forked from enlightenment/efl
638a790c63
By using a single XShmImage we avoid round trips to X and avoid having kernel to allocate (and zero) memory on every redraw. This also enable us to issue a single XShmPutImage() with the whole XShmImage just by using X Region and setting it as clip on Graphics Context (GC). On Nokia N800, expedite gains is about 10fps, while my other test with fewer objects (and thus drawing areas) I could go from 50fps to 160fps. Drawback is that we hold XShmImage until evas is resized or destroyed, we need a new API to flush engine memory so when it is idle for time we flush this memory, but it is kept alive during animations. SVN revision: 30390 |
||
---|---|---|
legacy |