final version of convolution filter ape

attached is the final version of my convolution filter ape.

it includes documentation and example files.

hope you enjoy it.

tom

this thing definately has some cool potential. there is glaring big one bug though. you didn't put the new ape in the zip file!!!

btw maybe you should make a quick re-release which puts the .Cff files into a subdirectory? The AVS directory is already supercluttered, with bitmaps, presets, apes, etc.

am i the only one getting everything but the ape in this zip file?

Jheriko is right, I also didn't find the ape in the zip, but if it really does what is promised in the documentation, then it's definitely neat stuff 🙂

Well there is a development version in another thread, but it's probably best to wait for cfp to repost because they might not be compatible.

They aren't compatible, i've tried the presets from the zip and I have the development version. The last development version was a bit crashy.

sorry about that...

i'm a muppet, i'd been up all night writing the documentation and doing the examples so i wasn't really thinking straight...

here's the final version, with ape (^_^)

no here is the final version with ape.

is there no end to my muppetry?

Thank you.

I love this thing, you can use it to make mini bitmaps for particles made from superscopes, its only gonna be 7x7 pixels but thats enough for me... I'm gonna go make something now.

this truely is awsome

check out what i made with it

Well, Karnov posted his, so I'll post a link to mine.

BUG: Sometimes the filter phuxup the RGB channels; when there's a WHITE render it'll somehow take a tiny bit out of one or two of the channels. The first time this happened I thought it was just the red channel (things were going red/cyan), but just today I had it going rygcbm (which meant that two of the channels where being changed, which means it's probably not a code error).
Also, could you have the load button default to open the last folder you loaded a CCF from, not the last folder you loaded a file to the WA playlist from?
Lastly, could you have a button that will let you get rid of that annoying 'all data will be lost' pop-up when you load a new CCF? Not the one saying 'really clear data,' because that's rather important...Wish the -main button would have that 😔

could you tell me exactly what you were doing when it went wrong?
e.g. values of the filter entries. + did the error seem to spread from any particular point on the screen?

there is also an error where winamp may come up with an error box while you're doing something completely unrelated to the ape. i can think of no reason why it should do this as if you do not have its dialog box visible it will not be changing its code or anything, so this may be a winamp bug.

however if this happens YOU CAN SAVE before you leave winamp.

just thought that'd be useful to know.

i'll try and fix any bug people point out to me, but i'm going back to uni this week so i'm not going to be devoting quite the amount of time to it i have recently.

as people are posting there convolution avs's here's mine (just to prove i'm not just a coder (^_^) )

tom

That's the thing - it wasn't one specific action that made it do that. I think the first time I was switching 2-pass off, and the second time I was docking it in the window. So it's not an AVS thing, it's just a glitch.

The winamp window comes up when you overload the variables (you have too many variables in a specific component).

here is a version of the filter (without any of the documentation) which remembers your save directory etc in most situations.

tom

p.s. if you want a link to quote for people to download this ape, i guess it's going to have be my http:\\cfp.begeek.net address, however lame it is.

and here's a version which includes all the docs.

Nice!

Although this is just a more advanced version of the Custom Filter, right? I'm gonna stick with the custom filter because it's basic!

i would not advise using "custom filter". have you compared their speeds?? custom filter may be simpler but it's only a fraction of the speed of convolution.

All those pretty presets, I just thought I'd have to add my piece to it all. This is a particle-ish superscope which uses the convolution matrix to turn all of the pixels into nice little circles, it's one of the best uses i've found for this thing outside of all of the usual graphics things.

(This is one of the presets from my up and coming Pack VI)

Wow, the convulution filter is 👍

Good job!

UnConeD should redo some of his presets, which could use this for faster edge detect.

BTW, the APE still stalls AVS when scale is set to 0 or lower.

here's the negative scale bug cured for the second time. you know the problem was pretty much the same as it was last time. i really should learn to test for bugs before releasing things (^_^).

however, all a negative scale does is to swap the sign of every entry. it doesn't allow you to do anything you couldn't already.

tom

and here's the full release again.

As long as your bug fixes are all backward compatible then I'm happy.

This thing is turning out to be really useful, I've used it in 4 presets in my Pack VI now.

Why is it that scales of powers of 2 are so much faster than others?



Edit: W00T W00T *fweeeeeeeeeee* 400th POST W00T!!!!!!!!! ...

cos if it's a power of two it does a bit shift rather than a mult.

this makes an even bigger difference for division because there is no mmx division instruction so the number of instructions goes from 1 to about 20. and that is with a still highly optimised fixed point division algorithm which doesn't actually use the slow div operation.

tom

slightly faster version.
no docs.

slightly faster version.
with docs.

I found it!

Whenever you use 2-pass and either wrap-around or absolute, the very top-left pixel turns blue. You can see this by using the 'accurate lapachian' preset with a render ring or something.

i've played around with it and i think it's a problem with bias when 2 pass mode is enabled. i think i've found the bug in the code too so expect a bug fix in a bit.

tom