Lumpy Tapioca Posted July 9 Share Posted July 9 (edited) // I use this KVP wrapper in every script now. It's surprisingly fast, even at full clock rate for animation, // and all variables are available for easy external debugging, by name. Often, I can eliminate using any native SL variables, // with little to no impact on execution speed. // It's not type safe and you can mess up, but it implements pointers and an easy, persistent cross-script variable pool // makes for much simpler code once you get used to it. // Setting and examining the contents of any and all variables from the command line or other scripts is a godsend. // F_( ) is write, F() is read. _F() is a pointer (returns the value of the value), which I didn't define here for simplicity // add these 8 lines: float F(string name) { return (float)llLinksetDataRead(name); } integer I(string name) { return (integer)llLinksetDataRead(name); } string S(string name) { return llLinksetDataRead(name); } integer B(string name) { if(((float)llLinksetDataRead(name)) != 0.0) return TRUE; return FALSE; } float F_(string name, float value) { llLinksetDataWrite(name,(string)value); return value; } integer I_(string name, float value) { llLinksetDataWrite(name,(string)((integer)value)); return ((integer)value); } string S_(string name, string value) { llLinksetDataWrite(name,value); return value; } integer B_(string name, integer value) { llLinksetDataWrite(name,(string)((integer)value)); return value; } // now you can do S_("name","lumpy"); // assigns "lumpy" to a new variable "name" llOwnerSay(S("name")); // prints lumpy // and S_("query", "name"); llOwnerSay(_S("query")); // also prints lumpy or F_("pi", 3.14159); and F_("two_pi", F("pi") * 2.0); // and llSetText(S("pi")); // prints without having to typecase to (string) (always remember that everything is really a string, watch your precision) - Lots more to this. matrix math, mult/add, lists, delete, search. sort, trig, increment/decrement/double/half... -- That's enough for now. Edited July 9 by Lumpy Tapioca Link to comment Share on other sites More sharing options...
Bleuhazenfurfle Posted July 12 Share Posted July 12 (edited) Why?!? You're inserting a whole extra function plus a bunch of casting, to save a little typing? I_() for example, casts an integer to float, back to integer, then to string — presumably in the name of saving having to write (integer) when you're trying to store the truncation of a float, at the cost of pointless double-casting (there's an implicit cast when you pass an integer to a float argument) when you're actually storing an integer. If you're storing a float into an integer space, I'd personally argue you want to have to explicitly cast it down the manual way, to make it obvious to yourself that you're doing it intentionally. That said, if saving a little typing (and the quirks you have embedded in your assumptions there) are sufficiently relevant to you, I'd suggest using Firestorm and it's pre-processor, and do all that as #define's instead. But also, just: integer B(string name) { return (float)llLinksetDataRead(name) != 0.0; } Edited July 12 by Bleuhazenfurfle Link to comment Share on other sites More sharing options...
Love Zhaoying Posted July 12 Share Posted July 12 3 minutes ago, Bleuhazenfurfle said: Why?!? You're inserting a whole extra function plus a bunch of casting, to save a little typing? Plus, the String version is exactly equivalent to just calling llLinksetDataRead()... @Lumpy Tapioca, thanks for your post. My own scripts use as much script memory as possible, so the extra memory needed for each function would be a deal-killer. (I don't remember for certain, but I think adding extra functions can result in another 512 bytes memory used, depending on different factors.) Link to comment Share on other sites More sharing options...
Bleuhazenfurfle Posted July 12 Share Posted July 12 9 minutes ago, Love Zhaoying said: @Lumpy Tapioca, thanks for your post. My own scripts use as much script memory as possible, so the extra memory needed for each function would be a deal-killer. (I don't remember for certain, but I think adding extra functions can result in another 512 bytes memory used, depending on different factors.) The figure that comes to mind is somewhere around 200 bytes, plus information about the arguments, and whatever code the function contains… The 512 figure (and the "can result in", should probably be "often results in") comes from bytecode being allocated in 512 byte blocks, so however much space it ends up consuming, you round that up to the nearest 512 bytes. (Though it's been a while since I cared, so don't take that figure as gospel.) Link to comment Share on other sites More sharing options...
Love Zhaoying Posted July 12 Share Posted July 12 13 minutes ago, Bleuhazenfurfle said: The figure that comes to mind is somewhere around 200 bytes, plus information about the arguments, and whatever code the function contains… The 512 figure (and the "can result in", should probably be "often results in") comes from bytecode being allocated in 512 byte blocks, so however much space it ends up consuming, you round that up to the nearest 512 bytes. (Though it's been a while since I cared, so don't take that figure as gospel.) That's right, depending on your code size / # functions you see the memory used jump in blocks of 512, etc. Link to comment Share on other sites More sharing options...
Lumpy Tapioca Posted October 5 Author Share Posted October 5 (edited) I've made some reasonably complex systems with this, and running out of space hasn't been a problem, plus allocation is dynamic and persistent. Also what's cool about this is, with a tiny shell, you can query or change any variable in the set, from the chat window or other objects. Every defined variable. No more globals. I can also connect variables dynamically at run time. Software-defined software. It also gives me pointers and run-time sparse arrays with a little more code. I've not seen a speed penalty either. I can't really feel the difference and I'm animating 100s of objects at a time. If you run off of a timer, instead of balls-out fast as it lets you, it's smooth, with plenty compute time left. Edited October 5 by Lumpy Tapioca update Link to comment Share on other sites More sharing options...
Recommended Posts
Please sign in to comment
You will be able to leave a comment after signing in
Sign In Now