Zack Rusin [Sat, 24 Oct 2009 05:29:27 +0000 (01:29 -0400)]
st/xorg: stop overflowing yuv buffers
Zack Rusin [Sat, 24 Oct 2009 05:05:40 +0000 (01:05 -0400)]
st/xorg: add yuv vertex shader
plus some general fixes
Zack Rusin [Fri, 23 Oct 2009 20:18:10 +0000 (16:18 -0400)]
st/xorg: add yuv shaders
Marek Olšák [Mon, 26 Oct 2009 00:47:55 +0000 (01:47 +0100)]
r300g: read occlusion query results from both Z pipes on RV530
Marek Olšák [Mon, 26 Oct 2009 00:46:21 +0000 (01:46 +0100)]
r300g: add cubemap support
Also, set a pitch for rectangles only.
José Fonseca [Sun, 25 Oct 2009 21:11:54 +0000 (21:11 +0000)]
gallium: Move enum pipe_error into p_defines.h.
It's really just another define. No need for its own header.
José Fonseca [Sun, 25 Oct 2009 12:27:14 +0000 (12:27 +0000)]
llvmpipe: Human friendlier sampler state dump.
José Fonseca [Sun, 25 Oct 2009 12:26:39 +0000 (12:26 +0000)]
util: Human readable output of texture states.
José Fonseca [Sun, 25 Oct 2009 11:49:01 +0000 (11:49 +0000)]
llvmpipe: Dump the sampler state of the shader key.
José Fonseca [Sun, 25 Oct 2009 11:48:17 +0000 (11:48 +0000)]
llvmpipe: Immediate multiplication.
José Fonseca [Sun, 25 Oct 2009 09:51:57 +0000 (09:51 +0000)]
llvmpipe: Fast path for sampling rgba8 textures with linear filtering.
Implement Keith's suggestion of doing most of the sampling with 16x8
and 8x16 AoS, and only doing the conversion to floating point SoA at
the very last step.
Improves gloss performance by 10%.
José Fonseca [Sun, 25 Oct 2009 09:16:38 +0000 (09:16 +0000)]
llvmpipe: Make lerping work for 8.8 fixed point values.
José Fonseca [Sun, 25 Oct 2009 09:09:59 +0000 (09:09 +0000)]
llvmpipe: New module to help make assertions about formats.
José Fonseca [Sun, 25 Oct 2009 09:09:23 +0000 (09:09 +0000)]
llvmpipe: New function to unpack rgba8 formats into 4 x u8n AoS.
José Fonseca [Sun, 25 Oct 2009 09:06:05 +0000 (09:06 +0000)]
llvmpipe: Split the format swizzling step from the unpacking.
José Fonseca [Sun, 25 Oct 2009 09:03:50 +0000 (09:03 +0000)]
llvmpipe: Allow different signs when unpacking.
José Fonseca [Sun, 25 Oct 2009 09:03:18 +0000 (09:03 +0000)]
llvmpipe: Add inlines to quickly generate types matching the native SIMD register bitwidth.
José Fonseca [Thu, 22 Oct 2009 21:44:32 +0000 (22:44 +0100)]
llvmpipe: Merge lp_build_load_rgba_soa into lp_build_sample_texel.
José Fonseca [Thu, 22 Oct 2009 21:37:06 +0000 (22:37 +0100)]
llvmpipe: Factor our pixel offset computation.
José Fonseca [Thu, 22 Oct 2009 20:16:26 +0000 (21:16 +0100)]
llvmpipe: Share testing infrastructure with lp_test_format.
José Fonseca [Thu, 22 Oct 2009 20:06:03 +0000 (21:06 +0100)]
llvmpipe: Eliminate lp_build_store_rgba_aos.
José Fonseca [Thu, 22 Oct 2009 19:58:35 +0000 (20:58 +0100)]
llvmpipe: Eliminate lp_build_load_rgba_aos.
José Fonseca [Thu, 22 Oct 2009 19:42:16 +0000 (20:42 +0100)]
llvmpipe: Move a few format/sampling functions into better space.
Marek Olšák [Sun, 25 Oct 2009 08:57:53 +0000 (09:57 +0100)]
r300g: fix redefining mipmaps and fetching from them
Marek Olšák [Sun, 25 Oct 2009 04:05:27 +0000 (05:05 +0100)]
r300g: added support for 3D textures
Mipmaps not tested. Also, I am not sure why piglit/texturing/tex3d needs
to have color tolerance +-1 to pass. The classic Mesa driver doesn't
need that.
Joakim Sindholt [Sat, 24 Oct 2009 00:38:28 +0000 (02:38 +0200)]
r300g: fix scons build yet again
Brian Paul [Sat, 24 Oct 2009 00:38:43 +0000 (18:38 -0600)]
Merge branch 'mesa_7_6_branch'
Brian Paul [Sat, 24 Oct 2009 00:15:55 +0000 (18:15 -0600)]
mesa: simplify att->CubeMapFace assignment
Brian Paul [Fri, 23 Oct 2009 22:31:48 +0000 (16:31 -0600)]
mesa: refactor, new print_shader_info()
Brian Paul [Fri, 23 Oct 2009 19:20:03 +0000 (13:20 -0600)]
mesa: rework error checking code for glGetCompressedTexImage()
Do all error checking in new getcompressedteximage_error_check() func.
