[TI-BASIC] Help with Optimization

[Moose]

Hello, me again

I was just wondering if anyone could help me optimize this quick game I made. I tried every trick I can think of, but it still runs too slow for
too much enjoyment. Its just a quick one, so its very short

It's called glider, its just a game where you're a glider who flies over hills, and the closer you get to the ground, the more points you get.

Code: Select all

16->dim(L1
Fill(6,L1
6 -> E
5 -> A
0 -> B

Repeat A>L1(1 or K=45
ClrHome

Output(A,1,">
Output(8,1,"SCORE:
Output(8,9,B
For(F,1,16
Output(L1(F,F,"-
End

randInt(1,3 -> D
E+(D=1)-(D=2) -> E

If E>8
8 -> E

If E<2
2 -> E

E -> L(17
For(F,2,17
L1(F -> L1(F-1
End

getkey -> K
A-(K=25)+(K=34 -> A

If A<1
1->A

round((2^A)/10,0)+B -> B

End

There's a closing screen with a score, but that doesn't really matter.

[JoostinOnline]

Code: Select all

16->dim(L1
Fill(6,L1
DelVarB6 -> E
5 -> A

Repeat A>L1(1 or K=45
ClrHome

Output(A,1,">
Output(8,1,"SCORE:
Output(8,9,B
For(F,1,16
Output(L1(F,F,"-
End

randInt(1,3 -> D
E+(E<8)(D=1)-(E>1)(D=2) -> E

E -> L(17
For(F,2,17
L1(F -> L1(F-1
End

getkey -> K
A-(A>1)(K=25)+(K=34 -> A

B+ipart(2^A/10-> B

End

It would also speed it up alot if you used " " instead of ClrHome to erase where the glider had been, but I'm too lazy to add that in right now .

[burr]

Code: Select all

16→dim(L1:Fill(6,L1
DelVarB6→E:5→A
Repeat A>L1(1 or K=45
ClrHome
Output(A,1,">
Output(8,1,"SCORE:
Output(8,9,B
For(F,1,16
Output(L1(F),F,"-
End
1+int(3rand→D
max(2,min(8,E+(Ans=1)-(Ans=2→E
Ans→L(17
For(F,2,17
L1(F→L1(F-1
End
getkey→K
A-(Ans=25 and A>1)+(Ans=34→A
B+iPart(.1(2^A→B
End

[JoostinOnline]

@burr: If he wants speed, then he shouldn't use "and", even if it saves a byte. Also, I think I read somewhere that "/10" is faster than ".1"

[burr]

@burr: If he wants speed, then he shouldn't use "and", even if it saves a byte. Also, I think I read somewhere that "/10" is faster than ".1"

I don't know where you read that, but your assumptions are wrong -- and is faster than implicit multiplication, and multiplication is faster than division. The TI-84+/SE has a built-in timer command that allows you to test the speed of your code, so that's how I know.

[JoostinOnline]

@burr: If he wants speed, then he shouldn't use "and", even if it saves a byte. Also, I think I read somewhere that "/10" is faster than ".1"

I don't know where you read that, but your assumptions are wrong -- and is faster than implicit multiplication, and multiplication is faster than division. The TI-84+/SE has a built-in timer command that allows you to test the speed of your code, so that's how I know.

I have learned from experience that multiplication is faster than using "and". Its also in many guides such as http://www.ticalc.org/archives/files/fi ... 39433.html . However, you were right about ".1" being faster than "/10".

[Goplat]

I did an experiment in PindurTI to prove that "and" is faster. These numbers are from OS 1.12.

Code: Select all

Input		T states elapsed between ends of lines (breakpoint set at $18:67B1)

:(0)(0		15078
:(0 and 0	13700
:(0)(1		15126
:(0 and 1	13748
:(1)(0		15380
:(1 and 0	14002
:(1)(1		16721
:(1 and 1	13897

There you go - "and" is consistently faster than multiplying. (Edit: I had put the breakpoint in the wrong place so it was only measuring parsing speed, fixed)

[JoostinOnline]

I'm just going by all the guides on ticalc.org and personal experience.

[jimmothy]

idk, it seems it'd best to measure stuff on a real calc. just test to see which takes longer when executed 100 times or something.

[driesguldolf]

imo it sounds logic that and is faster, as it only has to check wether any of the arguments is zero, the output is zero. Multiplying as not so simple and thus takes longer.

BUT... have you checked 0*0 instead of (0)(0? Maybe there is a significant difference? Plus and doesn't need braces wich should make it faster & smaller than multiplying.

[JoostinOnline]

imo it sounds logic that and is faster, as it only has to check wether any of the arguments is zero, the output is zero. Multiplying as not so simple and thus takes longer.

BUT... have you checked 0*0 instead of (0)(0? Maybe there is a significant difference? Plus and doesn't need braces wich should make it faster & smaller than multiplying.

Yeah, I would think that and would be faster, but I just tested it on my calc and using multiplication saved 3 seconds when counting to 1000. But since its such a small time difference Moose may want to go with "and" to save a byte. I guess it just depends on how long your test will last.

[Goplat]

BUT... have you checked 0*0 instead of (0)(0? Maybe there is a significant difference? Plus and doesn't need braces wich should make it faster & smaller than multiplying.

Of course 0*0 would be faster than (0)(0, but that's irrelevant. To multiply conditionals you would need the parenthesis because the comparison operators have a lower precedence than multiplication.

I just tested it on my calc and using multiplication saved 3 seconds when counting to 1000.

What was the code you used? If you didn't put parenthesis around the terms, the comparison is meaningless.

I just tested it on my calc; the loop with multiplication took 7.0s while the loop with "and" took 6.7s.

[JoostinOnline]

I don't remember the code, but I know I used parenthesis. I don't really see why it matters though, since I already said that he should go ahead and use your advice (I had no idea that the time difference was so small)

[driesguldolf]

BUT... have you checked 0*0 instead of (0)(0? Maybe there is a significant difference? Plus and doesn't need braces wich should make it faster & smaller than multiplying.

Of course 0*0 would be faster than (0)(0, but that's irrelevant. To multiply conditionals you would need the parenthesis because the comparison operators have a lower precedence than multiplication.

You may want to do "If AnsP" instead of "If Ans and P" where it does matter, but in this case go for the "and" method (saves a byte and it's slightly faster)
it's not entirely 0*0 but wether and is faster than implicit multiplication or maybe explicit multiplication is the fastest?

And the slight difference makes me think that to calculate "and" it effectively multiplies them instead of ORing all the bytes together :S

[King Harold]

Shouldn't "AND" AND the numbers together?

"AND" doesn't even need to do that, it could be faster to check one parameter first, if it's false just go to ELSE, if it's true check the other parameter and go to ELSE if it's true
But that's too complicated for TI
although multiplying goes a bit far imo

note also that "IF" and "IF THEN" are very different, one of them is faster on FALSE, the other is faster on TRUE, which one though I don't remember..