Screen Fill on key press, project 4, part 1 (N2T)

fill.asm

@state
M=0;       //state is 0 at start(screen white)    

(ILOOP)
@SCREEN
D=A
@d_add
M=D        //SCREEN(add)=d_add 

@KBD
D=M        //checks for KBD scan code, feeds in D
@PRESSED
D;JNE      //change current state to -1, else current state 0
@c_state
M=0;       //else current state 0

(SUB)     //checks if current state is same as old
@state
D=M;
@c_state
D=M-D;    //state - current state = D
@ILOOP
D;JEQ    //if state = current state, goto ILOOP
@state
M=!M    //else, state is flipped(same as,  state = current state)


@24575     //last screen address
D=A
@i
M=D        //i=8192
(LOOP)
@i
D=M        //D=i
@ILOOP
D;JLE      //if i<=0; goto ILOOP
@state    // change from state to *KBD* for quicker updates, but KBD might update (if processing speed is slow) from the state and run half or either loop 
D=M;
@BLACK
D;JNE      //if D!=0; goto black, else white


(WHITE)
@i  
AD=M      //set display address as A
M=0      //set the selected register to 0
A=D-1
M=0
@2
D=A
@i
M=M-D     //i--
@LOOP
0;JMP    //goto LOOP

(BLACK)   //same as while, but set register to -1
@i
AD=M
M=-1
A=D-1
M=-1
@2
D=A
@i
M=M-D
@LOOP
0;JMP

(PRESSED)
@c_state
M=-1
@SUB
0;JMP


//scope: when (black) or (white) is runnning and mode is switched before completion, the remaning loop should run in reverse to earse/write remaining. (instead of starting over again)
//the M-1 and M-2 was me trying to find a marginal difference in reducing the instructions before the next register print, to see a perfomance gain.

The idea is such,
when the state is same, the writing to registers in loop doesn't occur. (this isn't in a improvement in speed but just reducing processing overhead)

Rest is pretty much straightforward. The last part with M-2 was an experiment, as mentioned in last comment