I couldn't get to sleep tonight (it's 6 in the morning) so I did what all software developers do when idle - I created an assembly language. No, really, I wanted to try my hand at creating some sort of toy interpreted assembly prototype for a while now.
So, without further ado, behold the next trendy programming language, aptly codenamed "asm"!
[size=1]here be dragons[code=:0]WORDSIZE 32REGISTER 3; Calculates the GCD of two integers@start ; Pop arguments off the stack (we know how many we expect) ; Now R0 and R1 contain the two arguments (in correct order) POP R1 POP R0 JMP loop@loop ; If R1 is zero, we're done and gcd is R0 CMP R1 0 EQL end ; Else do standard operation MOV R2 R1 MOV R1 R0 MOD R1 R2 MOV R0 R2 JMP loop@end ; Clear the stack CLR PUSH 1 ; Size of output array (a single integer) PUSH R0 ; The GCD of the two inputs RET ; We're done![code=:0]WORDSIZE 32REGISTER 4; Outputs the first N terms of the Fibonacci sequence@start ; Pop arguments off the stack (we read the number of elements desired) ; Now R0 contains the number of Fibonnaci elements desired POP R0 ; Initialize R1 and R2 to the Fibonnaci sequence MOV R1 0 MOV R2 1 ; Push the counter as the stack output size CLR PUSH R0 JMP head; This just takes care of outputting the first two terms if necessary@head ; If we are asked for zero terms.. make it so CMP R0 0 EQL end ; Output the first term PUSH R1 DEC R0 ; Do we want the second term? CMP R0 0 EQL end ; Output the second term PUSH R2 DEC R0 ; Begin main loop JMP loop@loop ; If R0 is zero, we've got all the terms we required CMP R0 0 EQL end ; Else, we need one more Fibonnaci term - calculate it in R3 MOV R3 R1 ADD R3 R2 ; R3 is now equal to R1 + R2 (the next term) MOV R1 R2 MOV R2 R3 ; Shift down the two elements PUSH R2 ; Output the new term to the stack DEC R0 JMP loop ; Next term!@end RET ; Nothing else to do here, we've streamed out the outputs[code=:0]WORDSIZE 32REGISTER 3; Adds a list of numbers passed as argument@start CNT R0 ; Get the number of input elements MOV R1 0 ; R1 will contain the sum JMP loop ; Start processing@loop ; If R0 is zero, we've added all the numbers CMP R0 0 EQL end ; Else pop the integer and add it POP R2 ADD R1 R2 ; Next element DEC R0 JMP loop@end CLR PUSH 1 PUSH R1 RET
And the best thing is that it actually works! This prints out the first seven Fibonacci terms:[code=:0]$ ./bin/asm fibonacci.asm 70112358
The parsing and interpreting code is absolutely horrible, though, it started out nice but stuff got out of control after adding the third instruction type. It should be pretty easy to write the code elegantly, especially in C++ (yes, I am also suicidal and wrote the parser in C)
Anyway, the input format is basically, the first stack word is an integer representing the number of input words, those input words are then pushed on the stack. So for the Fibonnaci example, the stack upon running the program looks like this:
  [...]
And the output format is very much the same, except the program basically erases the stack and outputs its stuff there. So after returning, it looks like this:
        [...]
Yes, it's not very convenient, but you need to start somewhere right? It'll probably need an input and output stream distinct from the actual stack eventually.
Also, the POP instruction will segfault if the stack is empty, and the stack is assumed to be infinite when really, it's limited to 1024 words because I needed an arbitrary limit (later it will need to grow dynamically). The language also supports arbitrarily many registers (defined by the REGISTER keyword at the top - WORDSIZE is supposed to indicate the word size of the registers in bits but they are always 32 bits wide at the moment)
Ultimately my goal is to have some sort of prototyping language between assembly and C that I can use to play with algorithms and stuff and get valuable behavioural info (e.g. highest stack/heap usage achieved, instruction parallelism level, etc..) which would be otherwise difficult to obtain via the x86 instruction set or other (though I suppose tools already exist for this, but whatever, this is for fun too)
I have to say it's pretty thrilling to actually create some sort of parser/compiler that takes your programs and makes them work somehow. Now I know why people are so interested in designing languages.. they are turned on by compilers! Anyway, that's it for tonight.
By the way, feel free to voice your disgust at the misshapen horror I have just brought into existence