New commits:
http://git.directfb.org/?p=core/DirectFB.git;a=commit;h=6d00624797dd52e54a06cdb596bdc9f3db28291f
commit 6d00624797dd52e54a06cdb596bdc9f3db28291f
Author: Denis Oliver Kropp <[EMAIL PROTECTED]>
Date: Sun Aug 12 03:02:45 2007 -0700
FBDev System now with integrated surface manager...
Moved old surface manager code to frame buffer backend, removing a lot
of cruft, making the API more generally usable, probably moving it to a
utility library soon.
FBDev Framebuffer Memory surface pool uses the surface manager to allocate
and deallocate buffers in video memory. Added TestConfig() which uses a
new check only mode of the surface manager code.
Added VMWare acceleration driver, which does not really use acceleration
yet, but implements rectangle filling and simple blitting in software.
I used it for testing the new surface core on a 2.4.35-vmwarefb kernel,
quite nice...
-------------
Added the surface buffer to be allocated as another argument to
TestConfig().
The configuration alone is often not enough. The surface manager code needs
a buffer for checking, but it could probably be changed.
Added special handling for DFB_NOVIDEOMEMORY during surface pool
negotiation.
If any of the pools returned DFB_NOVIDEOMEMORY it does not return
DFB_UNSUPPORTED
in case of failure, but DFB_NOVIDEOMEMORY. No errors are printed.
Reenabled graphics drivers in build, but only VMWare by default.
Other drivers still need modifications, but trivial ones.
-------------
Another graphics driver API break:
CheckState() was being called before every repeated software fallback with
the same settings. The problem was that state->modified only gets reset in
the graphics driver's SetState(), but if CheckState() fails, it is not
called. Every time the state is to be checked, the modified flags are still
set and dfb_gfxcard_state_check() assumes the values have been changed and
activates rechecking...
To fix it, there's a new state member "mod_hw" which MUST be used by the
drivers instead of "modified". Every time the state is checked, the flags
are moved from "modified" to "mod_hw" after doing the recheck calculations.
Similarly, the software renderer could be changed to work the same way like
hardware drivers, utilizing another member like "mod_sw", instead of going
through the whole state everytime.
-------------
Added dfb_gfxcard_get_limits() and dfb_gfxcard_calc_buffer_size().
Added "min_size" to CoreSurfaceConfig, though not written yet, just read.
Updated TODO.
Makefile.am | 2 +-
TODO | 16 +-
configure.in | 42 ++-
gfxdrivers/Makefile.am | 9 +-
gfxdrivers/vmware/.gitignore | 6 +
gfxdrivers/vmware/Makefile.am | 34 +++
gfxdrivers/vmware/vmware_2d.c | 402 +++++++++++++++++++++++++++++
gfxdrivers/vmware/vmware_2d.h | 67 +++++
gfxdrivers/vmware/vmware_gfxdriver.c | 128 ++++++++++
gfxdrivers/vmware/vmware_gfxdriver.h | 58 +++++
src/core/coretypes.h | 4 -
src/core/gfxcard.c | 148 ++++++++---
src/core/gfxcard.h | 8 +
src/core/prealloc_surface_pool.c | 9 +-
src/core/state.h | 1 +
src/core/surface.h | 2 +
src/core/surface_buffer.c | 26 +-
src/core/surface_pool.c | 21 ++-
src/core/surface_pool.h | 1 +
systems/fbdev/Makefile.am | 14 +-
systems/fbdev/fbdev.c | 3 +
systems/fbdev/fbdev.h | 15 +-
systems/fbdev/fbdev_surface_pool.c | 148 ++++++++++-
systems/fbdev/surfacemanager.c | 466 ++++++++++++++++++++++++++++++++++
systems/fbdev/surfacemanager.h | 94 +++++++
systems/sdl/sdl_surface_pool.c | 1 +
26 files changed, 1608 insertions(+), 117 deletions(-)
_______________________________________________
directfb-cvs mailing list
[email protected]
http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-cvs
