profile/ivi/mesa.git
15 years agodemos: add LDFLAGS for corender target, bug 20844
Brian Paul [Tue, 24 Mar 2009 22:06:27 +0000 (16:06 -0600)]
demos: add LDFLAGS for corender target, bug 20844

15 years agoprogs/vp: add pointsize + clamp test
Keith Whitwell [Tue, 24 Mar 2009 17:50:36 +0000 (17:50 +0000)]
progs/vp: add pointsize + clamp test

15 years agomesa/st: bump gallium version to 0.3
Keith Whitwell [Tue, 24 Mar 2009 16:38:07 +0000 (16:38 +0000)]
mesa/st: bump gallium version to 0.3

To distinguish from the -0.2 version still being maintained on the
gallium-mesa-7.4 branch.  There are already greater interface changes
between these two branches than there were between -0.2 and -0.1.

Also stop injecting Tungsten into the vendor string - the Gallium in
the renderer string should be sufficient.

15 years agoprogs/trivial: add fflushes for cygwin
Keith Whitwell [Tue, 24 Mar 2009 16:35:29 +0000 (16:35 +0000)]
progs/trivial: add fflushes for cygwin

15 years agoi965: fix point rasterization when rendering to FBO
Robert Ellison [Tue, 24 Mar 2009 15:53:22 +0000 (09:53 -0600)]
i965: fix point rasterization when rendering to FBO

The FBO pixel coordinate system, with (0,0) as the
upper-left pixel, is inverted in Y compared to the
normal OpenGL pixel coordinate system, which has
(0,0) as its lower-left pixel.

Viewport and polygon stipple are sensitive to this
inversion; so is point rasterization.  The basic
fix is simple: when rendering to an FBO, instead
of the normal RASTRULE_UPPER_RIGHT that's
appropriate for OpenGL windows, use the Y inversion
RASTRULE_LOWER_RIGHT.

Unfortunately, current Intel documentation has this
value listed as "Reserved, but not seen as useful".
It does work on at least some i965-class devices,
though; and the worst that could happen if an
older device didn't support it would be incorrect
point rasterization to FBOs, which is what happens
already, so this fix is at least no worse than what
happens presently, and is better for some (and possibly
all) i965-class devices.

15 years agoprogs/trivial: draw non-interleaved arrays out of one vbo
Keith Whitwell [Tue, 24 Mar 2009 15:27:21 +0000 (15:27 +0000)]
progs/trivial: draw non-interleaved arrays out of one vbo

15 years agoprogs/tests: Build texcompress2 with scons
Jakob Bornecrantz [Sun, 22 Mar 2009 07:46:18 +0000 (08:46 +0100)]
progs/tests: Build texcompress2 with scons

15 years agoi965: Fix glFrontFacing in twoside GLSL demo.
Eric Anholt [Mon, 23 Mar 2009 23:29:31 +0000 (16:29 -0700)]
i965: Fix glFrontFacing in twoside GLSL demo.

This also cuts instructions by just using the existing bit in the payload
rather than computing it from the determinant in the SF unit and passing it
as a varying down to the WM.  Something still goes wrong with getting the
backface color right, but a simpler shader appears to get the right result.

15 years agoi965: Fix fog coordinate g,b,a values when glFrontFacing isn't used.
Eric Anholt [Tue, 24 Mar 2009 05:35:03 +0000 (22:35 -0700)]
i965: Fix fog coordinate g,b,a values when glFrontFacing isn't used.

Previously, we would sample (f,glFrontFacing,undef,undef) instead of the
(f,0,0,1) that fragment.fogcoord is supposed to return.  Due to
glFrontFacing's presence in FOGC.y, we'll still give bad results there when
glFrontFacing is used.

Bug #19122, piglit testcase fp-fog.

15 years agoi965: Clean up a bit of mess with unneeded variables in emit_interp.
Eric Anholt [Tue, 24 Mar 2009 05:30:12 +0000 (22:30 -0700)]
i965: Clean up a bit of mess with unneeded variables in emit_interp.

