14 replies to this topic

[Random Terrain]

I know we can't use this:

if a{0} <> a{1} then blah blah blah

If I knew how to compare two bits, I forgot and I don't see it on the bB page.

The only thing I can think of is this:

if a{0} then if !a{1} then goto __Pause_Game

if !a{0} then if a{1} then goto __Pause_Game

Is that the best way to do it?



Thanks.

Edited by Random Terrain, Wed Sep 28, 2011 2:46 AM.

[RevEng]

You can xor them together. If they differ, the result is 1. If they are the same, the result is 0.

As far as efficiency, if the bits are neighbors you come out ahead 9 bytes of rom space, for your example...

temp1=a^(a/2)
if temp1{0} then goto __Pause_Game

[Random Terrain]

RevEng, on Wed Sep 28, 2011 8:54 AM, said:

You can xor them together. If they differ, the result is 1. If they are the same, the result is 0.

As far as efficiency, if the bits are neighbors you come out ahead 9 bytes of rom space, for your example...

temp1=a^(a/2)
if temp1{0} then goto __Pause_Game

Thanks. Since it uses division, does that mean it's a little slower?

[RevEng]

Yours is a bit faster because the comparisons avoid some work that mine does in constant time. (a division by 2 in this case only adds 2 cycles) I think your worst case scenario is 24 cycles, and mine is 33 cycles.

[Mr SQL]

Random Terrain, on Wed Sep 28, 2011 9:23 AM, said:

RevEng, on Wed Sep 28, 2011 8:54 AM, said:

You can xor them together. If they differ, the result is 1. If they are the same, the result is 0.

As far as efficiency, if the bits are neighbors you come out ahead 9 bytes of rom space, for your example...

temp1=a^(a/2)
if temp1{0} then goto __Pause_Game

Thanks. Since it uses division, does that mean it's a little slower?

RT,
Bitwise operations are fast but I think it would be easier to just evaluate the bitwise expression if it is handled:
if (a{0} ^ !a{1}) then goto __Pause_Game

[Random Terrain]

RevEng, on Wed Sep 28, 2011 9:50 AM, said:

Yours is a bit faster because the comparisons avoid some work that mine does in constant time. (a division by 2 in this case only adds 2 cycles) I think your worst case scenario is 24 cycles, and mine is 33 cycles.

OK, thanks. I better put this on the bB page too.

Mr SQL, on Wed Sep 28, 2011 10:59 AM, said:

RT,
Bitwise operations are fast but I think it would be easier to just evaluate the bitwise expression if it is handled:
if (a{0} ^ !a{1}) then goto __Pause_Game

Thanks, but I don't think we can do that with batari basic:

http://www.randomter...mmands.html#bit

[Emehr]

RevEng, on Wed Sep 28, 2011 8:54 AM, said:

You can xor them together. If they differ, the result is 1. If they are the same, the result is 0.

As far as efficiency, if the bits are neighbors you come out ahead 9 bytes of rom space, for your example...

temp1=a^(a/2)
if temp1{0} then goto __Pause_Game

Can you just bit-shift right once instead of dividing by 2? Or would that use the same number of cycles? (does bBasic support bit shifting?)

[Random Terrain]

Emehr, on Wed Sep 28, 2011 2:21 PM, said:

Can you just bit-shift right once instead of dividing by 2? Or would that use the same number of cycles? (does bBasic support bit shifting?)

We can use assembly language:

http://www.randomter...s.html#assembly

But I usually try to keep things in bB mode for the bB page to avoid confusing new users. Unless it's something cool like this:

http://www.randomter...#points_roll_up

[Mr SQL]

Random Terrain, on Wed Sep 28, 2011 2:01 PM, said:

RevEng, on Wed Sep 28, 2011 9:50 AM, said:

Yours is a bit faster because the comparisons avoid some work that mine does in constant time. (a division by 2 in this case only adds 2 cycles) I think your worst case scenario is 24 cycles, and mine is 33 cycles.

OK, thanks. I better put this on the bB page too.

Mr SQL, on Wed Sep 28, 2011 10:59 AM, said:

RT,
Bitwise operations are fast but I think it would be easier to just evaluate the bitwise expression if it is handled:
if (a{0} ^ !a{1}) then goto __Pause_Game

Thanks, but I don't think we can do that with batari basic:

http://www.randomter...mmands.html#bit

Sorry RT,
I meant:
if(a{0}^a{1}) then ...
must have left the other operator in there; should work without the parenthesis if the compiler doesn't like them since only a single term is in the expression.

[Random Terrain]

Mr SQL, on Wed Sep 28, 2011 2:37 PM, said:

Sorry RT,
I meant:
if(a{0}^a{1}) then ...
must have left the other operator in there; should work without the parenthesis if the compiler doesn't like them since only a single term is in the expression.

Check out the "Did You Know?" box here:

http://www.randomter...mmands.html#bit

Unless we were never told about it or something has changed, I think we are very limited when it comes to bit operations and if-thens.

Edited by Random Terrain, Wed Sep 28, 2011 2:45 PM.

[Mr SQL]

Random Terrain, on Wed Sep 28, 2011 2:44 PM, said:

Mr SQL, on Wed Sep 28, 2011 2:37 PM, said:

Sorry RT,
I meant:
if(a{0}^a{1}) then ...
must have left the other operator in there; should work without the parenthesis if the compiler doesn't like them since only a single term is in the expression.

Check out the "Did You Know?" box here:

http://www.randomter...mmands.html#bit

Unless we were never told about it or something has changed, I think we are very limited when it comes to bit operations and if-thens.

You're right, you can't use ^ that way, it flips two arguments I thought it was a comparison operator
I think your first solution has the least overhead.

[RevEng]

Emehr, on Wed Sep 28, 2011 2:21 PM, said:

Can you just bit-shift right once instead of dividing by 2? Or would that use the same number of cycles? (does bBasic support bit shifting?)

bB automatically turns a divide with a power-of-2 into a number of LSR bit-shifts.

[ScumSoft]

[edit]Very nice bogax!
My solution was way off

Edited by ScumSoft, Sat Oct 1, 2011 3:45 AM.

[bogax]

Not sure I remember the conventions rightly, but assuming bit 0 is the
least significant bit ...

If you add a 1 to a bit you can, in effect, propagate it to a more significant
position via carry

If you add two bits together the sum is the XORing of the two

You can do both at once in this case by adding 1 to a

Then use & to isolate bit 1 to test it (ie a bitwise AND with 2)

if (a + 1)& 2 then goto __pause_game
 
__pause_game
 

produces

.L00 ; if ( a + 1 ) & 2 then goto __pause_game
 
; complex statement detected
LDA a
CLC
ADC #1
AND #2
BEQ .skipL00
.condpart0
jmp .__pause_game
 
.skipL00
.
;
 
.__pause_game
; __pause_game
 

That ought to be something like what you want

Edited by bogax, Fri Sep 30, 2011 4:50 PM.

[bogax]

if a/2 ^ a & 1 then goto __pause_game
 
__pause_game
 

.L00 ; if a / 2 ^ a & 1 then goto __pause_game
 
; complex condition detected
; complex statement detected
LDA a
lsr
EOR a
AND #1
  BEQ .skipL00
.condpart0
jmp .__pause_game
 
.skipL00
.
;
 
.__pause_game
; __pause_game