mesa.git
13 years ago[g3dvl] cleanup and documentation
Christian König [Wed, 16 Mar 2011 22:09:52 +0000 (23:09 +0100)]
[g3dvl] cleanup and documentation

13 years ago[g3dvl] move idct out of mc code
Christian König [Wed, 9 Mar 2011 22:40:08 +0000 (23:40 +0100)]
[g3dvl] move idct out of mc code

iDCT and MC now look good, but sp_video_context is a total mess

13 years ago[g3dvl] start to cleanup the mess
Christian König [Tue, 8 Mar 2011 20:30:33 +0000 (21:30 +0100)]
[g3dvl] start to cleanup the mess

Move the vertex buffer out of the mc code

13 years ago[g3dvl] use a single vertex buffer for both idct and mc
Christian König [Tue, 8 Mar 2011 17:34:05 +0000 (18:34 +0100)]
[g3dvl] use a single vertex buffer for both idct and mc

13 years agor600g: set start instance correctly
Christian König [Tue, 8 Mar 2011 15:48:39 +0000 (16:48 +0100)]
r600g: set start instance correctly

13 years ago[g3dvl] use instanced drawing to reduce the vertex buffer payload
Christian König [Sat, 5 Mar 2011 15:36:38 +0000 (16:36 +0100)]
[g3dvl] use instanced drawing to reduce the vertex buffer payload

13 years agoMerge remote branch 'origin/master' into pipe-video
Christian König [Sat, 5 Mar 2011 14:46:56 +0000 (15:46 +0100)]
Merge remote branch 'origin/master' into pipe-video

13 years agonv50: support the InstanceID system value
Christoph Bumiller [Sat, 5 Mar 2011 13:41:08 +0000 (14:41 +0100)]
nv50: support the InstanceID system value

13 years agor600g: simplify instance addr calculation
Christian König [Sat, 5 Mar 2011 12:40:55 +0000 (13:40 +0100)]
r600g: simplify instance addr calculation

Use MULHI_UINT instead of the more complex
INT_TO_FLT->MUL->TRUNC->FLT_TO_INT

13 years agonvc0: Update SConscript.
Vinson Lee [Sat, 5 Mar 2011 01:15:21 +0000 (17:15 -0800)]
nvc0: Update SConscript.

13 years agonv50: Update SConscript.
Vinson Lee [Sat, 5 Mar 2011 01:10:57 +0000 (17:10 -0800)]
nv50: Update SConscript.

13 years agor600g: fix fragment shader size calculation
Christian König [Sat, 5 Mar 2011 00:46:31 +0000 (01:46 +0100)]
r600g: fix fragment shader size calculation

bc.ndw is altered in r600_bc_build, respect that
in fragment shader size calculation.

13 years agoglsl: Process redeclarations before initializers
Ian Romanick [Sat, 5 Mar 2011 00:15:20 +0000 (16:15 -0800)]
glsl: Process redeclarations before initializers

If an array redeclaration includes an initializer, the initializer
would previously be dropped on the floor.  Instead, directly apply the
initializer to the correct ir_variable instance and append the
generated instructions.

Fixes bugzilla #34374 and piglit tests glsl-{vs,fs}-array-redeclaration.

NOTE: This is a candidate for stable release branches.  0292ffb8 and
8e6cb9fe are also necessary.

13 years agoglsl: Refactor AST-to-HIR code handling variable initializers
Ian Romanick [Fri, 4 Mar 2011 23:29:33 +0000 (15:29 -0800)]
glsl: Refactor AST-to-HIR code handling variable initializers

13 years agoglsl: Refactor AST-to-HIR code handling variable redeclarations
Ian Romanick [Fri, 4 Mar 2011 23:28:40 +0000 (15:28 -0800)]
glsl: Refactor AST-to-HIR code handling variable redeclarations

13 years agonv50,nvc0: get format desc for TIC entry from sampler view format
Christoph Bumiller [Fri, 4 Mar 2011 21:17:36 +0000 (22:17 +0100)]
nv50,nvc0: get format desc for TIC entry from sampler view format

Fixes piglit/tex-srgb.