15 years agoi965: Fix trailing "d" in debug output for 3DSTATE_VERTEX_ELEMENTS.
Eric Anholt [Mon, 23 Mar 2009 20:48:24 +0000 (13:48 -0700)]
i965: Fix trailing "d" in debug output for 3DSTATE_VERTEX_ELEMENTS.

15 years agoi965: Fix occlusion query when no other WM state updates occur.
Eric Anholt [Mon, 23 Mar 2009 19:29:02 +0000 (12:29 -0700)]
i965: Fix occlusion query when no other WM state updates occur.

Turns out that XXX comment was important.  We weren't flagging the WM to
re-update with the statistics enable, so we got zeroes out of our query.

Bug #20740, fixes piglit occlusion_query test.

Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agovpglsl: add missing SConscript
Keith Whitwell [Mon, 23 Mar 2009 19:01:40 +0000 (19:01 +0000)]
vpglsl: add missing SConscript

15 years agodraw: update aa points shader comment
Keith Whitwell [Mon, 23 Mar 2009 18:38:11 +0000 (18:38 +0000)]
draw: update aa points shader comment

15 years agoprogs/fp: pass texcoord to triangle, add a test shader
Keith Whitwell [Mon, 23 Mar 2009 18:37:33 +0000 (18:37 +0000)]
progs/fp: pass texcoord to triangle, add a test shader

15 years agogallium: Remove remnants of reference counting internals outside of p_refcnt.h.
Michel Dänzer [Mon, 23 Mar 2009 17:03:13 +0000 (18:03 +0100)]
gallium: Remove remnants of reference counting internals outside of p_refcnt.h.

15 years agodebug: Add function for writing transfers to files
Jakob Bornecrantz [Sat, 21 Mar 2009 13:23:04 +0000 (14:23 +0100)]
debug: Add function for writing transfers to files

15 years agotrace: Fix args to buffer write
Jakob Bornecrantz [Fri, 20 Mar 2009 22:06:05 +0000 (23:06 +0100)]
trace: Fix args to buffer write

15 years agoutil: Add a new macro for testing empty lists.
José Fonseca [Mon, 23 Mar 2009 12:05:07 +0000 (12:05 +0000)]
util: Add a new macro for testing empty lists.

15 years agosoftpipe: reformatting, comments, minor clean-ups
Brian Paul [Mon, 23 Mar 2009 00:11:12 +0000 (18:11 -0600)]
softpipe: reformatting, comments, minor clean-ups

15 years agotgsi: minor comments
Brian Paul [Mon, 23 Mar 2009 00:10:10 +0000 (18:10 -0600)]
tgsi: minor comments

15 years agotgsi: Document KIL, KILP instructions.
Michal Krol [Sat, 21 Mar 2009 11:18:09 +0000 (12:18 +0100)]
tgsi: Document KIL, KILP instructions.

15 years agogallium: remove remaining references to origin_lower_left
Keith Whitwell [Sat, 21 Mar 2009 10:37:33 +0000 (10:37 +0000)]
gallium: remove remaining references to origin_lower_left

15 years agost: Silence compiler warnings.
Michal Krol [Sat, 21 Mar 2009 10:46:54 +0000 (11:46 +0100)]
st: Silence compiler warnings.

15 years agogallium: remove use of origin_lower_left
Brian Paul [Sat, 21 Mar 2009 02:25:34 +0000 (20:25 -0600)]
gallium: remove use of origin_lower_left

This was used to indicate OpenGL's lower-left origin for fragment window
coordinates for polygon stipple and gl_FragCoord.

Now:
  - fragment coordinate origin is always upper-left corner
  - GL polygon stipple is inverted and shifted before given to gallium
  - GL fragment programs that use INPUT[WPOS] are modified to use an
    inverted window coord which is placed in a temp register.

Note: the origin_lower_left field still exists in pipe_rasterizer_state.
Remove it when all the drivers, etc. no longer reference it.

