Optimization for speed/efficiency BASIC XE

[Ricky Spanish]

Would this be as optimized as it will get, or is there still room for improvement with OSS BASIC XE

for speed/efficiency etc. Thanks.

 

1 FAST
165 DIM Q(158)
170 LS=10:MC=81:FOR I=0 TO LS-1:FOR J=0 TO 7:Q(78+I*8+J)=8-J:NEXT J:NEXT I
180 FOR I=0 TO 77:READ X:Q(I)=X:NEXT I:Z=Q(24):L=0:T=L:Q(158)=L:DC=T
190 CX=Q(L):CY=Q(L+12):D=64:D2=D*2:H=D-1:I=L*16+1
200 GRAPHICS 11
210 IF D<=1 THEN GOTO 270
220 FOR Y=T TO 191 STEP D
230   B=CY+(Y-96)/4*Z
240   FOR X=0 TO 79
250     A=CX+(X-40)*Z
260     COLOR 1
270     PLOT X,Y
280     R=0:I=R:R2=R:I2=R
290     FOR C1=0 TO MC-1
300       J=R2-I2+A
310       I=(2*R*I)+B
320       R=J:R2=R*R:I2=I*I
330       C=C1
340       IF (R2+I2)<4 THEN NEXT C1
350     COLOR Q(C+78)
360     PLOT X,Y
370     DRAWTO X,Y+H
380 NEXT X
390 NEXT Y
400 GOTO 270
410 Data 0,-0.170337,-0.745107298,-.462160300,-1.6487359367
411 Data -0.17485,-1.287769,.3150751,-1.9990959,-1.192101,-0.747227,-1.457758
420 Data 0,-1.06506,-0.15020534,-.5823998,0,-1.071623,.063761,.02918,-0.000000063,0.3061769,0.105766,-0.001353
430 Data 0.05,.000001,.000002,.0001,.000001,0.000045,.000004,.0004,.00000003,.00001,0.000147,0.00002
440 Data 254,2,32,14,18,16,238,242,240,1,1,1,0,11,11,11
450 Data 0,0,0,0,2,.5,0,0,2,.5,0,0,2,.5,0,1,1,1,0,1,1,1,0,0,0,0

[_The Doctor__]

if you aren't restoring to or pointing to specific data block, putting as many things on a line as possible helps. Same thing goes for command lines get as many on a line as possible. only using another line when something needs to go to it specifically

you still want to follow all the order of operation speed ups of Atari Basic, it helps when FAST mode fails. Looks like you've done pretty well in that regard.

You still want to compress line numbers down to as small a series of digits and get as much on a line as possible, while FAST mode does do addressing of the line ahead of time it still helps for the process of doing that or if FAST fails keeps things moving as best as possible.

 

 

[pirx]

the algo is flawed, it needs a rewrite. maybe this hacky way of jumping inside FOR and skipping NEXT is OK for a small program, but it will definitely fail in the long run (return stack gets clobbered).

smol opty:  D*2 is slower than D+D.

divide very bad.

floating point generally extremely bad, especially such a smol number as .00000003. at the end everything is integers, so maybe some kind of a rethink with integers would help. some kind of a precalced data maybe

[Ricky Spanish]

35 minutes ago, pirx said:

the algo is flawed, it needs a rewrite. maybe this hacky way of jumping inside FOR and skipping NEXT is OK for a small program, but it will definitely fail in the long run (return stack gets clobbered).

smol opty:  D*2 is slower than D+D.

divide very bad.

floating point generally extremely bad, especially such a smol number as .00000003. at the end everything is integers, so maybe some kind of a rethink with integers would help. some kind of a precalced data maybe

Thanks for the heads up. I'll look closer.

[Rybags]

Probably some gain to be had by stacking statements on a line rather than just 1 per line.

Possibly reordering so that the main loops are closer to the start though I think Basic XE doesn't suffer the same bad search speed for branches that Atari Basic does.

Where there's complex calculations - if there's stuff like integer multiplies on a sequence of numbers in a loop you can often get speed gain by just precalculating and storing the results in an array for later reference.

Edited March 26 by Rybags

[Ricky Spanish]

1 hour ago, Rybags said:

Probably some gain to be had by stacking statements on a line rather than just 1 per line.

