A while back I did an entry on Faster Initialization for the 2600. Today's entry is about squeezing the bytes out of the Initialization routine to the bare minimum. Here is a routine I came up with a while back:

;Economical 8 byte initialization routine
;By Omegamatrix
     cld
.loopClear:
     ldx    #$0A          ; ASL opcode = $0A
     inx    
     txs    
     pha    
     bne    .loopClear+1  ; jump between operator and operand to do ASL
                          ; A=0, X=0, Y=random, SP=$FF, carry is clear

While there are no illegal opcodes used in this routine, there are still a few tricks:

1) The accumulator is never loaded. Instead it gets left shifted, and eventually after 8 shifts A=0.

2) The opcode for ASL is used to load the X register with 10 the first time through. Subsequent loops jump between the operator and operand for LDX #$0A to perform ASL.

3) The writes with PHA start at $0B (REFP0), but the TIA registers will all get cleared through their mirrored addresses at $40-$7F.

 SP    REGISTER    WRITE VALUE (FROM ACCUMULATOR, which gets ASL'd)
 $0B     REFP0      %XXXXXXXX   <--- first time through
 $0C     REFP1      %XXXXXXX0   <--- second pass, etc...
 $0D     PF0        %XXXXXX00
 $0E     PF1        %XXXXX000
 $0F     PF2        %XXXX0000
 $10     RESP0      %XXX00000
 $11     RESP1      %XX000000
 $12     RESM0      %X0000000
 $13     RESM1      %00000000
 $14     RESBL      A=0 for now on
 
 ;writes continue to start of TIA mirrors
 SP    REGISTER    WRITE VALUE (FROM ACCUMULATOR)
 $40     VSYNC       0
 $41     VBLANK      0
 $42     WSYNC       0...
 ;Writes continue through ZP $80-$FF clearing RIOT RAM
 ;At end of routine TIA registers and RIOT RAM cleared,
 ;A=X=0, SP = $FF, Y=random, Carry is clear 

You probably have noticed that I have skipped using SEI. I did this because there is no interrupt line on the 2600. However, there is always a chance one of the weird clones will. Insert SEI at the top of the initialization routine if you feel the need.