:
call PrintChar
inc byte [EnterCount]
cmp byte [EnterCount], 1
je WaitForKey
mov byte [EnterCount], 0
call Execute
jmp WaitForKey
KeyBackspace:
call PrintChar
cmp word [CodeSize], 0
je WaitForKey
dec word [CodeSize]
jmp WaitForKey
TestProgram:
mov si, TestCode
mov di, Code
mov cx, TestCodeEnd - TestCode
mov [CodeSize], cx
rep movsb
call Execute
jmp WaitForKey
Execute:
mov si, Code ; code start
xor bp, bp ; instruction index
mov di, Data ; data start
mov cx, MAX_DATA_SIZE
xor al, al
rep stosb
sub di, MAX_DATA_SIZE
xor bx, bx ; data index
ExecuteLoop:
cmp bp, [CodeSize]
jae ExecuteDone
mov al, [bp+si]
cmp al, ">"
je IncPtr
cmp al, "<"
je DecPtr
cmp al, "+"
je IncData
cmp al, "-"
je DecData
cmp al, "."
je PrintData
cmp al, ","
je InputData
cmp al, "["
je While
cmp al, "]"
je EndWhile
mov si, MsgInvalidChar
call PrintStr
call PrintChar
mov al, 10
call PrintChar
jmp ExecuteDone
IncPtr:
inc bx
jmp ExecuteContinue
DecPtr:
dec bx
jmp ExecuteContinue
IncData:
inc byte [bx+di]
jmp ExecuteContinue
DecData:
dec byte [bx+di]
jmp ExecuteContinue
PrintData:
mov al, [bx+di]
call PrintChar
jmp ExecuteContinue
InputData:
call GetKey
or al, al
jz InputData
mov [bx+di], al
jmp ExecuteContinue
While:
cmp byte [bx+di], 0
jne ExecuteContinue
mov ax, 1
mov dx, "[]"
call FindMatchingBracket
ExecuteContinue:
inc bp
jmp ExecuteLoop
EndWhile:
mov ax, -1
mov dx, "]["
call FindMatchingBracket
jmp ExecuteLoop
ExecuteDone:
mov word [CodeSize], 0
mov si, MsgCompleted
jmp PrintStr
FindMatchingBracket:
xor cx, cx
FindMatchingBracket1:
cmp byte [bp+si], dl
jne FindMatchingBracket2
inc cx
jmp FindMatchingBracket3
FindMatchingBracket2:
cmp byte [bp+si], dh
jne FindMatchingBracket3
dec cx
jnz FindMatchingBracket3
ret
FindMatchingBracket3:
add bp, ax
jmp FindMatchingBracket1
; Inputs:
; AL = ASCII character code
PrintChar:
; assuming it's a color text mode (not monochrome or graphics)
pusha
push es
push word 0xb800
pop es
mov bx, [CursorPos]
cmp al, 8
je PrintCharBackSpace
cmp al, 10
je PrintCharBackLF
cmp al, 13
je PrintCharBackCR
mov [es:bx], al
call AdvanceCursorPosition
jmp PrintCharDone
PrintCharBackSpace:
; move the cursor back and erase the last character
or bx, bx
jz PrintCharDone
dec bx
dec bx
mov word [es:bx], 0x0720
jmp PrintCharSetCursorPos
PrintCharBackLF:
; move the cursor to the beginning of the next line - '\n' behavior
add bx, SCREEN_WIDTH * 2
cmp bx, SCREEN_WIDTH * SCREEN_HEIGHT * 2
jc PrintCharBackCR
sub bx, SCREEN_WIDTH * 2
call ScrollUp
PrintCharBackCR:
; move the cursor to the beginning of the current line - '\r' behavior
mov ax, SCREEN_WIDTH * 2
xchg ax, bx
xor dx, dx
div bx
mul bx
mov bx, ax
PrintCharSetCursorPos:
mov [CursorPos], bx
shr bx, 1
call SetCursorPosition
PrintCharDone:
PopEsAllRet:
pop es
popa
ret
ClearScreen:
; assuming it's a color text mode (not monochrome or graphics)
pusha
push es
push word 0xb800
pop es
xor di, di
mov cx, SCREEN_WIDTH * SCREEN_HEIGHT
mov ax, 0x0720 ; character = space, color = lightgray on black