Possibly reordering so that the main loops are closer to the start though I think Basic XE doesn't suffer the same bad search speed for branches that Atari Basic does.

Where there's complex calculations - if there's stuff like integer multiplies on a sequence of numbers in a loop you can often get speed gain by just precalculating and storing the results in an array for later reference.

Thanks. I'll rewrite it.

[FifthPlayer]

5 hours ago, Rybags said:

 

Possibly reordering so that the main loops are closer to the start though I think Basic XE doesn't suffer the same bad search speed for branches that Atari Basic does.

 

It does not.  The FAST keyword precalculates the branch points, so it does not have to search for the destination of a branch the way Atari Basic does.

[Ricky Spanish]

So, rewrote the program.

 

10 Fast :Dim Q(158):Graphics 11
15 Ls=10:Mc=81:For I=0 To Ls-1:For J=0To 7:Q(78+I*8+J)=8-J:Next J:Next I
20 For I=0 To 77:Read X:Q(I)=X:Next I:Z=Q(24):L=0:T=L:Q(158)=L:Dc=T
25 Cx=Q(L):Cy=Q(L+12):D=64:D2=D*2:H=D-1:I=L*16+1
30 On D<=1 Goto 55
35 For Y=T To 191 Step D:B=Cy+(Y-96)/4*Z:For X=0 To 79:A=Cx+(X-40)*Z:Color 1:Plot X,Y
40     R=0:I=R:R2=R:I2=R:For C1=0 To Mc-1:J=R2-I2+A
45       I=(2*R*I)+B:R=J:R2=R*R:I2=I*I:C=C1:If (R2+I2)<4 Then Next C1
50     Color Q(C+78):Plot X,Y:Drawto X,Y+H
55     Poke 77,0:Goto 65
60     Goto 25
65     If D>1 Then Next X:Next Y:D2=D2/2:D=D2:T=D/2:H=T-1:Goto 35
70 Goto 70
75 Data 0,-0.170337,-0.745107298,-.462160300,-1.6487359367,-0.17485,-1.287769
80 Data .3150751,-1.9990959,-1.192101,-0.747227,-1.457758,0,-1.06506,-0.15020534,-.5823998,0
85 Data -1.071623,.063761,.02918,-0,0.3061769,0.105766,-0.001353,0.05,0,0,0,0,0,0,0,0,0,0,0
90 Data 254,2,32,14,18,16,238,242,240,1,1,1,0,11,11,11,0,0,0,0,2,.5,0,0,2,.5,0,0,2,.5,0,1,1,1,0,1,1,1,0,0,0,0

[dmsc]

Hi!

2 hours ago, Ricky Spanish said:

So, rewrote the program.

 

10 Fast :Dim Q(158):Graphics 11
15 Ls=10:Mc=81:For I=0 To Ls-1:For J=0To 7:Q(78+I*8+J)=8-J:Next J:Next I
20 For I=0 To 77:Read X:Q(I)=X:Next I:Z=Q(24):L=0:T=L:Q(158)=L:Dc=T
25 Cx=Q(L):Cy=Q(L+12):D=64:D2=D*2:H=D-1:I=L*16+1
30 On D<=1 Goto 55
35 For Y=T To 191 Step D:B=Cy+(Y-96)/4*Z:For X=0 To 79:A=Cx+(X-40)*Z:Color 1:Plot X,Y
40     R=0:I=R:R2=R:I2=R:For C1=0 To Mc-1:J=R2-I2+A
45       I=(2*R*I)+B:R=J:R2=R*R:I2=I*I:C=C1:If (R2+I2)<4 Then Next C1
50     Color Q(C+78):Plot X,Y:Drawto X,Y+H
55     Poke 77,0:Goto 65
60     Goto 25
65     If D>1 Then Next X:Next Y:D2=D2/2:D=D2:T=D/2:H=T-1:Goto 35
70 Goto 70
75 Data 0,-0.170337,-0.745107298,-.462160300,-1.6487359367,-0.17485,-1.287769
80 Data .3150751,-1.9990959,-1.192101,-0.747227,-1.457758,0,-1.06506,-0.15020534,-.5823998,0
85 Data -1.071623,.063761,.02918,-0,0.3061769,0.105766,-0.001353,0.05,0,0,0,0,0,0,0,0,0,0,0
90 Data 254,2,32,14,18,16,238,242,240,1,1,1,0,11,11,11,0,0,0,0,2,.5,0,0,2,.5,0,0,2,.5,0,1,1,1,0,1,1,1,0,0,0,0

You did not rewrite the program, only rearranged it a little.

