I saw the stars =-)
Matrox G400 with DirectFB accelerators installed. You can find my
GGI_DEBUG=255 output here:
http://plug.org/~thayne/files/stars.out
Right now I'm porting a DirectFB benchmark to GGI so that I can have a
closely matched benchmark comparison: same accelerators, same benchmark,
different graphics libraries.
On 22 Aug 2001 00:10:46 -0400, Brian S. Julin wrote:
>
> Heylo,
>
> I made some updates to the new fbdev-directfb which may have corrected
> some problems with accels not being loaded quite properly. Now all
> the DirectFB global symbols are in their own sublib. (Thayne, for
> some reason this new sublib is not building in .libs, but out in the
> open, though it seems to install/work just fine.)
>
> In addition, it is no longer necessary to have the directfb source
> in order to build. Recent DirectFB versions seem to install the necessary
> headers to compile the target. So the --with-directfb configure directive
> has been coopted to set the directory in which DirectFB driver .so files
> are looked for. Unfortunately, the Debian -dev package does not yet include
> these header files, but as the packagers catch up I expect it will
> be possible to build with a .rpm or .deb of DirectFB. If you have
> a directfb-internal directory in your include path after installing DFB,
> odds are you can compile this just fine.
>
> Still no luck with my Riva 128, but if anyone wants to try it, preset
> your fbdev to a 16bpp mode and try stars. The screen should not clear,
> but the stars should be there. If you see a black rectangle painted
> at 100,100, then you have made it farther than me (currently fillscreen
> is hacked to only draw a small rectangle.) GGI_DEBUG=255 output should
> contain something like the following:
>
>
> LibGGI: display-fbdev: accel: 27, supported: 33
> LibGGI: display-fbdev: Have accel: "directfb"
> LibGGI: display-fbdev: Parsing modedb: /etc/fb.modes
> LibGGI: display-fbdev: begin: "640x480-60"
> LibGGI: display-fbdev: got 640x480 mode
>
> [...]
>
> LibGGI: Success in loading generic-linear-16 ()
> ^^
> You must be in 16-bit mode, DFB doesn't support pseudocolor, and only
> on a few cards is 32-bit supported.
>
> LibGGI: _ggiAddDL(0x8072290, "tgt-fbdev-directfb", "(null)", 0x1) called
> LibGGI: _ggiLoadDL("/usr/local/lib/ggi/default/fbdev/directfb/directfb.so", 0x1)
>called
> LibGGI: hand.handle=0x8078260
> LibGGI: &(GGIdl_directfb) = 0x4119f9dc
> LibGGI: hand.open = 0x4119f0f8
> LibGGI: hand.exit = (nil)
> LibGGI: hand.close = 0x4119f9a0
> LibGGI: _ggiLoadDL returned 0 (0x8078508)
> LibGGI: GGIopen for DirectFB started!
> LibGGI: _ggiAddDL(0x8072290, "helper-fbdev-directfb-global", "(null)", 0x4) called
> LibGGI: _ggiLoadDL("/usr/local/lib/ggi/default/fbdev/directfb/directfbglobal.so",
>0x4) called
> LibGGI: hand.handle=0x80787e8
> LibGGI: &(GGIdl_directfbglobal) = 0x411a2a50
> LibGGI: hand.open = 0x411a29c0
> LibGGI: hand.exit = (nil)
> LibGGI: hand.close = (nil)
> LibGGI: _ggiLoadDL returned 0 (0x8078ac0)
> Globals 0x411a3cc8 0x411a3ba8 0x411a2b1c 0x411a2b20 0x411a2b24 0x411a2b2c 0x411a2b28
> LibGGI: 0 = dlh->open(0x8072290, 0x8078ac0, "(null)", 0x8078530, 0) -
>/usr/local/lib/ggi/default/fbdev/directfb/directfbglobal.so
> Opened the helper
> globals 0x411a3cc8 0x411a3ba8 0x411a2b1c 0x411a2b20 0x411a2b24 0x411a2b2c 0x411a2b28
> card = 0x80785fc, config = 0x807854c
> LibGGI: Opening '/usr/local/lib/directfb/gfxdrivers/libdirectfb_neomagic.so'.