15 years agomesa: add new internal state var for window size
Brian Paul [Sat, 21 Mar 2009 02:15:58 +0000 (20:15 -0600)]
mesa: add new internal state var for window size

Actually, window width - 1, height - 1

15 years agomesa: linear scan register allocation for shader programs
Brian Paul [Fri, 20 Mar 2009 23:08:30 +0000 (17:08 -0600)]
mesa: linear scan register allocation for shader programs

This is a check-point commit; not turned on yet.

Use the linear scan register allocation algorithm to re-allocate temporary
registers.  This is done by computing the live intervals for registers and
reallocating temps with that information.

For some shaders this dramatically reduces the number of temp registers
needed.

For the time being we give up on a few cases such as relative-indexed temps
and subroutine calls (but we inline most GLSL functions anyway).

15 years agodocs: updated Mesa extension enum info
Brian Paul [Fri, 20 Mar 2009 15:20:53 +0000 (09:20 -0600)]
docs: updated Mesa extension enum info

15 years agor300-gallium: Misspelled macro name.
Corbin Simpson [Fri, 20 Mar 2009 21:47:49 +0000 (14:47 -0700)]
r300-gallium: Misspelled macro name.

*pulls paper bag down over head*

15 years agor300-gallium: Put r300_cs_inlines to bed.
Corbin Simpson [Fri, 20 Mar 2009 07:48:53 +0000 (00:48 -0700)]
r300-gallium: Put r300_cs_inlines to bed.

Guess it was a mistake in the first place. Oops.

15 years agor300-gallium: Simplify/neaten up packet3.
Corbin Simpson [Fri, 20 Mar 2009 07:43:29 +0000 (00:43 -0700)]
r300-gallium: Simplify/neaten up packet3.

Deck chairs on the Hindenburg. :3

15 years agor300-gallium: Clean up surface_fill, prep for surface_copy code.
Corbin Simpson [Fri, 20 Mar 2009 07:35:38 +0000 (00:35 -0700)]
r300-gallium: Clean up surface_fill, prep for surface_copy code.

15 years agor300-gallium: Properly offset scissors.
Corbin Simpson [Fri, 20 Mar 2009 07:15:03 +0000 (00:15 -0700)]
r300-gallium: Properly offset scissors.

As per r300_reg, classic Mesa, and xf86-video-ati.

15 years agor300-gallium: Clean up r300_swtcl_emit.
Corbin Simpson [Fri, 20 Mar 2009 03:36:59 +0000 (20:36 -0700)]
r300-gallium: Clean up r300_swtcl_emit.

Some compile warnings, some statements without effect.

15 years agor300-gallium: Clean up some emit, and some state handlers.
Corbin Simpson [Fri, 20 Mar 2009 03:32:08 +0000 (20:32 -0700)]
r300-gallium: Clean up some emit, and some state handlers.

15 years agor300-gallium: A bit more invariant state.
Corbin Simpson [Thu, 19 Mar 2009 19:29:03 +0000 (12:29 -0700)]
r300-gallium: A bit more invariant state.

15 years agogallium: Explain what happens if buffer_flush_mapped_range isn't called.
José Fonseca [Wed, 18 Mar 2009 16:54:25 +0000 (16:54 +0000)]
gallium: Explain what happens if buffer_flush_mapped_range isn't called.

15 years agogallium/util: add upload manager helper module
Keith Whitwell [Fri, 20 Mar 2009 15:08:59 +0000 (15:08 +0000)]
gallium/util:  add upload manager helper module

Add a module that will manage uploading and coalescing multiple
user-buffers, malloc-buffers and other random data that doesn't
happen to be in a GPU buffer already.  The module stuffs multiple
little uploads into larger GPU buffers to reduce create/destroy
overheads, etc.

15 years agoFix DRI2 accelerated EXT_texture_from_pixmap with GL_RGB format.
Eric Anholt [Wed, 18 Mar 2009 19:07:09 +0000 (12:07 -0700)]
Fix DRI2 accelerated EXT_texture_from_pixmap with GL_RGB format.

This requires upgrading the interface so that the argument to
glXBindTexImageEXT isn't just dropped on the floor.  Note that this only
fixes the accelerated path on Intel, as Mesa's texture format support is
missing x8r8g8b8 support (right now, GL_RGB textures get uploaded as a8r8gb8,
but in this case we're not doing the upload so we can't really work around it
that way).