13 years agonv50,nvc0: share sampler state creation
Christoph Bumiller [Fri, 4 Mar 2011 21:15:17 +0000 (22:15 +0100)]
nv50,nvc0: share sampler state creation

13 years agonv50,nvc0: update the format tables
Christoph Bumiller [Fri, 4 Mar 2011 22:54:42 +0000 (23:54 +0100)]
nv50,nvc0: update the format tables

Removed sampler view support for USCALED/SSCALED, the texture unit
refuses to convert to non-normalized float. The enums are treated
like UNORM.

Removed duplicate format related headers.

13 years agonvc0: use m2mf for resource_copy_region if formats are equal
Christoph Bumiller [Fri, 4 Mar 2011 17:17:05 +0000 (18:17 +0100)]
nvc0: use m2mf for resource_copy_region if formats are equal

Which is always the case, but we'll keep the 2D engine blitter
nonetheless.

13 years agonv50,nvc0: fix texture layer issues
Christoph Bumiller [Fri, 4 Mar 2011 23:05:14 +0000 (00:05 +0100)]
nv50,nvc0: fix texture layer issues

13 years agoi915g: Use tgsi_info from fragment shader instead
Jakob Bornecrantz [Fri, 4 Mar 2011 22:57:16 +0000 (23:57 +0100)]
i915g: Use tgsi_info from fragment shader instead

13 years agoi915g: use passthough shader for empty fragment programs
Daniel Vetter [Fri, 4 Mar 2011 22:44:39 +0000 (23:44 +0100)]
i915g: use passthough shader for empty fragment programs

The hw doesn't like it - demos/shadowtex is broken. The emitted shader
isn't totally empty though, the depth write fixup gets emitted instead.
Maybe that one is somewhat fishy, too?

Idea for this patch from Jakob Bornecrantz.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
13 years agoegl_dri2: Fix incompatible vfunc-pointer warning
Benjamin Franzke [Thu, 3 Mar 2011 10:22:09 +0000 (11:22 +0100)]
egl_dri2: Fix incompatible vfunc-pointer warning

13 years agoegl/wayland: Move wayland-egl into a subdir
Benjamin Franzke [Thu, 3 Mar 2011 10:05:03 +0000 (11:05 +0100)]
egl/wayland: Move wayland-egl into a subdir

This hopefully fixes wayland-egl's dependency
resolution for autogenerated wayland-drm headers.

13 years agoi965: Apply a workaround for the Ironlake "vertex flashing".
Eric Anholt [Tue, 22 Feb 2011 07:46:52 +0000 (23:46 -0800)]
i965: Apply a workaround for the Ironlake "vertex flashing".

This is an awful hack and will hurt performance on Ironlake, but we're
at a loss as to what's going wrong otherwise.  This is the only common
variable we've found that avoids the problem on 4 applications
(CelShading, gnome-shell, Pill Popper, and my GLSL demo), while other
variables we've tried appear to only be confounding.  Neither the
specifications nor the hardware team have been able to provide any
enlightenment, despite much searching.

https://bugs.freedesktop.org/show_bug.cgi?id=29172
Tested by: Chris Lord <chris@linux.intel.com> (Pill Popper)
Tested by: Ryan Lortie <desrt@desrt.ca> (gnome-shell)

13 years agor300g: preliminary implementation of clamping controls
Marek Olšák [Tue, 15 Feb 2011 23:50:25 +0000 (00:50 +0100)]
r300g: preliminary implementation of clamping controls

13 years agor300g: implement FP16 alpha test
Marek Olšák [Fri, 4 Mar 2011 16:28:44 +0000 (17:28 +0100)]
r300g: implement FP16 alpha test

13 years agor300g: implement blending for some of non-RGBA8 formats
Marek Olšák [Fri, 4 Mar 2011 15:39:51 +0000 (16:39 +0100)]
r300g: implement blending for some of non-RGBA8 formats

Blending is now fully supported with:
- R8_UNORM
- R8G8_UNORM
- B8G8R8A8_UNORM
- R16G16B16A16_FLOAT (r500-only)

Blending is partially supported (DST_ALPHA not working) with:
- L8A8_UNORM
- I8_UNORM
- B5G5R5A1_UNORM
- B10G10R10A2_UNORM

The other formats can't do blending.

13 years agodraw: Silence tgsi_emit_sse2 failed messages.
José Fonseca [Fri, 4 Mar 2011 16:29:13 +0000 (16:29 +0000)]
draw: Silence tgsi_emit_sse2 failed messages.

