| Age | Commit message (Collapse) | Author |
|---|
|
Add additional checks for the *DRIRec info structure passed in from the
device driver. This ensures that things fallback to indirect rendering if
the DDX driver has had modifications (i.e. removal of the drmAddress field).
|
|
extension wasn't enabled. I have verified its correct function with Mesa's
stencil_wrap test. It is enabled on both Savage4 and Savage3D. Since
Savage3D uses a software fallback for *all* stencil operations, this is safe.
|
|
_glapi_add_entrypoint has been replaced by a new routine called
_glapi_add_dispatch. This new routine dynamically assignes dispatch offsets
to functions added. This allows IHVs to add support for extension functions
that do not have assigned dispatch offsets.
It also means that a driver has no idea what offset will be assigned to a
function. The vast majority of the changes in this commit account for that.
An additional table, driDispatchRemapTable, is added. Functions not in the
Linux OpenGL ABI (i.e., anything not in GL 1.2 + ARB_multitexture) has a
fixed offset in this new table. The entry in this table specifies the
offset in of the function in the real dispatch table.
The internal interface was also bumped from version 20050725 to 20050727.
This has been tested with various programs in progs/demos on:
radeon (Radeon Mobility M6)
r128 (Rage 128 Pro)
mga (G400)
|
|
that are currently obtained via glXGetProcAddress and all of the XF86DRI
functions are replaced with a funciton table. This table will be passed to
__driCreateNewScreen.
One of the functions in the table is getProcAddress. This allows some
loaders to expose functionality not in all loaders. This will be immediatly
used for glxEnableExtension (formerly known to drivers as
__glXScrEnableExtension). libGL (and in the future libglx) expose this
function so that drivers can enable GLX extensions. libEGL should exposed
eglEnableExtension to enable EGL extensions. The same function cannot be
used for both because the extensions have different names and (possibly)
different semantics. Drivers can optionally use one, both, or neither.
The key parts are in the __DRIinterfaceMethodsRec structure in
dri_interface.h. A pointer to one of these structures is passed into
__driCreateNewScreen. Because of this, the version of the API is bumped to
20050725. Since the previous version(s) were never in a release, their
existance is erased.
I was actually a little surprised by how much code this cuts from the
drivers. A lot of glXGetProcAddress calls disappear, and a lot of
version checks go with them. Nice.
The one thing I'm not sure of is removing __glXInitialize. For some
reason that function was in the glXGetProcAddress table, but *nothing*
in the Mesa tree used it. Did something with DRI conf. use this
function? It seems odd...
|
|
1,402 lines of code to be removed from Mesa (drivers and libGL). The
big winner is dri_util.c.
Primary changes are:
1. Remove all "deprecated" entry-points from the various structures in
dri_interface.h.
2. Rename the remaining fields to removed "version numbers." So,
bindContext3 becomes bindContext. Functions with "New" in the name
(e.g., CreateNewContext) were *not* changed, but that is an option.
Having "New" in the name is less annoying to me than having "3" in the name.
3. Remove all compatibility code that handles cases where the driver or
the loader is too old to support the latest interfaces.
4. Append the API version to the __driCreateNewScreen function name.
This is currently done by hand. In the future (i.e., the next time we
make an incompatible change to the interface) we'll want to come up with
a better way to do this. This prevents old loaders from being able to load
new (incompatible) drivers.
5. Bump the API version to 20050722. All drivers (by way of dri_util.c)
require this version.
6. All drivers are *required* to expose GLX_SGIX_fbconfig and
GLX_OML_swap_method (or the moral equivalents). Support for these
functions in implicit in the use of the "new" interface.
7. Some cases still exist that need to be compiled differently in a loader
or core Mesa versus in a driver. These are identified by the define
IN_DRI_DRIVER.
|
|
individual drivers and put them in common code. It is still possible for a driver to define its own macros if it has special needs. This affects CLIPPIXEL, CLIPSPAN, HW_CLIPLOOP, HW_ENDCLIPLOOP, and for drivers using the spantmp2 template also GET_SRC_PTR and GET_DST_PTR.
|
|
device-specific code. A new Python script
(src/mesa/glapi/extension_helper.py) generates a list of all
entry-points for all known extensions. Each driver the selects only
the extensions that it needs and enables the via either
driInitExtensions or driInitSingleExtension.
This code has been compile-tested on a drivers, but has only been
run-tested on mga and i915 (on i830 hardware).
These changes were discussed at length on the mesa3d-dev mailing list.
http://marc.theaimsgroup.com/?t=111947074700001&r=1&w=2
|
|
|
|
|
|
|
|
Main driver impacts:
- new code for creating the Mesa GLframebuffer
- new span/pixel read/write code
Some drivers not yet updated/tested.
|
|
|
|
a marginal speedup, but I'm not sure this has the same effect on all
hardware. Tested on Savage IX and ProSavageDDR. For experimenting with
different values see the macros at the start of savagestate.c.
|
|
S3TC support for an older driver version. On Savage3D/IX/MX
GL_EXT_texture_compression_s3tc can't be enabled because DXT3 and DXT5
are not supported. However GL_S3_s3tc is enabled on all Savage chips.
