Saving bytes, byte by byte - an example

[GARTHWILSON]

Bregalad wrote:

GARTHWILSON wrote:

Another one to add even to the NMOS part is removing the CMP #0 after anything affecting the accumulator, or the CPY #0 after anything affecting Y, or the CPX #0 after anything affecting X.

This is quite obvious, not even an "optimisation".

Totally agreed; but the waste of these instructions is epidemic. One person posting there recently said he just wanted to make the code more clear (which it does not do).

[RichTW]

A favourite one of mine is, to divide by 64, instead of:

Code: Select all

LSR A
LSR A
LSR A
LSR A
LSR A
LSR A

do this:

Code: Select all

ASL A
ROL A
ROL A
AND #3

Save 1 byte and 4 cycles. Not a big space-saver, but definitely worthwhile.

[CurtisP]

I'm working on a 6502-targeted compiler, which is generating a lot of STA/LDA to the same location, so I plan to write some sort of peep-hole optimizer. I've decided to use straight BRx instructions in the IF/ELSE structures, even though that will limit the number of instructions in the blocks, because they are so common. FOR/WHILE/DO loops will use BRx/JMP combinations.

[White Flame]

For doing an in-range test, instead of testing both the low bound & high bound, offset it so the range starts at 0 and do an unsigned test with the high bound.

Code: Select all

  lda value
  sec
  sbc #low_bound
  cmp #high_bound-low_bound
  bcs out_of_bounds

If you don't need the SEC in there, it saves a fair percentage of bytes & cycles.

[CurtisP]

As long as you don't need to preserve the accumulator, of course

White Flame wrote:

For doing an in-range test, instead of testing both the low bound & high bound, offset it so the range starts at 0 and do an unsigned test with the high bound.