The DX exec has a software cache. When you run the Sequencer, the graph (net) is evaluated for each of the inputs (values coming from the Sequencer). When the loop is finished and if you have enought memory+swap the end results are pulled from the cache. If you are using s/w rendering, then it's bitmaps. If you are using h/w rendering, it will be geometry to be passed to your graphics card. The default memory usage is 7/8 of physical RAM, which can be overridden by dx -memory n, which n is in MB. If not everything fits in cache and you can't increase memory+swap or reduce number/size of images, by default other parts of your net might be cached and taking up memory. The first thing you can do is use the built-in optimization available in the VPE. You can hand-tune nets, but you'll need to understand more about the execution model for this to work correctly, which will work even better.
If all you want to do is look at a post-processed animation, you can always dump the frames to disk (e.g., miff for IM), and use animate for playback or MPEG encode them and use an MPEG player... Hans Fangohr <[EMAIL PROTECTED]>@opendx.watson.ibm.com on 07/03/2001 07:08:38 AM Please respond to [email protected] Sent by: [EMAIL PROTECTED] To: [email protected] cc: Subject: [opendx-users] Speed of sequenced animations Greetings, I try to visualize a series of vector fields with nicely rendered Glyphs. If I start the sequencer, then the rate with which the different elements of the sequence are displayed is rather small, i.e. the animation is slow. I get a speed up, if I only display 20 pictures: After DX went trhougth them once, it speeds up a bit. However, having more pictures the speed stays small. I guess this has somehow to do with memory, and swapping data to disk/or recomputing it (?) ... Here is the question: Are there any general guidelines for making animations (using the sequencer) fast? Or: How does it work? Does DX render the data into bitmaps, and then store and redisplay those bitmaps (if you repeat the sequence)? Or does it do the rendering again and again ? What is the bottleneck here (if this can be said on such a general basis)? Any comments welcome. Many thanks, Hans