Move some additional PBO checks out of the driver fallbacks into the
error checking functions.
Brian Paul [Fri, 23 Oct 2009 19:06:13 +0000 (13:06 -0600)]
mesa: remove misplaced VERBOSE_TEXTURE tests
Brian Paul [Fri, 23 Oct 2009 17:34:14 +0000 (11:34 -0600)]
mesa: more detailed error messages in fbo code
Brian Paul [Sat, 24 Oct 2009 00:21:59 +0000 (18:21 -0600)]
mesa: remove FBO texture depth/stencil test
The texture format should not be checked until validation time since
the format might be changed by a subsequent glTexImage() call.
Zack Rusin [Fri, 23 Oct 2009 13:35:36 +0000 (09:35 -0400)]
st/xorg: fix text with component alpha rendering
Brian Paul [Fri, 23 Oct 2009 22:36:15 +0000 (16:36 -0600)]
mesa: fix broken _mesa_str_checksum()
Eric Anholt [Fri, 23 Oct 2009 21:19:52 +0000 (14:19 -0700)]
Merge remote branch 'origin/mesa_7_6_branch'
Conflicts:
src/mesa/drivers/dri/intel/intel_fbo.c
src/mesa/drivers/dri/intel/intel_mipmap_tree.c
src/mesa/drivers/dri/intel/intel_mipmap_tree.h
src/mesa/drivers/dri/intel/intel_tex_copy.c
src/mesa/drivers/dri/intel/intel_tex_image.c
Eric Anholt [Wed, 8 Jul 2009 02:33:55 +0000 (19:33 -0700)]
i965: Annotate _NEW_COLOR dependency for brw_update_renderbuffer_surface().
Eric Anholt [Thu, 9 Jul 2009 16:32:21 +0000 (09:32 -0700)]
intel: Keep track of x,y offsets in miptrees and use them for blitting.
By just using offsets, we confused the hardware's tiling calculations,
resulting in failures in miptree validation and blit clears.
Fixes piglit fbo-clearmipmap.
Bug #23552. (automatic mipmap generation)
Brian Paul [Fri, 23 Oct 2009 20:52:10 +0000 (14:52 -0600)]
progs/tests: print requested/internal format info
Brian Paul [Fri, 23 Oct 2009 20:49:47 +0000 (14:49 -0600)]
docs: document updated GL extension header files
Brian Paul [Fri, 23 Oct 2009 20:48:49 +0000 (14:48 -0600)]
gl: updated wglext.h to version 17
Brian Paul [Fri, 23 Oct 2009 20:48:39 +0000 (14:48 -0600)]
gl: updated glxext.h to version 25
Brian Paul [Fri, 23 Oct 2009 20:48:30 +0000 (14:48 -0600)]
gl: updated glext.h to version 56
Brian Paul [Fri, 23 Oct 2009 20:42:32 +0000 (14:42 -0600)]
docs: document cell, glGetTexLevelParameter() fixes
Brian Paul [Fri, 23 Oct 2009 20:40:40 +0000 (14:40 -0600)]
mesa: fix GL_TEXTURE_INTERNAL_FORMAT query for compressed formats
Need to return the actual compressed format when the user originally
requested a generic compressed format.
Brian Paul [Fri, 23 Oct 2009 20:37:41 +0000 (14:37 -0600)]
mesa: added _mesa_compressed_format_to_glenum()
Maps a compressed MESA_FORMAT_x to correspding GLenum. Needed for
querying a texture's actual format when a generic format was originally
requested.
Brian Paul [Fri, 23 Oct 2009 20:37:14 +0000 (14:37 -0600)]
mesa: change s3tc vs. fxt1 priority when choosing compressed formats
Christoph Bumiller [Fri, 23 Oct 2009 20:00:06 +0000 (22:00 +0200)]
nv50: do SIGN_SET as one instruction
Christoph Bumiller [Fri, 23 Oct 2009 19:57:42 +0000 (21:57 +0200)]
nv50: fix saturation outside of tx_insn case
Christoph Bumiller [Fri, 23 Oct 2009 19:38:37 +0000 (21:38 +0200)]
nv50: allow all 127 TEMP regs
We should really learn to not waste so many though.
Christoph Bumiller [Fri, 23 Oct 2009 20:17:44 +0000 (22:17 +0200)]
nv50: fix address reg code
Contained some rather obvious thinking errors before,
and didn't consider offsets from TGSI ADDRESS regs.