13 years agoMerge remote branch 'origin/master' into pipe-video
Christian König [Fri, 4 Mar 2011 16:15:43 +0000 (17:15 +0100)]
Merge remote branch 'origin/master' into pipe-video

13 years agotgsi: Disable SSE2 code generation.
José Fonseca [Fri, 4 Mar 2011 14:54:24 +0000 (14:54 +0000)]
tgsi: Disable SSE2 code generation.

It's broken now that tgsi_exec_machine::Inputs/Ouputs are pointers.

Temporary if anybody still cares about tgsi_sse2.c. Permanent otherwise.

13 years agoscons: Unbreak mingw cross compilation.
José Fonseca [Fri, 4 Mar 2011 14:29:06 +0000 (14:29 +0000)]
scons: Unbreak mingw cross compilation.

13 years agost/mesa: set PIPE_BIND_RENDER_TARGET for sRGB formats if UNORM is supported
Marek Olšák [Fri, 4 Mar 2011 05:38:13 +0000 (06:38 +0100)]
st/mesa: set PIPE_BIND_RENDER_TARGET for sRGB formats if UNORM is supported

Because the format can be changed to UNORM in a surface.

This fixes:
state_tracker/st_atom_framebuffer.c:163:update_framebuffer_state:
Assertion `framebuffer->cbufs[i]->texture->bind & (1 << 1)' failed.

13 years agoscons: Get glsl2 and glcpp programs building correctly.
José Fonseca [Fri, 4 Mar 2011 13:11:49 +0000 (13:11 +0000)]
scons: Get glsl2 and glcpp programs building correctly.

13 years agoglsl/glcpp: Use stdio.h instead of unistd.h.
José Fonseca [Fri, 4 Mar 2011 12:53:14 +0000 (12:53 +0000)]
glsl/glcpp: Use stdio.h instead of unistd.h.

13 years agoglsl: Define YY_NO_UNISTD_H on MSVC.
José Fonseca [Fri, 4 Mar 2011 12:49:55 +0000 (12:49 +0000)]
glsl: Define YY_NO_UNISTD_H on MSVC.

13 years agogallium: Define __func__ on MSVC.
José Fonseca [Fri, 4 Mar 2011 11:46:13 +0000 (11:46 +0000)]
gallium: Define __func__ on MSVC.

13 years agoMerge remote branch 'origin/nvc0'
Christoph Bumiller [Fri, 4 Mar 2011 10:02:10 +0000 (11:02 +0100)]
Merge remote branch 'origin/nvc0'

13 years agoi965: Fix extending VB packets
Chris Wilson [Thu, 3 Mar 2011 22:34:22 +0000 (22:34 +0000)]
i965: Fix extending VB packets

Computation of the delta of this array from the last had a silly little
bug and ignored any initial delta==0 causing grief in Nexuiz and
friends.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
13 years agoi965: Handle URB_FENCE erratum for Broadwater
Chris Wilson [Thu, 3 Mar 2011 18:15:03 +0000 (18:15 +0000)]
i965: Handle URB_FENCE erratum for Broadwater

There is a silicon bug which causes unpredictable behaviour if the
URB_FENCE command should cross a cache-line boundary. Pad before the
command to avoid such occurrences. As this command only applies to
gen4/5, do the fixup unconditionally as the specs do not actually state
for which chip it was fixed (and the cost is negligible)...

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
13 years agoi965: Align index to type size and flush if the type changes
Chris Wilson [Thu, 3 Mar 2011 15:56:39 +0000 (15:56 +0000)]
i965: Align index to type size and flush if the type changes

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
13 years agointel: Add couple of missing gen6 commands to decode
Chris Wilson [Sun, 27 Feb 2011 19:21:54 +0000 (19:21 +0000)]
intel: Add couple of missing gen6 commands to decode

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
13 years agoi965: Prevent using a zero sized (or of unknown type) vertex array
Chris Wilson [Thu, 3 Mar 2011 11:25:30 +0000 (11:25 +0000)]
i965: Prevent using a zero sized (or of unknown type) vertex array

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
13 years agor600g: disable tiling by default again.
Dave Airlie [Thu, 3 Mar 2011 22:42:33 +0000 (08:42 +1000)]
r600g: disable tiling by default again.

we still have a lot of corner cases that aren't working.

Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agotgsi: Update assert.
José Fonseca [Thu, 3 Mar 2011 19:15:24 +0000 (19:15 +0000)]
tgsi: Update assert.

Elements(mach->Inputs) is wrong now that mach->Inputs is dynamically
allocated.

13 years agoglcpp: Remove trailing contexts from #if rules.
Kenneth Graunke [Thu, 3 Mar 2011 17:52:36 +0000 (09:52 -0800)]
glcpp: Remove trailing contexts from #if rules.

These are now unnecessary.

13 years agoglcpp: Rework lexer to use a SKIP state rather than REJECT.
Kenneth Graunke [Tue, 1 Mar 2011 20:24:58 +0000 (12:24 -0800)]
glcpp: Rework lexer to use a SKIP state rather than REJECT.

Previously, the rule deleted by this commit was matched every single
time (being the longest match).  If not skipping, it used REJECT to
continue on to the actual correct rule.

The flex manual advises against using REJECT where possible, as it is
one of the most expensive lexer features.  So using it on every match
seems undesirable. Perhaps more importantly, it made it necessary for
the #if directive rules to contain a look-ahead pattern to make them
as long as the (now deleted) "skip the whole line" rule.

This patch introduces an exclusive start state, SKIP, to avoid REJECTs.
Each time the lexer is called, the code at the top of the rules section
will run, implicitly switching the state to the correct one.

Fixes piglit tests 16384-consecutive-chars.frag and
16385-consecutive-chars.frag.

13 years agoglcpp/tests: Update 063-comments.c.expected to match output.
Kenneth Graunke [Thu, 3 Mar 2011 07:26:24 +0000 (23:26 -0800)]
glcpp/tests: Update 063-comments.c.expected to match output.

The expected result has been out of sync with what glcpp produces for
some time; glcpp's actual result seems to be correct and is very close to
GCC's cpp.  Updating this will make it easier to catch regressions in
upcoming commits.

13 years agorbug: Fix depth stencil surface not being sent to the client
Jakob Bornecrantz [Thu, 3 Mar 2011 18:25:49 +0000 (18:25 +0000)]
rbug: Fix depth stencil surface not being sent to the client

13 years agoscons: More tweaks to fix MinGW build.
José Fonseca [Thu, 3 Mar 2011 16:57:38 +0000 (16:57 +0000)]
scons: More tweaks to fix MinGW build.

13 years agoscons: Ensure generated headers are in the include path.
José Fonseca [Thu, 3 Mar 2011 15:43:18 +0000 (15:43 +0000)]
scons: Ensure generated headers are in the include path.

13 years agoscons: Add human friendlier build messages for lex/yacc.
José Fonseca [Thu, 3 Mar 2011 15:42:58 +0000 (15:42 +0000)]
scons: Add human friendlier build messages for lex/yacc.

13 years agoscons: Always load lex/yacc tool.
José Fonseca [Thu, 3 Mar 2011 15:28:36 +0000 (15:28 +0000)]
scons: Always load lex/yacc tool.

lex/yacc is not loaded by default when toolchain is not default either,
e.g., when toolchain=crossmingw.

13 years agonv50: check grclass instead of chipset for 3D caps
Christoph Bumiller [Thu, 3 Mar 2011 11:31:35 +0000 (12:31 +0100)]
nv50: check grclass instead of chipset for 3D caps

13 years agonv50: increase size of shader code bo
Christoph Bumiller [Thu, 3 Mar 2011 11:25:12 +0000 (12:25 +0100)]
nv50: increase size of shader code bo

512 KiB should be quite enough, but dynamic resize might be nicer.

13 years agonouveau: allow pipe driver to define which buffers should start in sysmem
Ben Skeggs [Thu, 3 Mar 2011 05:56:20 +0000 (15:56 +1000)]
nouveau: allow pipe driver to define which buffers should start in sysmem

PIPE_BIND_CONSTANT_BUFFER alone was OK for nv50/nvc0, but nv30 will need
to be able to set others on certain chipsets.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
13 years agoi965: SNB GT1 has only 32k urb and max 128 urb entries.
Zou Nan hai [Thu, 3 Mar 2011 02:30:06 +0000 (10:30 +0800)]
i965: SNB GT1 has only 32k urb and max 128 urb entries.

