From: Carsten H. (T. R. <ra...@ra...> - 2009-02-14 04:46:58
|
On Fri, 13 Feb 2009 11:46:38 +0200 Oguz Yarimtepe <com...@gm...> said: > Hi, > > > > this pretty much tells the problem. the ONLY test it seems to properly > > accelerate is the textblock intl test. unless something drastic went wrong > > here, but those numbers are very low for accelerated rendering on any > > device - > > that smells like software. very bad software too. > > > > try software_x11 or fb egnine of evas to compare with "pure software": > > > > expedite -e x11 > > expedite -e fb > > > > > > I tried with expedite -e fb and got better results. I didn't enable x11 > related options while compiling thats why i only tried with fb. Here are > some results. > > 3.35 , Image Blend Unscaled > 5.49 , Image Data ARGB Alpha > 25.00 , Image Blend Solid Unscaled > 2.00 , Rect Blend > 7.61 , Image Crossfade > > So the same results are slower when ran with DirectFB engine. When you said > "very bad software" you meaned the DirectFB libraries are bad or something > else? > > I may recompile after i installed a new version of DirectFB. dfb may not be accelerating at all - or its so bad at it that just plain old software in evas is better (if dfb isn't accelerating it is using its OWN software rendering code of course - and this likely is slower than evas's software). from what i can see of the results you have - dfb doesn't seem to be accelerating at all. just software. which makes dfb 100% useless. if it is accelerating - either the way it's doing it is really bad, or the graphics accelerator chip/core is just abysmal and useless. -- ------------- Codito, ergo sum - "I code, therefore I am" -------------- The Rasterman (Carsten Haitzler) ra...@ra... |