 

Your program is an adaptation of an Atari BASIC 10-liner, from this post: https://workshop88.com/oldblog/index.php/2021/03/21/an-interactive-mandelbrot-set-explorer-in-10-lines-of-atari-basic/

 

Being a 10-liner, it is written to be as concise as possible, without much regard for speed - but the algorithm is a standard mandelbrot plotter. Try to understand the code and write it again without the NEXTs inside the IFs, and without all the unneeded gotos.

 

Have Fun!

Edited March 27 by dmsc
clarify it is for Atari BASIC.

[Ricky Spanish]

12 minutes ago, dmsc said:

Try to understand the code and write it again without the NEXTs inside the IFs, and without all the unneeded gotos.

will do !

[Ricky Spanish]

curiously though, what's the problem with putting NEXT inside a IF statement ?

[_The Doctor__]

55     Poke 77,0:Goto 65
60     Goto 25

delete :GOTO 65

DELETE 60

 

they serve no purpose to get started, I see DMSC pointed out another possible issue

IF with a NEXT

it's possible to call for for a NEXT after the FOR count has been expended. This usually not a great idea.

[Ricky Spanish]

1 hour ago, _The Doctor__ said:

55     Poke 77,0:Goto 65
60     Goto 25

delete :GOTO 65

DELETE 60

 

they serve no purpose to get started, I see DMSC pointed out another possible issue

IF with a NEXT

it's possible to call for for a NEXT after the FOR count has been expended. This usually not a great idea.

yeah I got rid of the goto's. Why isn't it a great idea ?  I couldn't find any reason for it in my books (PDF).

It's only interrupting the FOR/NEXT, no ?

[Ricky Spanish]

Apon further research, the program is skipping the rest of the code within the "FOR" loop and moving on to the next iteration when the condition in the "IF" statement is met. So, how would it be rewritten to accomplish the same task ? Assuming it needs to be rewritten.

 

10 Fast :Dim Q(158):Graphics 11
15 Ls=10:Mc=81:For I=0 To Ls-1:For J=0
To 7:Q(78+I*8+J)=8-J:Next J:Next I
20 For I=0 To 77:Read X:Q(I)=X:Next I:Z=Q(24):L=0:T=L:Q(158)=L:Dc=T
25 Cx=Q(L):Cy=Q(L+12):D=64:D2=D*2:H=D-1:I=L*16+1
30 On D<=1 Goto 55
35 For Y=T To 191 Step D:B=Cy+(Y-96)/4*Z:For X=0 To 79:A=Cx+(X-40)*Z:Color 1:Plot X,Y
40     R=0:I=R:R2=R:I2=R:For C1=0 To Mc-1:J=R2-I2+A
45       I=(2*R*I)+B:R=J:R2=R*R:I2=I*I:C=C1:If (R2+I2)<4 Then Next C1
50     Color Q(C+78):Plot X,Y:Drawto X,Y+H
55     Poke 77,0
65   Next X:Next Y:D2=D2/2:D=D2:T=D/2:H=T-1:Goto 35
70 Goto 70

 

same data statement as above. Thanks.

Edited March 28 by Ricky Spanish
my grammar sucks.

[_The Doctor__]

What does the display look like at this point?

[Ricky Spanish]

1 hour ago, _The Doctor__ said:

What does the display look like at this point?

Best if you run it with Altirra or if you have real hardware & BASIC XE.

[dmsc]

Hi!

4 hours ago, Ricky Spanish said:

Apon further research, the program is skipping the rest of the code within the "FOR" loop and moving on to the next iteration when the condition in the "IF" statement is met. So, how would it be rewritten to accomplish the same task ? Assuming it needs to be rewritten.