Fixes bugs with compositors trying to use shaders that use alpha channels, on
windows without a valid alpha channel.  Bug #19910 and likely others as well.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

15 years agoprogs/wgl: Send a resize message after context have been created.
Michal Krol [Fri, 20 Mar 2009 15:48:54 +0000 (16:48 +0100)]
progs/wgl: Send a resize message after context have been created.

15 years agostw: Do not err on nil context handle in MakeCurrent().
Michal Krol [Fri, 20 Mar 2009 15:36:23 +0000 (16:36 +0100)]
stw: Do not err on nil context handle in MakeCurrent().

15 years agoprogs/wgl: Create GL context in a thread that actually uses it.
Michal Krol [Fri, 20 Mar 2009 15:16:12 +0000 (16:16 +0100)]
progs/wgl: Create GL context in a thread that actually uses it.

15 years agoprogs/wgl: Make context current to get GL_RENDERER string.
Michal Krol [Fri, 20 Mar 2009 15:13:39 +0000 (16:13 +0100)]
progs/wgl: Make context current to get GL_RENDERER string.

15 years agostw: Use u_handle_table to maintain context list.
Michal Krol [Fri, 20 Mar 2009 14:45:00 +0000 (15:45 +0100)]
stw: Use u_handle_table to maintain context list.

15 years agowinsys/gdi: Init state tracker's per-thread data.
Michal Krol [Fri, 20 Mar 2009 12:06:28 +0000 (13:06 +0100)]
winsys/gdi: Init state tracker's per-thread data.

15 years agostw: Keep per-thread storage for current context and pixel format.
Michal Krol [Fri, 20 Mar 2009 12:05:51 +0000 (13:05 +0100)]
stw: Keep per-thread storage for current context and pixel format.

15 years agoslang: Use _mesa_snprintf() wrapper.
Michal Krol [Fri, 20 Mar 2009 11:19:40 +0000 (12:19 +0100)]
slang: Use _mesa_snprintf() wrapper.

15 years agoslang: initialize the context
Alan Hourihane [Thu, 19 Mar 2009 22:38:01 +0000 (22:38 +0000)]
slang: initialize the context

15 years agoslang: support uniform arrays
Alan Hourihane [Thu, 19 Mar 2009 22:06:10 +0000 (22:06 +0000)]
slang: support uniform arrays

15 years agotrace: Formalize on a standard for data size pair
Jakob Bornecrantz [Thu, 19 Mar 2009 15:11:05 +0000 (16:11 +0100)]
trace: Formalize on a standard for data size pair

15 years agoglslcompiler: added new options to override debug/optimization pragmas
Brian Paul [Thu, 19 Mar 2009 16:26:50 +0000 (10:26 -0600)]
glslcompiler: added new options to override debug/optimization pragmas

15 years agoglsl: change GLSL #pragma initialization
Brian Paul [Thu, 19 Mar 2009 16:25:24 +0000 (10:25 -0600)]
glsl: change GLSL #pragma initialization

Initialize the shader's pragma settings before calling the compiler.
Added pragma "Ignore" fields to allow overriding the #pragma directives found
in shader source code.

15 years agoegl: fix comment
Alan Hourihane [Thu, 19 Mar 2009 15:39:01 +0000 (15:39 +0000)]
egl: fix comment

15 years agoegl: try harder for eglGetProcAddress()
Alan Hourihane [Thu, 19 Mar 2009 15:34:45 +0000 (15:34 +0000)]
egl: try harder for eglGetProcAddress()

15 years agoglsl: when debug pragma is on, emit comments about function calls/inlines
Brian Paul [Thu, 19 Mar 2009 15:26:20 +0000 (09:26 -0600)]
glsl: when debug pragma is on, emit comments about function calls/inlines

BTW, the debug pragma syntax is "#pragma debug(on)"

15 years agomesa: check renderbuffer is defined before use.
Alan Hourihane [Thu, 19 Mar 2009 13:58:04 +0000 (13:58 +0000)]
mesa: check renderbuffer is defined before use.