rep stosw
xor bx, bx
mov [CursorPos], bx
call SetCursorPosition
jmp PopEsAllRet
ScrollUp:
; assuming it's a color text mode (not monochrome or graphics)
pusha
push es
push ds
push word 0xb800
pop es
push es
pop ds
mov si, SCREEN_WIDTH * 2
xor di, di
mov cx, SCREEN_WIDTH * (SCREEN_HEIGHT - 1)
rep movsw
mov cx, SCREEN_WIDTH
mov ax, 0x0720 ; character = space, color = lightgray on black
rep stosw
pop ds
jmp PopEsAllRet
; Inputs:
; DS:SI = address of NUL-terminated ASCII string
PrintStr:
pusha
PrintStr1:
lodsb
or al, al
jz PrintStrDone
call PrintChar
jmp PrintStr1
PrintStrDone:
popa
ret
; Inputs:
; BX = Y * SCREEN_WIDTH + X
SetCursorPosition:
; assuming it's a color text mode (not monochrome or graphics)
pusha
%if 0
mov dx, 0x3d4
mov al, 0x0f
out dx, al
inc dx
mov al, bl
out dx, al
dec dx
mov al, 0x0e
out dx, al
inc dx
mov al, bh
out dx, al
%else
mov dx, 0x3d4
mov al, 0x0f
mov ah, bl
out dx, ax
dec al
mov ah, bh
out dx, ax
%endif
popa
ret
AdvanceCursorPosition:
; assuming it's a color text mode (not monochrome or graphics)
pusha
mov ax, [CursorPos]
inc ax
inc ax
cmp ax, SCREEN_WIDTH * SCREEN_HEIGHT * 2
jc AdvanceCursorPosition1
sub ax, SCREEN_WIDTH * 2
call ScrollUp
AdvanceCursorPosition1:
mov [CursorPos], ax
shr ax, 1
xchg ax, bx
call SetCursorPosition
popa
ret
; Outputs:
; AH = scan code
; AL = character
GetKey:
push bx
push si
GetKeyRepeat:
mov ax, [ScanWriteIdx]
mov si, [ScanReadIdx]
sub ax, si
jz GetKeyRepeat
mov bx, si
mov ax, [ScanBuf + bx + si]
inc si
and si, SCAN_BUF_SIZE - 1
mov [ScanReadIdx], si
pop si
pop bx
ret
Irq1Isr:
pusha
push ds
push cs
pop ds
; read keyboard scan code
in al, 0x60
cmp al, 0x2a ; Left Shift down
jne Irq1Isr1
or byte [Shift], 1
Irq1Isr1:
cmp al, 0x36 ; Right Shift down
jne Irq1Isr2
or byte [Shift], 2
Irq1Isr2:
cmp al, 0xaa ; Left Shift up
jne Irq1Isr3
and byte [Shift], ~1
Irq1Isr3:
cmp al, 0xb6 ; Right Shift up
jne Irq1Isr4
and byte [Shift], ~2
Irq1Isr4:
test al, 0x80
jnz Irq1IsrEois ; key released
mov ah, al
cmp al, 58
jc Irq1Isr5
xor al, al ; don't translate non-character keys
jmp Irq1Isr7
Irq1Isr5:
mov bx, ScanToChar
cmp byte [Shift], 0
je Irq1Isr6
add bx, ScanToCharShift - ScanToChar
Irq1Isr6:
xlatb
Irq1Isr7:
mov bx, [ScanWriteIdx]
mov di, bx
mov [ScanBuf + bx + di], ax
inc bx
and bx, SCAN_BUF_SIZE - 1
mov [ScanWriteIdx], bx
Irq1IsrEois:
%if 0
; send EOI to XT keyboard
in al, 0x61
mov ah, al
or al, 0x80
out 0x61, al
mov al, ah
out 0x61, al
%endif
; send EOI to master PIC
mov al, 0x20
out 0x20, al
pop ds
popa
iret
ScanToChar:
db 0 ; unused
db 0 ; Escape
db "1234567890-="
db 8 ; Backspace
db 9 ; Tab
db "qwertyuiop[]"
db 10 ; Enter
db 0 ; Ctrl
db "asdfghjkl;'`"
db 0 ; Left Shift
db "\zxcvbnm,./"
db 0 ; Right Shift
db 0 ; Print Screen
db 0 ; Alt
db " " ; Space
ScanToCharShift:
db 0 ; unused
db 0 ; Escape
db "[email protected]#$%^&*()_+"
db 8 ; Backspace
db 9 ; Tab
db "QWERTYUIOP{}"
db 10 ; Enter
db 0 ; Ctrl
db 'ASDFGHJKL:"~'
db 0 ; Left Shift
db "|ZXCVBNM<>?"
