Not getting expected performance improvement after accelerating bitBlit through KAA
prudhviraj.v at gmail.com
Thu Oct 29 22:34:57 PDT 2009
I have added the acceleration support to the Xfbdev server (in X windows
version 11 release 7.4) in order to improve the performance on ARM 11 board.
Among the accelerated hooks supported by KAA we have implemented
polyFillRect and CopyArea.
But when we run some client application on this accelerated server the calls
to the accelerated copyArea (which internally calls the hardware bitBlit
function) are coming very less frequently compared to the software blit
During my observation i have found some partially implemented functions like
kaaFillRegionTiled() etc which may internally call the blit operation
through (*pKaaScr->info->Copy) hook.
If at all this function could help to improve the performance of blit, can
you provide the fully implemented kaaFillRegionTiled().
When i have run one sample client application (gtkDemo) the fbBlt() function
got called around 40000 times where as the hardware accelerated blit
function called through kaaCopyNtoN (which calls (*pKaaScr->info->Copy)
internally) very few times (around 2500 times only) compared to fbBlt.
Can you provide more clarity why are most of the Blit calls are routed
through software fallbacks.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the xorg-devel