15 years agost/gl: Only transfer needed parts in st_TexSubimage
Jakob Bornecrantz [Wed, 18 Mar 2009 18:10:32 +0000 (19:10 +0100)]
st/gl: Only transfer needed parts in st_TexSubimage

15 years agoslang: if/else/break & if/else/continue work for unrolled loops
Alan Hourihane [Thu, 19 Mar 2009 10:16:28 +0000 (10:16 +0000)]
slang: if/else/break & if/else/continue work for unrolled loops

15 years agowgl: Add an `-s' option to wglthreads to force single-threaded operation.
Michal Krol [Thu, 19 Mar 2009 10:16:01 +0000 (11:16 +0100)]
wgl: Add an `-s' option to wglthreads to force single-threaded operation.

15 years agoswrast: Silence compiler warnings, give better structure to the code as a result.
Michal Krol [Thu, 19 Mar 2009 09:39:57 +0000 (10:39 +0100)]
swrast: Silence compiler warnings, give better structure to the code as a result.

15 years agotgsi: Document vs_2_0 instruction set operations.
Michal Krol [Thu, 19 Mar 2009 08:52:50 +0000 (09:52 +0100)]
tgsi: Document vs_2_0 instruction set operations.

15 years agotgsi: Document vs_1_1 instruction set operations.
Michal Krol [Thu, 19 Mar 2009 08:49:38 +0000 (09:49 +0100)]
tgsi: Document vs_1_1 instruction set operations.

15 years agotgsi: Begin documenting ps_2_x instruction set operations.
Michal Krol [Thu, 19 Mar 2009 08:44:51 +0000 (09:44 +0100)]
tgsi: Begin documenting ps_2_x instruction set operations.

15 years agor300-gallium: Emit viewport state.
Corbin Simpson [Wed, 18 Mar 2009 22:17:38 +0000 (15:17 -0700)]
r300-gallium: Emit viewport state.

Note that this will break you, hard, if you're not using RADEON_NO_TCL.
I really need to start vertex shaders soon.

15 years agor300-gallium: Viewport state storage.
Corbin Simpson [Wed, 18 Mar 2009 20:37:59 +0000 (13:37 -0700)]
r300-gallium: Viewport state storage.

15 years agor300-gallium: Fixup registers for viewport state.
Corbin Simpson [Wed, 18 Mar 2009 20:37:08 +0000 (13:37 -0700)]
r300-gallium: Fixup registers for viewport state.

15 years agomesa: use the IROUND() macro in pixel packing code
Brian Paul [Wed, 18 Mar 2009 21:39:11 +0000 (15:39 -0600)]
mesa: use the IROUND() macro in pixel packing code

It turns out some tests are sensitive to rounding vs. truncating when
converting float color values to integers in glReadPixels().  In particular,
this matters when the destination format is 5/6/5 or 4/4/4/4, etc.

15 years agoslang: redo the last commit for if/break & if/continue tests as it
Alan Hourihane [Wed, 18 Mar 2009 21:42:14 +0000 (21:42 +0000)]
slang: redo the last commit for if/break & if/continue tests as it
wasn't good enough for deeply nested if's.

15 years agoRevert "slang: if we detect an if/break or if/continue within a loop and we're"
Alan Hourihane [Wed, 18 Mar 2009 21:40:03 +0000 (21:40 +0000)]
Revert "slang: if we detect an if/break or if/continue within a loop and we're"

This reverts commit 752296b8f311c5e3844f3ce89d17ba57224ce5ba.

15 years agoslang: if we detect an if/break or if/continue within a loop and we're
Alan Hourihane [Wed, 18 Mar 2009 21:16:35 +0000 (21:16 +0000)]
slang: if we detect an if/break or if/continue within a loop and we're
trying to unroll, bail, and fallback to doing the real loop.

15 years agost: call _glapi_check_multithread from st_make_current
Keith Whitwell [Wed, 18 Mar 2009 19:14:45 +0000 (19:14 +0000)]
st: call _glapi_check_multithread from st_make_current

This function is called from many OS-dependent versions of MakeCurrent.

