From: Mukund JB. <muk...@es...> - 2005-03-07 18:07:33
|
Hello Burian, > The framebuffer implementation you get in the unofficial cirrus tree > (ep93xx-2.6) is not accelerated. The latest cirrus release (for > linux-2.4.21-rmk1) seems acclerated. Diff the source for details. >=20 > Note that the unaccelerated 2.6 driver and the accelerated > 2.4.21-rmk1-crus-whatever driver do not have a common anchestor, > the 2.6. was implemented from scratch, using fb-skeleton.c >=20 OK. > To explain it short: >=20 > if hardware acceleration is not available you'll have to emulate the > missing features in software (always possible), which is usually slow. Yes, right. This will be painful (performance wise). >=20 > if hardware acceleration is available but the driver does not use it > (emulates it as if it wasn't there) you still have a software-only driver. =20 You mean if the existing driver support is not accelerated, we need to emulate it in software again. But Why, can't we rewrite the driver at some places & make it work? > if hardware acceleration is available and the driver makes use of > it you've got a "accelerated" driver. In the ep93xx's case > the "raster engine" is used then. What is the raster engine? I goggled it, but did not find any thing strong enough. =20 My actual requirement is to rotate the screen by 90 degrees. Maillists ask me to use Xfree86 as it already has such kind of a support to rotate the screen.=20 Are there any side affects if I make use of the X server & rotate the screen. I mean is there any application that makes use of the video buffer directly without using X server and may fail to rotate the screen as we are rotating the screen at X server level and not at fb level. Also, if my device uses the console the screen rotation at X server level will not work. I mean if. Thanks & Regards, mukund jampala |