Signed-off-by: Zou Nan hai <nanhai.zou@intel.com>
13 years agoMerge remote branch 'origin/master' into pipe-video
Christian König [Wed, 2 Mar 2011 23:59:12 +0000 (00:59 +0100)]
Merge remote branch 'origin/master' into pipe-video

13 years agoglsl: Remove unused glcpp/Makefile.am.
Kenneth Graunke [Wed, 2 Mar 2011 23:23:10 +0000 (15:23 -0800)]
glsl: Remove unused glcpp/Makefile.am.

This is a remnant of when glsl2 lived in its own repository.

13 years agoglsl: Remove 'tests' subfolder.
Kenneth Graunke [Wed, 2 Mar 2011 23:22:46 +0000 (15:22 -0800)]
glsl: Remove 'tests' subfolder.

These have long since moved to piglit and aren't useful to have here.

13 years agor600g: correct mega_fetch_count in fetch shader
Christian König [Wed, 2 Mar 2011 22:47:41 +0000 (23:47 +0100)]
r600g: correct mega_fetch_count in fetch shader

13 years agotgsi: defer allocation of huge inputs/outputs until we have a gs
Zack Rusin [Wed, 2 Mar 2011 03:50:42 +0000 (22:50 -0500)]
tgsi: defer allocation of huge inputs/outputs until we have a gs

13 years agodocs: added news item for 7.9.2 and 7.10.1 releases
Ian Romanick [Wed, 2 Mar 2011 22:33:59 +0000 (14:33 -0800)]
docs: added news item for 7.9.2 and 7.10.1 releases

13 years agodocs: All links to 7.9.2 and 7.10.1 release notes
Ian Romanick [Wed, 2 Mar 2011 22:30:40 +0000 (14:30 -0800)]
docs: All links to 7.9.2 and 7.10.1 release notes

13 years agodocs: Import 7.10.1 release notes from 7.10 branch
Ian Romanick [Wed, 2 Mar 2011 22:28:08 +0000 (14:28 -0800)]
docs: Import 7.10.1 release notes from 7.10 branch

13 years agodocs: Import 7.9.2 release notes from 7.9 branch
Ian Romanick [Wed, 2 Mar 2011 22:27:52 +0000 (14:27 -0800)]
docs: Import 7.9.2 release notes from 7.9 branch

13 years agonv50: fix IB index buffer path
Christoph Bumiller [Wed, 2 Mar 2011 21:36:22 +0000 (22:36 +0100)]
nv50: fix IB index buffer path

Add missing VERTEX_END and treat unaligned offsets correctly.

13 years agonv50: fix POINT_COORD_REPLACE_MAP method size
Christoph Bumiller [Wed, 2 Mar 2011 20:07:33 +0000 (21:07 +0100)]
nv50: fix POINT_COORD_REPLACE_MAP method size

Introduced in 223d98bb8d49c9e52e498a12980722467ae2bf87.

13 years agonv50: primitive restart trick for vertex data through FIFO mode
Christoph Bumiller [Wed, 2 Mar 2011 19:48:20 +0000 (20:48 +0100)]
nv50: primitive restart trick for vertex data through FIFO mode

Also, on nv50 the VERTEX_BEGIN method doesn't follow VERTEX_END,
which was erroneously taken over from nvc0 and is fixed now.

13 years agonv50: fix depth clamp for disabled primitive clipping
Christoph Bumiller [Wed, 2 Mar 2011 19:15:01 +0000 (20:15 +0100)]
nv50: fix depth clamp for disabled primitive clipping

13 years agonv50: implement independent blend functions for nva3+ and fix cap
Christoph Bumiller [Wed, 2 Mar 2011 18:41:08 +0000 (19:41 +0100)]
nv50: implement independent blend functions for nva3+ and fix cap

13 years agonv50: fix tile size calculations
Christoph Bumiller [Wed, 2 Mar 2011 18:21:46 +0000 (19:21 +0100)]
nv50: fix tile size calculations

13 years agonv50: fix point sprite state validation
Christoph Bumiller [Wed, 2 Mar 2011 18:02:59 +0000 (19:02 +0100)]
nv50: fix point sprite state validation

Wasn't updated if the FP didn't change, and coordinate replacement
wasn't disabled anymore.

