Any suggestions for better Performance of my Slide Show script?

[bigboss97]

I'm writing a program which generates a slide show script from an image database. The program uses Avisynth API to display the slide show frame by frame. Everything is working. But it takes sometime to process the script (before the slide show).

Since the script is generated automatically I have to use some variables to make the codes repeatable.

global ScrWidth= 640
global ScrHeight= 480
FPS= 16
TransLen= 32 #---- Frames of a transition
ImgLen= 64/2 +32 #---- Frames for a single image

import("head.avs")

B= KillAudio( BlankClip( ImgLen, ScrWidth, ScrHeight, color=$000000, pixel_type="RGB24"))

Img00064=""
A= B
B= Image2Clip("folksKJ_1024.jpg", ImgLen, 1)
V00= Dissolve( A,B, TransLen)

Img00160="02/12/2007 13:53"
A= B
B= Image2Clip(".\..\..\..\Photos\20071202_HydePark\r0010778.jpg", ImgLen, 1)
V01= TransBubbles( A,B, TransLen)

*snip*

Img01888="30/03/2011 10:34"
A= B
B= Image2Clip(".\..\..\..\Photos\20110326_Mexico\0329_San_Cristobal\img_2191.jpg", ImgLen, 1)
V19= TransBubbles( A,B, TransLen)

V20= B.FadeOut( TransLen)

V00+ V01+ V02+ V03+ V04+ V05+ V06+ V07+ V08+ V09+ V10+ V11+ V12+ V13+ V14+ V15+ V16+ V17+ V18+ V19+ V20



Since the resolution of the screen is usually lower than the digital photos most of the time they got downsized. I don't know how Avisynth operations are executed. I guess if I downsize the image at the right place all following operations can save a lot of time.

Any suggestions for optimization?


LoadPlugin("TransAll.dll")
SetMemoryMax(64)

function Image2Clip( string FName, int ImgLen, int Ortn)
{
B= KillAudio( BlankClip( 1, ScrWidth, ScrHeight, color=$001144, pixel_type="RGB24"))
C= Ortn== 3 ? ImageSource( FName).Turn180(): \
Ortn== 6 ? ImageSource( FName).TurnRight(): \
Ortn== 8 ? ImageSource( FName).TurnLeft(): ImageSource( FName)
Width= C.Width
# Width= C.Width/1.333 # Fix distortion of 4:3 resolution on 16:9 screen

ImgIsWide = float(Width)/C.Height >= float(ScrWidth)/ScrHeight
Fact= ImgIsWide ? 1.0 *ScrWidth/Width : 1.0 *ScrHeight/C.Height
NewWidth= Round( Fact* Width)
NewHeight= Round( Fact* C.Height)
O= C.BilinearResize( NewWidth, NewHeight)
return B.Overlay( O, (ScrWidth-NewWidth)/2, (ScrHeight-NewHeight)/2).Loop( ImgLen)
}



Thanks
Phuoc

[Gavino]

Any suggestions for optimization?
return B.Overlay( O, (ScrWidth-NewWidth)/2, (ScrHeight-NewHeight)/2).Loop( ImgLen)
It would be both quicker and less memory-intensive to use Layer instead of Overlay.
You could even instead use StackHorizontal/StackVertical to build up the sections of the final image, but that would require more tricky logic to cover the possible combinations.

[bigboss97]

It would be both quicker and less memory-intensive to use Layer instead of Overlay.
You could even instead use StackHorizontal/StackVertical to build up the sections of the final image, but that would require more tricky logic to cover the possible combinations.

Thanks, I'm going to try stacking. I assume it's quicker than layer, right?

One more performance question, will the comments in a function be read only once, i.e. more comments won't slow down the loop?

[Gavino]

Thanks, I'm going to try stacking. I assume it's quicker than layer, right?
I would expect it to be faster, but I haven't measured it.
If using this method, note that in your code you have to be careful to cover special cases like the image being placed at the edge of the background clip, or completely covering it.
One more performance question, will the comments in a function be read only once, i.e. more comments won't slow down the loop?
That's right. Comments are stripped out on the first pass of the parser.