Zugg Software :: View topic

Wizard Joined: 26 Mar 2008 Posts: 1547

Is there anything that would make scripts run slower in 3.20? I have a Clan Highlighter Script and am now working on a Damage Script + Damage Calculator.

Both of them do a lot of stuff, such that I would almost expect them to lag but they both ran perfectly in 2.37. However, having now upgraded to 3.20 BETA these scripts are suddenly running awfully slow. I could paste the whole scripts, but they require a lot of data from the MUD itself to function properly. However, if you could tell what might be slowing them down by looking at them I will.

GURU Joined: 23 Dec 2000 Posts: 4834 Location: USA

Yes. One of the bigger ones is the variable-by-reference in functions. Consider the following:

varname = %additem("item",@varname)

In this example, a second copy of @varname is created so it can be sent to the %additem() function. In addition, a third copy of the variable is created that includes "item" so it can be returned as a result. This is a by-value call and obviously it's pretty inefficient.

To call a variable by reference, you drop the @ just like when using #commands:

varname = %additem("item",varname)

Wizard Joined: 26 Mar 2008 Posts: 1547

So, dropping the @ will allow me to increase the speed and efficiency back to 2.37 levels?

GURU Joined: 23 Dec 2000 Posts: 4834 Location: USA

There are many more things, but yeah. For example, do you use the %eval() function in any of your code? If you do, remove the %eval and just use (). %eval() is a runtime function and thus about as slow as #EXEC.

Wizard Joined: 26 Mar 2008 Posts: 1547

I have a tendency to use %eval(), but I also have a tendency to see odd weird errors when I don't use it. Still, I'll try using () where I can. Thanks.

EDIT: Does this by-reference thing matter with var=@var too?

Wizard Joined: 26 Mar 2008 Posts: 1547

This feature isn't backwards compatible with 2.37, is it? It seems any scripts I've updated with it no longer function with 2.37.

Posted: Mon Jun 28, 2010 12:37 am

No, it's a feature that's new with the latest beta versions. Really, you should be on 3.21, by the way, and not 3.20.

Also, if you experience anything with %eval not working, you need to post about it as well. Keep in mind that using (expressions) within {} or "" will cause them to print literally instead of evaluating the expression, so you'll need to use %eval in situations such as:

#SAY {3+4 is %eval(3+4)}

or

#SUB {The answer is %eval(3+4)}

and so on.

Charneus

Magician Joined: 02 Feb 2004 Posts: 364 Location: USA

Concat works as well and should be faster..

#Say %concat( "3+4 is ", (3+4))

Wizard Joined: 26 Mar 2008 Posts: 1547

I'm on 3.21 Razz

Posted: Tue Jun 29, 2010 4:35 pm

Wizard Joined: 26 Mar 2008 Posts: 1547

Well, oddly enough, the script speed went back to 2.37 levels after changing all the variables to by-reference instead of what they were. But we're talking about lists with close to 2000 members, maybe that makes a difference?

Posted: Wed Jun 30, 2010 4:28 pm

It's possible that for some function combinations that the old hash tables in 2.37 didn't make a copy of the table and were essentially using the "by reference" concept. Of course, this would cause bugs in 2.37 because changing a by-reference list that hasn't been copied might end up changing the table stored in a different variable. v2.37 was rather inconsistent about all of that. So what you said makes sense.

It would be nice to have the functions use "by reference" by default for speed, but unfortunately everybody in older versions of CMUD are using the @ in front of the list and there is no way @varname should ever do a direct "by reference". So for backwards compatibility, there is no way to get the speed improvement automatically. People will need to edit their scripts to remove the @ where they want the by-reference speed.

Apprentice Joined: 15 Jun 2010 Posts: 121

is there a list of which functions support variable by reference?

Posted: Thu Jul 01, 2010 4:17 pm

Actually, the more I thought about this, the more confused that I got.

In 3.2x, any function that accepts a list/table for "read only" access (like %db, %item, %numitems, etc) always uses the "by-reference" regardless of whether you use @ or not. So these functions should not be slower than 2.37 even with the presence of the @ character.

Here is the list of functions effected by "by-reference":

Write-access list: @list makes a copy, list is "by-reference"
%additem
%delitem
%delnitem
%push
%replaceitem

Write-access table/dbvar: @table makes a copy, table is "by-reference"
%addkey
%delkey

Read-only access list or tables: no difference between list and @list. Both are "by-reference"
%db
%dbitems
%dbkey
%dbkeys
%dbvalues
%iskey
%ismember
%isvalue
%item
%numitems
%numkeys
%numwords

So your script from 2.37 will only be slower in some cases when using @list with the "Write Access" functions shown above. And any case where this is slower than in 2.37 indicates a bug in v2.37 where it was not making a copy of the table and could cause weird script side-effects with changing the underlying list value.