13 years agonv50: allow accidentally disabled IB index buffers again
Christoph Bumiller [Wed, 2 Mar 2011 14:41:19 +0000 (15:41 +0100)]
nv50: allow accidentally disabled IB index buffers again

Must have sneaked in from debugging.

13 years agonv50: apply relocations to shader code
Christoph Bumiller [Wed, 2 Mar 2011 14:39:57 +0000 (15:39 +0100)]
nv50: apply relocations to shader code

On nv50, branches are absolute, so we need to adjust them according
to the shader's position in the code buffer.

13 years agonv50: fix wrong miptree tile flags taken over from nvc0
Christoph Bumiller [Wed, 2 Mar 2011 19:57:29 +0000 (20:57 +0100)]
nv50: fix wrong miptree tile flags taken over from nvc0

13 years agor600g: merge fix
Christian König [Wed, 2 Mar 2011 19:48:03 +0000 (20:48 +0100)]
r600g: merge fix

13 years agoegl_dri2 x11: Workaround device_name xcb-dri2 bug
Benjamin Franzke [Wed, 2 Mar 2011 20:23:04 +0000 (21:23 +0100)]
egl_dri2 x11: Workaround device_name xcb-dri2 bug

This commit is basically a copy-over of the fix
Chia-I Wu's commited to wayland:
   http://cgit.freedesktop.org/wayland/wayland-demos/commit/?id=1b6c0ed95
   "Workaround an xcb-dri2 bug.
    xcb_dri2_connect_device_name generated by xcb-proto 1.6 is broken.
    It only works when the length of the driver name is a multiple of 4."

13 years agoegl/wayland: build subdirs (wayland-drm) before depend
Benjamin Franzke [Wed, 2 Mar 2011 19:17:26 +0000 (20:17 +0100)]
egl/wayland: build subdirs (wayland-drm) before depend

Autogenerated files need to be generated first.

13 years agor300g: require DRM 2.3.0 (kernel 2.6.34)
Marek Olšák [Wed, 2 Mar 2011 16:51:19 +0000 (17:51 +0100)]
r300g: require DRM 2.3.0 (kernel 2.6.34)

Running any older kernel is not recommended anyway.

13 years agor300g: do not use ioctl thread offloading on single-core machines
Marek Olšák [Wed, 2 Mar 2011 16:37:52 +0000 (17:37 +0100)]
r300g: do not use ioctl thread offloading on single-core machines

13 years agomesa: added gl_program_constants::MaxAddressOffset
Brian Paul [Wed, 2 Mar 2011 16:32:45 +0000 (09:32 -0700)]
mesa: added gl_program_constants::MaxAddressOffset

See https://bugs.freedesktop.org/show_bug.cgi?id=29418

13 years agomesa: increase INST_INDEX_BITS to 12
Brian Paul [Wed, 2 Mar 2011 16:20:57 +0000 (09:20 -0700)]
mesa: increase INST_INDEX_BITS to 12

For more info see fd.o bug 29418.

13 years agoRevert "mesa: reduce calls to _mesa_test_framebuffer_completeness()"
Brian Paul [Wed, 2 Mar 2011 16:11:43 +0000 (09:11 -0700)]
Revert "mesa: reduce calls to _mesa_test_framebuffer_completeness()"

This reverts commit 1f9a0a4e6e5566c36c781add5f1e62af3efdfb58.

This caused trouble with Lightsmark w/ i965 driver and fbo/fbo-blit-d24s8
(see bug 34894).  It's probably something simple but no time to debug now.

13 years agovbo: fix error parameter
Brian Paul [Wed, 2 Mar 2011 02:11:44 +0000 (19:11 -0700)]
vbo: fix error parameter

Spotted by Ian.

13 years agor300g: Silence 'control reaches end of non-void function' warning.
Vinson Lee [Wed, 2 Mar 2011 08:43:09 +0000 (00:43 -0800)]
r300g: Silence 'control reaches end of non-void function' warning.

Fixes this GCC warning.
r300_hyperz.c: In function 'r300_get_hiz_func':
r300_hyperz.c:65: warning: control reaches end of non-void function

13 years agogallium: Add u_format_rgtc.c to SConscript.
Vinson Lee [Wed, 2 Mar 2011 07:02:50 +0000 (23:02 -0800)]
gallium: Add u_format_rgtc.c to SConscript.