> LibGGI: Probing driver at 0x411a4e58...
> LibGGI: Opening '/usr/local/lib/directfb/gfxdrivers/libdirectfb_matrox.so'.
> LibGGI: Probing driver at 0x411a6584...
> LibGGI: Opening '/usr/local/lib/directfb/gfxdrivers/libdirectfb_ati128.so'.
> LibGGI: Probing driver at 0x411a6478...
> LibGGI: Opening '/usr/local/lib/directfb/gfxdrivers/libdirectfb_tdfx.so'.
> LibGGI: Probing driver at 0x411a63c8...
> LibGGI: Opening '/usr/local/lib/directfb/gfxdrivers/libdirectfb_nvidia.so'.
> LibGGI: Probing driver at 0x411a5a08...
> LibGGI: DirectFB driver, Init:0x411a517c InitLayers:0x411a5a4c DeInit:0x411a5474
> LibGGI: DirectFB/GfxCard: nVidia RIVA TNT/TNT2/GeForce 0.0 (convergence integrated
>media GmbH)
> LibGGI: Setting DSPF_RGB16
> LibGGI: 0 = dlh->open(0x8072290, 0x8078508, "(null)", (nil), 4) -
>/usr/local/lib/ggi/default/fbdev/directfb/directfb.so
>
> [...]
>
> LibGGI: fillrectangle = 0x411a5758, called
>
> A line like this means that the DirectFB fillscreen accel has been called.
>
> LibGGI: GGI_fbdev_idleaccel(0x8072290) called
> LibGGI: directfb_idleaccel(0x8072290) called
>
> [...]
>
> After closing the app:
>
> LibGGI: Disposing "helper-fbdev-directfb-global"
> LibGGI: Closing handle: 0x80787e8
> LibGGI: Disposing "helper-linux-vtswitch"
> LibGGI: L/vtswitch: close OK.
> LibGGI: Closing handle: 0x80732e0
> LibGGI: _ggiZapDL(0x8072290, 0x8072634) called
> LibGGI: Disposing "tgt-fbdev-directfb"
> (-) DirectFB/nvidia: FIFO Performance Monitoring:
> (-) DirectFB/nvidia: 0 nv_waitfifo calls
> (-) DirectFB/nvidia: 0 register writes (nv_waitfifo sum)
> (-) DirectFB/nvidia: 0 FIFO wait cycles (depends on CPU)
> (-) DirectFB/nvidia: 0 IDLE wait cycles (depends on CPU)
> (-) DirectFB/nvidia: 0 FIFO space cache hits(depends on CPU)
> (-) DirectFB/nvidia: Conclusion:
> (-) DirectFB/nvidia: Average register writes/nvidia_waitfifocall:nan
> (-) DirectFB/nvidia: Average wait cycles/nvidia_waitfifo call: nan
> (-) DirectFB/nvidia: Average fifo space cache hits: -2147483648%
> LibGGI: fbdev-directfb: Starting cleanup
> (-) DirectFB/nvidia: FIFO Performance Monitoring:
> (-) DirectFB/nvidia: 0 nv_waitfifo calls
> (-) DirectFB/nvidia: 0 register writes (nv_waitfifo sum)
> (-) DirectFB/nvidia: 0 FIFO wait cycles (depends on CPU)
> (-) DirectFB/nvidia: 0 IDLE wait cycles (depends on CPU)
> (-) DirectFB/nvidia: 0 FIFO space cache hits(depends on CPU)
> (-) DirectFB/nvidia: Conclusion:
> (-) DirectFB/nvidia: Average register writes/nvidia_waitfifocall:nan
> (-) DirectFB/nvidia: Average wait cycles/nvidia_waitfifo call: nan
> (-) DirectFB/nvidia: Average fifo space cache hits: -2147483648%
>
> (Stats are obviously broken in the Riva driver.)
>
> This driver should only be considered about 50% complete, but give it a try
> if for no other reason than to ruin your uptime record :-) (actually, although
> I have had to ALT-SysRq-k the app a couple of times, it hasn't hung me up yet.)
>
> Enjoy.
>
> --
> Brian
--
Thayne Harbaugh
