efficient compilation of #batcktab()

[artrag]

I'm compiling this line of code

A = PEEK($0200+(X-8)/8+ (Y+3 AND $F8)/2*5)

and I get this

    MVI var_X,R1
    SUBI #8,R1
    SLR R1,2
    SLR R1,1
    ADDI #512,R1
    MVI var_Y,R2
    ADDI #3,R2
    ANDI #248,R2
    SLR R2,1
    MULT R2,R4,5
    ADDR R2,R1
    MVI@ R1,R0
    MVO R0,var_&A
 

Later I compile this line:

A = PEEKAT(X-8,Y+3) 

where I've defined

DEF FN PEEKAT(x,y) = #backtab((x)/8+ ((y) AND $F8)/2*5)

and I get this:

    MVI var_X,R1
    SUBI #8,R1
    SLR R1,2
    SLR R1,1
    MVI var_Y,R2
    ADDI #3,R2
    ANDI #248,R2
    SLR R2,1
    MULT R2,R4,5
    ADDR R2,R1
    MVII #Q2,R2
    ADDR R2,R1
    MVI@ R1,R0
    MVO R0,var_&A
 

It seems that this latter solution using backtab is slower and larger in ROM

Is this correct?

And if yes, why?

 

[+nanochess]

The IntyBASIC compiler isn't smart enough to discover that #backtab is a constant, so it loads it separately inside a register, instead of adding it directly.

[artrag]

can't the 200h address be embedded in the asm generated by the compiler?

[intvnut]

On 10/15/2019 at 10:45 AM, artrag said:

((y) AND $F8)/2*5)

I keep staring at that and it hurts my eyes.  I know you're losing 6 cycles to "ADDI" being split into "MVII + ADDR", but it feels like you're losing as much to "SLR + MULT."

 

This sequence:

    ANDI #248,R2
    SLR R2,1
    MULT R2,R4,5
    ADDR R2,R1

expands to:

    ANDI #248,R2
    SLR R2,1
    MOVR R2,R4
    SLL R2,2
    ADDR R4,R2
    ADDR R2,R1

That's 40 cycles and 7 words.

 

If you coded this in assembly yourself you could do:

    ANDI #$F8, R2
    SLR R2, 1
    ADDR R2, R1
    SLL R2, 2
    ADDR R2, R1

That's 34 cycles and 6 words, and leaves R4 untouched.

 

*shrug*

 

If you care about 6 cycles and one word, why not code directly in assembly?