13 years agoi965: Maxinum the usage of urb space on SNB.
Zou Nan hai [Wed, 2 Mar 2011 06:23:17 +0000 (14:23 +0800)]
i965: Maxinum the usage of urb space on SNB.
      SNB has 64k urb space, we only use piece of them.
      The more urb space we alloc,
      the more concurrent vs threads we can run.
      push the urb space usage to the limit.

Signed-off-by: Zou Nan hai <nanhai.zou@intel.com>
13 years agomesa/st: fix softpipe npot compressed mipmaps.
Dave Airlie [Wed, 2 Mar 2011 06:13:59 +0000 (16:13 +1000)]
mesa/st: fix softpipe npot compressed mipmaps.

this fixes fbo-generatemipmap-formats rgtc and s3tc in NPOT mode
with softpipe.

r600g fails to even get level 0 correct so have to look into that
a bit further.

Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agosoftpipe: enable RGTC now that we have u_format support.
Dave Airlie [Wed, 2 Mar 2011 04:36:22 +0000 (14:36 +1000)]
softpipe: enable RGTC now that we have u_format support.

13 years agomesa/st: fix generate mipmap for signed compressed formats.
Dave Airlie [Wed, 2 Mar 2011 05:28:32 +0000 (15:28 +1000)]
mesa/st: fix generate mipmap for signed compressed formats.

This was always converting to 8-bit per channel unsigned formats,
which isn't suitable for RGTC signed formats, this special cases
those two formats and converts to floats for those.

Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agogallium: add RGTC UNORM support to u_format.
Dave Airlie [Wed, 2 Mar 2011 04:34:14 +0000 (14:34 +1000)]
gallium: add RGTC UNORM support to u_format.

SNORM needs a bit of work in the state tracker in order for mipmap
generation to work I believe.

I'm also not sure that having unorm fetches for an snorm format is
sane.

13 years agorgtc: remove GL types from this file.
Dave Airlie [Wed, 2 Mar 2011 04:33:35 +0000 (14:33 +1000)]
rgtc: remove GL types from this file.

I'd like to share this file with gallium u_format stuff.

Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agorgtc: move the texel fetch into common unsigned/signed code.
Dave Airlie [Wed, 2 Mar 2011 04:16:39 +0000 (14:16 +1000)]
rgtc: move the texel fetch into common unsigned/signed code.

This function can be done in the include file also.

Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agorgtc: fix issues with compressor and signed types.
Dave Airlie [Wed, 2 Mar 2011 04:08:59 +0000 (14:08 +1000)]
rgtc: fix issues with compressor and signed types.

With signed types we weren't hitting this test however the comment
stating this doesn't happen often doesn't apply when using signed
types since an all 0 block is quite common which isn't abs min or max.

this fixes the limits correctly again also.

Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agorgtc: don't try to access off the end of the block.
Dave Airlie [Wed, 2 Mar 2011 03:14:34 +0000 (13:14 +1000)]
rgtc: don't try to access off the end of the block.

if the values are all in the last dword, the high bits can be 0,

This fixes a valgrind warning I saw when playing with mipmaps.

Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agorgtc: move to using ubyte for fetch instead of chan + fix limit
Dave Airlie [Wed, 2 Mar 2011 03:01:01 +0000 (13:01 +1000)]
rgtc: move to using ubyte for fetch instead of chan + fix limit

My previous fix to the byte max was incorrect.

Signed-off-by: Dave Airlie <airlied@redhat.com>
13 years agost/mesa: use RGTC for GL_COMPRESSED_RED/RG if possible
Marek Olšák [Tue, 1 Mar 2011 20:15:38 +0000 (21:15 +0100)]
st/mesa: use RGTC for GL_COMPRESSED_RED/RG if possible

With proper fallback formats.

13 years agosvga: reduce MAX_DMA_SIZE to 4MB
Brian Paul [Wed, 2 Mar 2011 00:40:27 +0000 (17:40 -0700)]
svga: reduce MAX_DMA_SIZE to 4MB

13 years agovbo: add vbo_always_unmap_buffers()
Brian Paul [Wed, 2 Mar 2011 00:16:53 +0000 (17:16 -0700)]
vbo: add vbo_always_unmap_buffers()

Drivers can call this function as needed.  It tells the VBO module to
always unmap the current glBegin/glEnd VBO when we flush.  Otherwise
it's possible to be in a flushed state but still have the VBO mapped.