Very important problem related to code size

[nathan999]

Hi! I just learned about this compiler a couple of weeks ago and began porting one of my 8 bits games to the Genesis. It's a flick-screen platformer where you have to step over all enemies in the map to clear a stage. There are several kinds of enemies, moving platforms, pushable blocks, keys and bolts, etcetera.

This were going pretty good until today. I was adding a new feature to the engine (to have timed "crushers" over a conveyor belt which would squish the player if not avoided). I hit F5 to compile and got this assembly error:

basic.s:8351: *** Error:  Word branch out of range ***
009DCA 6100 7FCC         bsr allign_heap2

Looking at the assembly listing Basiegaxorz generates, allign_heap2 must be an internal subroutine. It's called all over the code.

What's the problem? Well, the subroutine is being called with bsr, which is a relative jump. Relative means that it can only be a +-32K long jump. So if your code grows, as has happened to mine, the subroutine call cannot get to the actual subroutine. Which is a HUGE problem.

That means that your code can't grow much bigger than 32K (that is, just the code, not including graphics, data or variables).

It could be easily fixed modifying Basiegaxorz's code generator... but as this compiler is closed source and I've read the author is not further developing it I think I'm hopeless

Please take this in consideration before starting a big project. I have wasted so much time and effort into mine and it feels really bad when these things happen.

[Deleted]

Looks like allign_heap is called when passing a string as argument to a subroutine. Perhaps you could swap those out with a address / pointer to wherever the data is located .. or move the blocks of code that call these subroutines to the top of your source.

Pretty bad that the compiler doesn't check whether the generated BSR instruction is valid or not though.

[nathan999]

Yeah, I had the very same idea... I stripped my code first, making all variables globals and thus elliminating tons of parameter passing. Then I moved the subs which call subroutines with parameters to the top and got it working.

Phew!

Hopefully I won't need to add much more code, because I have the engine almost finished. Next time I'll take this important limitation in account

When the game is ready I will release it complete with full source code in case anybody is interested in taking a glance Right away it looks like this:



Thanks for the reply!

[Deleted]

Mar 12, 2013 2:56:43 GMT -5 nathan999 said:

Next time I'll take this important limitation in account

Subroutines that don't use any strings should be fine though .. so it's not a major limitation either.

Mar 12, 2013 2:56:43 GMT -5 nathan999 said:

Right away it looks like this

Looking good!

[nathan999]

It's not just the strings. The affected code is that which make calls to other Subs passing parameters

Now I'm struggling to code a decent sound driver.

[Deleted]

This is strange. I'm certain I've passed string variables to subroutines without any issues. It seems like it might be an issue with nested subroutines. One alternative would be to use a line label outside of a subroutine and use the shlink command, though I've never tried it so I don't know if it's bugged. I've not called other subroutines while in a sub or function though.

[sega16]

Wow that game looks amazing. I am very excited to play it.

[nathan999]

Glad to hear it I will keep you all posted when it's done.

Mar 12, 2013 14:04:24 GMT -5 @elusive said:

This is strange. I'm certain I've passed string variables to subroutines without any issues. It seems like it might be an issue with nested subroutines. One alternative would be to use a line label outside of a subroutine and use the shlink command, though I've never tried it so I don't know if it's bugged. I've not called other subroutines while in a sub or function though.

This certainly happens when I call SUBs/FUNCTIONs from inside SUBs/FUNCTIONs. I am very used to structured programming and I made my engine that way.

At first, not all my vars were global and the occurence of the offending bsr in the generated code was very high. When I activated the "all globals" directive in the compiler, bsr only appeared before calling SUBs or FUNCTIONS which take parameters.

I guess that the way to go is using just FUNCTIONs with parameters when needed, and don't nest calls, use GOSUB instead of SUBs and try that your code doesn't become a mess

[Deleted]

Mar 12, 2013 10:48:41 GMT -5 nathan999 said:

It's not just the strings. The affected code is that which make calls to other Subs passing parameters

You're not accidentally calling subroutines using the function syntax or something? The following snippet for example doesn't generate any allign_heap calls.

global time as Long
time& = 1000
addTime 60
print time&

declare sub extraTime(numToAdd as long)
time& = time& + numToAdd&
exit sub
end sub

declare sub addTime(numToAdd as long)
time& = time& + numToAdd&
extraTime 1
exit sub
end sub

[Deleted]

^was also thinking that.

On another note, if you have a dimensionalized array, you need the suffix or bex screams: global time&(10) as long

At least it did for me

[nathan999]

Just double-checked. It's when calling FUNCTIONs, not SUBs. I use a couple of functions to check collisions with the background in the SUBs which move stuff around. Thankfully, moving such SUBs up solved the problem. The "main" section isn't too big and BSR is more than enough to reach the allign_heap routine.

And yes, it does scream It shouldn't, but it does

[Deleted]

Functions are broken in BEX, I believe.

[Deleted]

Mar 14, 2013 5:27:21 GMT -5 nathan999 said:

Just double-checked. It's when calling FUNCTIONs, not SUBs.

I knew something was up Calling a function always aligns the heap, calling a subroutine doesn't always.

By the way, i recommend not using the "Make all variables in Subs / Functions GLOBAL" option. It's more flexible to leave that off and define the variables you want to be global using the global keyword yourself.

Mar 14, 2013 7:21:15 GMT -5 @elusive said:

Functions are broken in BEX, I believe.

Nothing wrong with functions.

print add(1,2)

declare function add(a as integer, b as integer) as integer
return a+b
end function

[Deleted]

Hmm, I thought they were, whoops I thought I read it somewhere on here, but maybe I'm confusing it with something else, or I'm just imagining things haha

[riper]

Hello forum,

I'm totally desperate. I spent hours and hours developing an engine that helps me to move the player accross rooms, printing a map and something else.

I have a piece of paper of a full map and I started to "make" rooms but... at the 8th room the same error as nathan999 appears on the screen.

Output File: basic.bin

Pass 1

Pass 2

basic.s:8181: *** Error:  Word branch out of range ***

00A24C 6100 7B4A        	bsr	allign_heap2

basic.s:8242: *** Error:  Word branch out of range ***

00A322 6100 7A74        	bsr	allign_heap2



00002 Total Error(s)





UnSuccessfully Compiled  !

2 Asmx errors

Have a nice day



Press ESC to close the log window

I can not continue with my game due this error.

I've readed this thread many times, but I'm not able to understand what is the problem (and the solution).
Is the lenght of the code the problem?
I have moved a lot of FUNCTIONS or/and SUBS from the bottom of my code to the top... but the problem is still there.

Its related to the global vars?

	GLOBAL moviment as integer 	
	moviment = 1 			

My code works with 1, 2... until 7 rooms... but I can not add any more.

Here my "game".

Game

(button C: from the game to the map... and from the map to the game)

I'm totally defeated... i'm near to give up.
What makes me ungry is that I don't understand the problem.

I don't know if someone wants to read my source code... I know that's a lot of job.

Well... thank you in any case.