db 0 ; Right Shift
db 0 ; Print Screen
db 0 ; Alt
db " " ; Space
MsgHello:
db "Brainfuck Interpreter", 10, 10
db "Press 0 to run test code OR", 10
db "Type your code.", 10
db "Use Esc to erase it all or Backspace to delete last character.", 10
db "Press Enter twice to run it.", 10, 10, 0
MsgCodeTooBig:
db 10, "Code's too big", 10, 0
MsgCompleted:
db 10, "Code's completed", 10, 0
MsgInvalidChar:
db 10, "Invalid character: ", 0
Shift db 0
CursorPos dw 0
ScanReadIdx dw 0
ScanWriteIdx dw 0
EnterCount db 0
CodeSize dw 0
TestCode:
; Hello World!
db "++++++++++[>+++++++>++++++++++>+++>+<<<<-]>++.>+.+++++++..+++.>++.<<+++++++++++++++.>.+++.------.--------.>+.>."
; Squares of 0 through 100
; db "++++[>+++++<-]>[<+++++>-]+<+[>[>+>+<<-]++>>[<<+>>-]>>>[-]++>[-]+>>>+[[-]++++++>>>]<<<[[<++++++++<++>>-]+<.<[>----<-]<]<<[>>>>>[>>>[-]+++++++++<[>-<-]+++++++++>[-[<->-]+[<<<]]<[>+<-]>]<<-]<<-]"
; ROT13
; db "+[,+[-[>+>+<<-]>[<+>-]+>>++++++++[<-------->-]<-[<[-]>>>+[<+<+>>-]<[>+<-]<[<++>>>+[<+<->>-]<[>+<-]]>[<]<]>>[-]<<<[[-]<[>>+>+<<<-]>>[<<+>>-]>>++++++++[<-------->-]<->>++++[<++++++++>-]<-<[>>>+<<[>+>[-]<<-]>[<+>-]>[<<<<<+>>>>++++[<++++++++>-]>-]<<-<-]>[<<<<[-]>>>>[<<<<->>>>-]]<<++++[<<++++++++>>-]<<-[>>+>+<<<-]>>[<<+>>-]+>>+++++[<----->-]<-[<[-]>>>+[<+<->>-]<[>+<-]<[<++>>>+[<+<+>>-]<[>+<-]]>[<]<]>>[-]<<<[[-]<<[>>+>+<<<-]>>[<<+>>-]+>------------[<[-]>>>+[<+<->>-]<[>+<-]<[<++>>>+[<+<+>>-]<[>+<-]]>[<]<]>>[-]<<<<<------------->>[[-]+++++[<<+++++>>-]<<+>>]<[>++++[<<++++++++>>-]<-]>]<[-]++++++++[<++++++++>-]<+>]<.[-]+>>+<]>[[-]<]<]"
TestCodeEnd:
section .bss
ScanBuf:
resw SCAN_BUF_SIZE
Code:
resb MAX_CODE_SIZE
Data:
resb MAX_DATA_SIZE
Si vous voulez prendre DOS (comme l'environnement d'hébergement) et MSNA Hors de l'image, vous êtes invités à encoder le code d'assemblage ci-dessus à la main, en faire une disquette de démarrage et le démarrer.
Un peu plus pratique serait d'implémenter Forth from scratch sur un métal nu. Et puis vous serez capable de l'utiliser pour démarrer un système plus complexe en plus. –
Ok ... ça ressemble à ce qui m'intéresse. Je suppose que par "métal nu" vous voulez dire uniquement du matériel? Y a-t-il un bon tutoriel pour le faire? Tout va bien ... comme je l'ai dit je ne suis intéressé qu'à des fins éducatives, donc si je comprends comment le faire avec Forth, je serai heureux. –
Oui, il y a beaucoup de tutoriels, comme celui-ci: http://lambda-the-ultimate.org/node/2452 ou ceci: http://hbrobotics.org/wiki/index.php?title=Forth_for_Robotics_and_Programming_at_the_Bare_Metal –