Tested on a Savage IX and a ProSavageDDR
|
|
channels are masked (disable draw update). Bumped driver date to
reflect recent correctness fixes. The driver now passes all glean
tests except exactRGBA on both Savage4 and Savage3D-based cards.
|
|
|
|
- force emitting texAddr when the texture image changed (flush caches)
- set transparent texture color key to something improbable
(couldn't find a way to disable it completely)
- fixed texture environment modes GL_DECAL and GL_REPLACE
- made texture environment mode GL_BLEND a software fallback
- added two custom texture formats for promoting from GL_ALPHA to ARGB8888
and 4444. Since the hardware can't be made to ignore the RGB color
components, set them to 1 instead of 0. This gives the correct results
- disabled GL_EXT_texture_env_add on Savage3D/MX/IX
- map both GL_CLAMP and GL_CLAMP_TO_EDGE to hardware mode "wrap". It doesn't
match either mode exactly by the spec, so we should either fall back on both
or none. I chose the latter.
- fall back to software when s and t wrapping modes differ (hardware has only
one bit for both)
|
|
|
|
compile-time-warnings on x86-64. Not tested on x86-64 yet, but a good
thing in general, so I'm giving up waiting for feedback. See also
https://bugs.freedesktop.org/show_bug.cgi?id=2924.
|
|
the use of hardware texture format I8 as it produces garbage at least on
ProSavageDDR.
|
|
|
|
|
|
set. Otherwise the 3D engine may still be using a texture that is
being overwritten. In order to minimize the cost of waiting, timestamp
textures only when needed: when a texture image changes, when a
different texture is bound to a texture unit or when a texture unit is
disabled. (Used to be after flushing every command buffer.)
|
|
not part of the DDX->DRI interface and should never have been in this
file in the first place.
|
|
kept around for reference. Now that command DMA is implemented in the
DRM module they are no longer needed.
|
|
- Flush before using a new texture or before disabling a texture unit,
because savageFlushCmdBuf can only update the timestamp of the last
used texture. This fixes corruption in quake2 with single-textured
lighting.
|
|
if it didn't change. This seems to flush texture cashes. Fixes
multi-textured lighting in quake2.
|
|
ahead of the CPU for more parallelism of CPU and GPU. Unfortunately
there seems to be some broken hardware (like my ProSavageDDR) on which
status register updates are delayed sometimes. This leads to very
jerky animation if the hardware can buffer more than the current
frame. A new option "sync_frames" can be used as a remedy on such
broken hardware.
|
|
new DRM version 2.4 and command DMA.
- Flush less.
- Bumped the driver date.
|
|
changed. Other drivers don't need to do this because they're swapping
modified textures out of texture memory, which implies a timestamp
update.
|
|
heap aging, similar to the way it's done in the i810 and i855 drivers.
This avoids idling the engine on every texture upload.
|
|
Add missing per-primitive fallback for polygon stipple.
|
|
Another small tweak to subtile upload (mostly cosmetic).
|
|
|
|
vertically incomplete tiles, but only if the color depth is
32bpp. Nobody said this was supposed to be logical!
|
|
|
|
|
|
- Flush buffered primitives before changing scissors.
- Require Savage DRM version 2.1.0, so that the broken single-cliprect
code can finally rest in peace.
- Removed some more dead code.
|
|
will be used. Implemented this option in the Savage driver. On my
ProSavageDDR uploads to AGP memory are about 1.5 times as fast as
uploads to card memory. On non-IGP hardware the difference may be even
bigger. Now mplayer -gl is getting really usable.
|
|
about a factor 1.3.
|
|
|
|
|
|
|
|
|
|
* When glTexSubImageND is used, track the set of changed tiles in a
bit vector and upload only dirty tiles later. This should improve
the performance of dynamic light maps and gl movie player plugins.
* Renamed debug item "lru" to "tex". Indicate which levels are
uploaded completely or partially.
|
|
|
|
buffered visuals. Also don't change span draw/read buffers in
savageDDDrawBuffer. Now glean's polygon offset test works (and PASSes
with fixed point depth buffer) with single buffered visuals without
winding up in an infinite loop.
|
|
range used by the savage driver by negating ctx->MRD broke polygon
offsets with software fallbacks. This one adds a REVERSE_DEPTH parameter
to t_dd_tritmp.h (defaults to 0) that allows reversing polygon offsets
for hardware rendering but not for software fallbacks. For software
fallbacks depth values are reversed after polygon offsets have been
applied by the depth span functions.
|
|
|
|
* Remove some unused (mostly empty) functions
* Added context parameter to WAIT_IDLE_EMPTY[_LOCKED] for consistency
* Added debug messages to WAIT_IDLE_EMPTY[_LOCKED]
* Don't flush empty command buffers
|