Move the check for multithreading to this central location to avoid
having to make this check from all the callers.

15 years agostw: Fix makeCurrent.
Thomas Hellstrom [Wed, 18 Mar 2009 11:28:31 +0000 (12:28 +0100)]
stw: Fix makeCurrent.

Flush if we change current context.

Signed-off-by: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
15 years agoprogs: Build the wgl examples with scons.
José Fonseca [Wed, 18 Mar 2009 16:37:20 +0000 (16:37 +0000)]
progs: Build the wgl examples with scons.

15 years agoprogs/trivial: Make quad-tex-sub repeat patterns
Jakob Bornecrantz [Wed, 18 Mar 2009 16:24:40 +0000 (17:24 +0100)]
progs/trivial: Make quad-tex-sub repeat patterns

15 years agotests: remove/update old comments
Brian Paul [Wed, 18 Mar 2009 16:23:31 +0000 (10:23 -0600)]
tests: remove/update old comments

15 years agost: need to free/realloc pipe texture when we're handed image level > last_level
Brian Paul [Wed, 18 Mar 2009 16:22:05 +0000 (10:22 -0600)]
st: need to free/realloc pipe texture when we're handed image level > last_level

Fixes progs/tests/tri-fbo-tex-mip.c

15 years agost: clean up some if-statment code
Brian Paul [Wed, 18 Mar 2009 16:21:05 +0000 (10:21 -0600)]
st: clean up some if-statment code

15 years agosoftpipe: more texture transfer fixes.
Brian Paul [Wed, 18 Mar 2009 16:17:48 +0000 (10:17 -0600)]
softpipe: more texture transfer fixes.

Need to null-out pointers after freeing transfer objects.
Fix mix-ups between tc->transfer and tc->tex_trans fields.

15 years agosoftpipe: need to set tc->transfer=NULL after destroying the transfer object
Brian Paul [Wed, 18 Mar 2009 15:55:00 +0000 (09:55 -0600)]
softpipe: need to set tc->transfer=NULL after destroying the transfer object

This fixes a number of crashes/regressions in programs such as lodbias.c,
mipmap_limits.c, etc.

15 years agost: update texture comments
Brian Paul [Wed, 18 Mar 2009 15:40:37 +0000 (09:40 -0600)]
st: update texture comments

15 years agoprogs/trivial: Add quad-tex-sub
Jakob Bornecrantz [Wed, 18 Mar 2009 16:15:40 +0000 (17:15 +0100)]
progs/trivial: Add quad-tex-sub

   Tests glTexSubImage

15 years agoUse the specified X11 headers for xlib mesa
Dan Nicholson [Wed, 18 Mar 2009 06:33:04 +0000 (23:33 -0700)]
Use the specified X11 headers for xlib mesa

The xlib build was using the system's Xlib headers or bombing if they
weren't available.

15 years agotests: add simplest mipgen test
Keith Whitwell [Wed, 18 Mar 2009 14:01:47 +0000 (14:01 +0000)]
tests: add simplest mipgen test

15 years agopipe/atomic: dont use ms interlock calls from gcc
Keith Whitwell [Wed, 18 Mar 2009 13:27:53 +0000 (13:27 +0000)]
pipe/atomic: dont use ms interlock calls from gcc

15 years agoutil: avoid clashprone UINT, UBYTE enum values
Keith Whitwell [Wed, 18 Mar 2009 11:35:58 +0000 (11:35 +0000)]
util: avoid clashprone UINT, UBYTE enum values

15 years agodraw: use AOS_ERROR rather than clash-prone ERROR() macro
Keith Whitwell [Wed, 18 Mar 2009 11:35:26 +0000 (11:35 +0000)]
draw: use AOS_ERROR rather than clash-prone ERROR() macro

15 years agopipe/atomic: clean up #ifdef maze
Keith Whitwell [Wed, 18 Mar 2009 11:29:01 +0000 (11:29 +0000)]
pipe/atomic: clean up #ifdef maze

15 years agoprogs/tests: Update mipmap_limits to show image and colors
Jakob Bornecrantz [Wed, 18 Mar 2009 12:29:03 +0000 (13:29 +0100)]
progs/tests: Update mipmap_limits to show image and colors

15 years agoxlib st: Fix makeCurrent.
Thomas Hellstrom [Wed, 18 Mar 2009 10:52:24 +0000 (11:52 +0100)]
xlib st: Fix makeCurrent.

