Hi, you need at least some kind of synchronisation, but 70% does seem a lot. I can give the following suggestions: - implement the primitives in hardware :) - try to group the software primitives together - perform all the software primitives on a separate surface, and then perform a single 'blit' to update the (front)buffer. Or perform it on a separate layer alltogether.. This can mean that you program towards your platform, and away from portability.
hth, Niels Yaowen Feng wrote: > Hi, All > > I have tested some primitives which are not accelerated by hardware. > Software fallbacks will lose almost 70% performance when hardware > acceleration is enabled than software only. > > I looked into the DirectFB and find out that extra FlushReadCache will > be called when hardware acceleration is enabled. > > DirectFB uses this function to sync the memory between HW and SW. And > this causes the performance losing when hardware acceleration is enabled. > > Am I right? And how to avoid such performance losing? > > -- > Thanks, > Yaowen.Feng > ------------------------------------------------------------------------ > > _______________________________________________ > directfb-dev mailing list > directfb-dev@directfb.org > http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-dev > -- .------------------------------------------. | DirectFB - Hardware accelerated graphics | | http://www.directfb.org/ | "------------------------------------------" _______________________________________________ directfb-dev mailing list directfb-dev@directfb.org http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-dev
