performance - How to efficiently display a large number of moving points -
i have large array of points, updates dynamically. part, (relatively small) parts of array updated. goal of program build , display picture using these points.
if build picture directly points 8192 x 8192 pixels in size. believe optimization reduce array in size. application has 2 screen areas (the 1 magnification/zooming in of other). additionally need pan picture in either of screen areas.
my approach optimization follows.
- take source array of points , reduce scaling factor first screen area
- same second area, larger scaling factor
- render there 2 arrays in 2 fbos
- using fbos textures (to provide ability pan picture)
- when updating picture re-render changed area.
suggest ways speed current implementation runs extremely slow.
you hardly able optimize lot if don't have hardware run @ adequate rate. if render in different threads fbos , compose result, bottleneck remain. 67 million data points nothing sneeze at, modern gpus.
try not update unnecessarily, update changes, render what's updated , visible, try minimize size of components, e.g. use shorter data type if possible.
Comments
Post a Comment