Flush if we change context.
Also reinstate the old optimization of doing nothing if
nothing changes.

Signed-off-by: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
15 years agoprogs/trivial: Add tri-fbo-tex-mip
Jakob Bornecrantz [Tue, 17 Mar 2009 14:04:07 +0000 (15:04 +0100)]
progs/trivial: Add tri-fbo-tex-mip

   Tests rendering to mip levels other then first

15 years agonouveau: rewrite winsys in terms of drm_api, support dri2 state tracker
Ben Skeggs [Tue, 17 Mar 2009 22:22:35 +0000 (08:22 +1000)]
nouveau: rewrite winsys in terms of drm_api, support dri2 state tracker

drm_api is a set of hooks used by the dri2 state tracker, this wraps our
dri1 code around the same set of hooks.

Currently the dri2 build will produce nouveau_dri2.so which you'll need
to install as nouveau_dri.so if you wish to try it.  The dri2 state
tracker doesn't make it easy for a driver to support both paths in the
same binary.

15 years agost/dri2: use pipe_screen.is_format_supported to choose depth/stencil format
Ben Skeggs [Tue, 17 Mar 2009 22:21:32 +0000 (08:21 +1000)]
st/dri2: use pipe_screen.is_format_supported to choose depth/stencil format

15 years agor300-gallium: Move all unsorted state into invariant state.
Corbin Simpson [Tue, 17 Mar 2009 20:11:55 +0000 (13:11 -0700)]
r300-gallium: Move all unsorted state into invariant state.

Gotta just slowly whittle this down.

15 years agor300-gallium: Emit invariant state, no matter what.
Corbin Simpson [Tue, 17 Mar 2009 20:10:41 +0000 (13:10 -0700)]
r300-gallium: Emit invariant state, no matter what.

It's called "invariant" for a reason. :3

15 years agor300-gallium: Quick little cleanup of surface_fill state.
Corbin Simpson [Tue, 17 Mar 2009 20:07:44 +0000 (13:07 -0700)]
r300-gallium: Quick little cleanup of surface_fill state.

15 years agotgsi: Restore statement ordering.
José Fonseca [Tue, 17 Mar 2009 19:50:27 +0000 (19:50 +0000)]
tgsi: Restore statement ordering.

15 years agor300-gallium: Debugging for the more sensitive card registers.
Corbin Simpson [Tue, 17 Mar 2009 18:49:29 +0000 (11:49 -0700)]
r300-gallium: Debugging for the more sensitive card registers.

These are nearly always the cause of hardlocks, so let's dump them.

15 years agor300-gallium: r500-fs: Fixup immediate->constant counting a bit.
Corbin Simpson [Tue, 17 Mar 2009 18:48:55 +0000 (11:48 -0700)]
r300-gallium: r500-fs: Fixup immediate->constant counting a bit.

15 years agor300-gallium: More query stuff.
Corbin Simpson [Tue, 17 Mar 2009 18:42:13 +0000 (11:42 -0700)]
r300-gallium: More query stuff.

Should work, but doesn't. At least it doesn't hardlock.

15 years agoswrast: use better _swrast_compute_lambda() function
Brian Paul [Tue, 17 Mar 2009 16:28:32 +0000 (10:28 -0600)]
swrast: use better _swrast_compute_lambda() function

The MAX-based function can produce values that are non-monotonic for a span
which causes glitches in texture filtering.  The sqrt-based one avoids that.

This is perhaps slightly slower than before, but the difference
probably isn't noticable given we're doing software mipmap filtering.

Issue reported by Nir Radian <nirr@horizonsemi.com>