dual head radeon -- again

Mark mark at harddata.com
Wed Mar 17 06:23:39 UTC 2004


On March 16, 2004 11:55 am, "Mike A. Harris" <mharris at redhat.com> wrote:
> On Tue, 16 Mar 2004 hansel at hansel.mnstate.edu wrote:
> >I've found many archived emails in many places about this problem with no
> >general solution and nothing that helps me now.
> >
> >This is the 3d OS install in about 8 months on this machine and the 3d
> >struggle to get the dual head display working properly.  This is a fedora
> >x86_64 install with XFree86 rpm=4.3.0-55.
>
> I presume "3d" above means "third" and not really "3D" as in 3D
> graphics, because dualhead and 3D acceleration are incompatible
> with each other.
 
Not with the newer DRI drivers. They added support for MergedFB to the Radeon 
Driver. This is similar to way FireGl Drivers work.

>From the Radeon Driver Option Page.

http://www.botchco.com/alex/radeon/mergedfb/cvs/DRI/final/radeon.4.html

<quote>
Option "MergedFB" "boolean"
    This enables merged framebuffer mode. In this mode you have a single 
shared framebuffer with two viewports looking into it. It is similar to 
Xinerama, but has some advantages. It is faster than Xinerama, the DRI works 
on both heads, and it supports clone modes.
    Merged framebuffer mode provides two linked viewports looking into a 
single large shared framebuffer. The size of the framebuffer is determined by 
the Virtual keyword defined on the Screen section of your XF86Config file. It 
works just like regular virtual desktop accept you have two viewports looking 
into it instead of one.
    For example, if you wanted a desktop composed of two 1024x768 viewports 
looking into a single desktop you would create a virtual desktop of 2048x768 
(left/right) or 1024x1536 (above/below), e.g.,
    Virtual 2048 768 or Virtual 1024 1536
    The virtual desktop can be larger than larger than the size of the 
viewports looking into it. In this case the linked viewports will scroll 
around in the virtual desktop. Viewports with different sizes are also 
supported (e.g., one that is 1024x768 and one that is 640x480). In this case 
the smaller viewport will scroll relative to the larger one such that none of 
the virtual desktop is inaccessable. If you do not define a virtual desktop 
the driver will create one based on the orientation of the heads and size of 
the largest defined mode in the display section that is supported on each 
head.
    The relation of the viewports in specified by the CRT2Position Option. The 
options are Clone , LeftOf , RightOf , Above , and Below. MergedFB is enabled 
by default if a monitor is detected on each output. If no position is given 
it defaults to clone mode (the old clone options are now deprecated, also, 
the option OverlayOnCRTC2 has been replaced by the Xv attribute XV_SWITCHCRT; 
the overlay can be switched to CRT1 or CRT2 on the fly in clone mode).
    The maximum framebuffer size that the 2D acceleration engine can handle is 
8192x8192. The maximum framebuffer size that the 3D engine can handle is 
2048x2048.
    Note: Page flipping does not work well in certain configurations with 
MergedFB. If you see rendering errors or other strange behavior, disable page 
flipping.
    The default value is undefined. 
Option "CRT2HSync" "string"
    Set the horizontal sync range for the secondary monitor. It is not 
required if a DDC-capable monitor is connected.
    For example, Option "CRT2HSync" "30.0-86.0"
    The default value is undefined. 
Option "CRT2VRefresh" "string"
    Set the vertical refresh range for the secondary monitor. It is not 
required if a DDC-capable monitor is connected.
    For example, Option "CRT2VRefresh" "50.0-120.0"
    The default value is undefined. 
Option "CRT2Position" "string"
    Set the relationship of CRT2 relative to CRT1. Valid options are: Clone , 
LeftOf , RightOf , Above , and Below
    For example, Option "CRT2Position" "RightOf"
    The default value is Clone. 
</quote>

This does work with Fedora Core 1. It appears that the Radeon Man Page needs 
to be updated. Hint, Hint.

regards, 


-- 
Mark Lane, CET mailto:mark at harddata.com 
Hard Data Ltd. http://www.harddata.com 
T: 01-780-456-9771   F: 01-780-456-9772
11060 - 166 Avenue Edmonton, AB, Canada, T5X 1Y3
--> Ask me about our Excellent 1U Systems! <--





More information about the amd64-list mailing list