CR-ASM by The Cronies
____
,----.. ,-.----. ,---, .--.--. ,' , `.
/ / \ \ / \ ' .' \ / / '. ,-+-,.' _ |
| : :; : \ ,---,. / ; '. | : /`. / ,-+-. ; , ||
. | ;. /| | .\ : ,' .' |: : \ ; | |--` ,--.'|' | ;|
. ; /--` . : |: | ,---.' ,: | /\ \| : ;_ | | ,', | ':
; | ; | | \ : | | || : ' ;. :\ \ `. | | / | | ||
| : | | : . / : : .' | | ;/ \ \`----. \' | : | : |,
. | '___ ; | | \ : |.' ' : | \ \ ,'__ \ \ |; . | ; |--'
' ; : .'|| | ;\ \`---' | | ' '--' / /`--' /| : | | ,
' | '/ :: ' | \.' | : : '--'. / | : ' |/
| : / : : :-' | | ,' `--'---' ; | |`-'
\ \ .' | |.' `--'' | ;/
`---` `---' '---'
Prod: CR-ASM
Size: 512b
Type: Demotool?
Platform: MS-DOS (386+)
Group: The Cronies
Date: July 6, 2015
Contact: [email protected]
-~=<|$| Introduction |$|>=~-----------------------------------------------------
This is a self-hosting assembler (i.e. it can assemble it's own source code)
in 512 bytes! It assembles a Turing-complete subset of the x86 instruction set.
It accepts the source file via STDIN and writes the machine code to STDOUT. Feel
free to give it a go:
cr-asm.com <cr-asm.asm >cr-asm2.com
Notice that cr-asm2.com is identical to cr-asm.com. You can do
cr-asm2.com <cr-asm.asm >cr-asm3.com
ad infinitum (if you're insane and you get your jollies from that sort of thing)
-~=<|$| Caveats |$|>=~----------------------------------------------------------
There are some caveats to the CR-ASM assembly language. Obviously, this was
never intended to be seriously used, but is rather a proof of concept. I'll
enumerate the short-comings up-front for the sake of fairness:
1. As mentioned earlier, it only implements a small subset of the x86
instruction set.
2. It is CaSe-SeNsItIvE. All alphabetic characters must be capitalized.
3. It does not tolerate any extraneous white-space (no blank lines, etc.)
4. There is no support for comments or labels :( This makes it about as fun
to code in as MS-DOS DEBUG.
5. There is no support for decimal values. All constants are assumed to be
hexadecimal.
6. The source file must end with two blank lines (this signals to the parser
to exit). I know. It's super lame.
7. If there is a syntax error in the source code, CR-ASM typically hangs.
Again, super lame, but what do you want out of a 512 byte assembler?
8. No support for UNIX style lines. <CR><LF> must immediately follow each
instruction (no white-space after the instruction).
9. Only registers AX, CX, DX, BX, SP, BP, SI, and DI are supported. No
support for segment registers.
10. Many of the mnemonics that CR-ASM uses are not the official mnemonics.
This was to make parsing the instructions easier.
11. Other snags that I'm not thinking of at the moment.
-~=<|$| Supported instructions |$|>=~-------------------------------------------
Each instruction must be entered EXACTLY as described here. No extra
white-space ANYWHERE! In particular, watch out for white space after the end of
the instruction.
MOV XX, YYYY
Move the hexadecimal value YYYY into the register XX. YYYY must be exactly
four characters long. Use leading zeros where needed. This is the only form
of the MOV instruction that is supported. If you need to move values between
registers, use pushes and pops.
INT YY
Call interrupt YY where YY is a hexadecimal value that is exactly two
characters long.
PSH XX
Push the register XX onto the stack. Notice the mnemonic deviates from the
official "PUSH".
POP XX
Pop the stack and store in register XX.
INC XX
Increment register XX.
DEC XX
Decrement register XX.
CMPD
Double-word compare DS:SI to ES:DI and increment SI and DI. Notice the
mnemonic deviates from the official "CMPSD".
CMPW
Word compare DS:SI to ES:DI and increment SI and DI. Notice the mnemonic
deviates from the official "CMPSW".
LODW
Read a word from DS:SI into AX. Notice the mnemonic deviates from the
official "LODSW".
LODB
Read a byte from DS:SI into AL. Notice the mnemonic deviates from the
official "LODSB".
STOW
Write a word from AX to ES:DI. Notice the mnemonic deviates from the
official "STOSW".
STOB
Write a byte from AL to ES:DI. Notice the mnemonic deviates from the
official "STOSB".
JNE YY
If the zero flag is not set, then add YY to the instruction pointer where
YY is a hexadecimal value that is exactly two characters long.
JMP YYYY
Add YYYY to the instruction pointer where YYYY is a hexadecimal value that
is exactly four characters long.
CAL YYYY
Push the current instruction pointer and add YYYY to the instruction
pointer where YYYY is a hexadecimal value that is exactly four characters
long. Notice the mnemonic deviates from the official "CALL".
RETN
Pop the stack and copy the value to the instruction pointer (near return).
Notice the mnemonic deviates from the official "RET".
-~=<|$| Pseudo-instructions |$|>=~----------------------------------------------
In addition to the aforementioned instructions, CR-ASM supports two commands
which allow the user to write data directly to the output. Notice these commands
will write the characters which follow precisely (that includes white-space,
new-lines, etc.).
DSW XX
Write the characters XX to the output file.
DSD XXXX
Write the characters XXXX to the output file.
-~=<|$| Bye-bye! |$|>=~---------------------------------------------------------
Well, that about wraps it up. Thanks for taking the time to check out my
little labour of love. I'd love to see someone pull this off in 256b. If you do,
make sure to drop me a line. Greets go out to sensenstahl, jmph, frag,
g0blinish, Rrrola, YOLP, and all size-coders.
orbitaldecay 2015[ back to the prod ]