 

10 Fast :Dim Q(158):Graphics 11
15 Ls=10:Mc=81:For I=0 To Ls-1:For J=0
To 7:Q(78+I*8+J)=8-J:Next J:Next I
20 For I=0 To 77:Read X:Q(I)=X:Next I:Z=Q(24):L=0:T=L:Q(158)=L:Dc=T
25 Cx=Q(L):Cy=Q(L+12):D=64:D2=D*2:H=D-1:I=L*16+1
30 On D<=1 Goto 55
35 For Y=T To 191 Step D:B=Cy+(Y-96)/4*Z:For X=0 To 79:A=Cx+(X-40)*Z:Color 1:Plot X,Y
40     R=0:I=R:R2=R:I2=R:For C1=0 To Mc-1:J=R2-I2+A
45       I=(2*R*I)+B:R=J:R2=R*R:I2=I*I:C=C1:If (R2+I2)<4 Then Next C1
50     Color Q(C+78):Plot X,Y:Drawto X,Y+H
55     Poke 77,0
65   Next X:Next Y:D2=D2/2:D=D2:T=D/2:H=T-1:Goto 35
70 Goto 70

 

same data statement as above. Thanks.

 

My question is: what are you trying to accomplish?

 

Your program is still very hard to understand, you still have lines (like line 30 and line 70) that do not do anything, you are using variables instead of numbers in line 40 (I=R instead of I=0), the Q stores 6 different arrays of data, most of that not necessary, the color table is not appropriate for graphics 11, etc.

 

¿Do you want a faster mandelbrot program? There are faster languages for that.

¿Do you want to learn BASIC XE? Then, your code is not written using any BASIC XE special statements.

 

Have Fun!

[Ricky Spanish]

Ok well just trying to learn. Accomplish ? Well, as you know that 10-liner goes much past the character per line limit.

Wanted to see if I could modify it without all the fancy original programs bells & whistles & perhaps speed it up.

 

Sorry for wasting time.

 

[_The Doctor__]

You realize when typing in POKE 82,0 then you use the abbreviations and skip any extra spaces etc when typing those in. Let the Atari expand them.

when listing it will all be there but you have to squish it all back down to make changes else it will cut everything off after 3 rows in a line number

 

Edited March 29 by _The Doctor__

[dmsc]

Hi!

25 minutes ago, Ricky Spanish said:

Ok well just trying to learn. Accomplish ? Well, as you know that 10-liner goes much past the character per line limit.

Wanted to see if I could modify it without all the fancy original programs bells & whistles & perhaps speed it up.

 

Then, I think you could start with this simplified program, IMHO it is easier to understand:

 

100 MC=15
110 Z=0.05
120 T=0
130 CX=0
140 CY=0
150 D=64:D2=D*2:H=D-1
200 GRAPHICS 11
210 WHILE D>1
220   FOR Y=T TO 191 STEP D
230     B=CY+(Y-96)/4*Z
240     FOR X=0 TO 79
250       A=CX+(X-40)*Z
260       COLOR 1
270       PLOT X,Y
280       R=0:I=0:R2=0:I2=0
290       C=MC
300       WHILE (R2+I2<4) AND C
310         J=R2-I2+A
320         I=(2*R*I)+B
330         R=J:R2=R*R:I2=I*I
340         C=C-1
350       ENDWHILE
360       COLOR C
370       PLOT X,Y
380       DRAWTO X,Y+H
390     NEXT X
400     POKE 77,0
400   NEXT Y
410   D2=D2/2:D=D2:T=D/2:H=T-1
420 ENDWHILE
430 GOTO 430

 

The inner loop is the same as the original, but now it uses a WHILE loop instead of FOR, so the exit condition is clearer.

 

Now, you could try to make it faster:

- You could change the starting condition (R=0, I=0,R2=0,I2=0) with the condition after the first iteration: R=A,I=B,R2=R*R,I2=I*I, this will make the code faster because you are doing one less loop

- You could lower MC and see if the mandelbrot has enough resolution still

 

You could also rewrite it in TurboBasic XL, it should be faster, and TurboBasic has better structured programming so could be simplified a little.

 

Have Fun!

[Ricky Spanish]

On 3/28/2024 at 9:34 PM, dmsc said:

Have Fun!

learned a lot from your example, thanks.

PS the higher the resolution, the better.

Edited March 30 by Ricky Spanish

[ivop]

If you want really fast, you could try porting this to the Atari: https://github.com/davidgiven/bogomandel

Interesting write ups on his blog. Links are in the README.md.

 

Have fun!

 

[Ricky Spanish]

3 hours ago, ivop said:

you could try porting this to the Atari

that's more @dmsc area of expertise. my programming abilities aren't to much above a apes.