Christoph Bumiller [Fri, 23 Oct 2009 18:05:31 +0000 (20:05 +0200)]
nv50: add depth texture formats, and a few others, too
Vinson Lee [Fri, 23 Oct 2009 19:49:02 +0000 (13:49 -0600)]
progs/tests: Fix MSVC build.
Patrice Mandin [Fri, 23 Oct 2009 16:42:21 +0000 (18:42 +0200)]
nouveau: nv30: Relax some limits. We can render to z24s8 buffer even if color buffer is 16 bits.
Patrice Mandin [Fri, 23 Oct 2009 16:40:13 +0000 (18:40 +0200)]
nouveau: nv30: use r5g6b5 as z16 format
Chia-I Wu [Fri, 16 Oct 2009 08:04:06 +0000 (16:04 +0800)]
glapi: Always build libglapi.a.
This is made possible by making glapioffsets.h and glapidispatch.h
internal headers of glapi. They should only be included indirectly
through dispatch.h by mesa.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Chia-I Wu [Fri, 16 Oct 2009 08:01:57 +0000 (16:01 +0800)]
glapi: Move dispatch marcos to glapidispatch.h.
dispatch.h is kept as a wrapper to glapidispatch.h.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Chia-I Wu [Thu, 8 Oct 2009 03:52:11 +0000 (11:52 +0800)]
mesa/dri: Remove extension_helper.h.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Chia-I Wu [Thu, 8 Oct 2009 02:33:57 +0000 (10:33 +0800)]
mesa: Enable remap table in core.
This enables the remap table in core. driInitExtensions is adapted to
use the remap table. All uses of extension_helper.h are replaced by
remap_helper.h. The chicken-egg problem of the DRI drivers is also
solved.
It is now also possible to pass NULL extensions to driInitExtensions.
It will cause driInitExtensions to map all known functions. This
functionality is used by software drivers and EGL_i915.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Chia-I Wu [Thu, 8 Oct 2009 02:33:32 +0000 (10:33 +0800)]
mesa/main: Add support for remap table.
This commit only adds the source files. It is supposed to replace the
remap table in DRI drivers.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Zack Rusin [Fri, 23 Oct 2009 05:05:23 +0000 (01:05 -0400)]
st/xorg: lots of fixes related to compositing
fixes transformations, rendering with multiple bound textures, xrender
matrix conversions plus some cleanups
Andre Maasikas [Fri, 23 Oct 2009 13:44:31 +0000 (16:44 +0300)]
r600: remove duplicate stride setting
Stride is set already in r700SetVertexFormat and there it works
correctly for 0 also
Andre Maasikas [Fri, 23 Oct 2009 11:25:09 +0000 (14:25 +0300)]
r600: for position invariant programs reading vert_pos is not always known at this point
Keith Whitwell [Fri, 23 Oct 2009 13:50:02 +0000 (14:50 +0100)]
gallium: remove extended negate also, and also the ExtSwz token
Likewise, the extended negate functionality hasn't been
used since mesa switched to using tgsi_ureg to build programs,
and has been translating the SWZ opcode internally to a single MAD.
Keith Whitwell [Fri, 23 Oct 2009 13:38:30 +0000 (14:38 +0100)]
cell: typo from ExtSwizzle commit
Keith Whitwell [Fri, 23 Oct 2009 13:31:24 +0000 (14:31 +0100)]
gallium: remove the swizzling parts of ExtSwizzle
These haven't been used by the mesa state tracker since the
conversion to tgsi_ureg, and it seems that none of the
other state trackers are using it either.
This helps simplify one of the biggest suprises when starting off with
TGSI shaders.
Keith Whitwell [Fri, 23 Oct 2009 12:49:04 +0000 (13:49 +0100)]
gallium: remove noise opcodes
Provide a dummy implementation in the GL state tracker (move 0.5 to
the destination regs).
At some point, a motivated person could add a better
implementation of noise. Currently not even the nvidia
binary drivers do anything more than this. In any case, the
place to do this is in the GL state tracker, not the poor
driver.
Keith Whitwell [Sat, 26 Sep 2009 07:39:57 +0000 (08:39 +0100)]
progs/demos: add fps to multiarb
Cooper Yuan [Fri, 23 Oct 2009 08:40:31 +0000 (16:40 +0800)]
r300g: last changes's typo, miss a include file
Cooper Yuan [Fri, 23 Oct 2009 06:46:29 +0000 (14:46 +0800)]
r300g: add flush_frontbuffer function to display video surface
Cooper Yuan [Fri, 23 Oct 2009 06:44:27 +0000 (14:44 +0800)]
g3dvl: pass display and screen to g3dvl when creating video private context
Alex Deucher [Fri, 23 Oct 2009 05:00:23 +0000 (01:00 -0400)]
r600: remove remains of old tnl pipeline
Alex Deucher [Fri, 23 Oct 2009 04:40:41 +0000 (00:40 -0400)]
r600: fix render size prediction
Alex Deucher [Thu, 22 Oct 2009 21:41:31 +0000 (17:41 -0400)]
r600: remove old tnl pipeline
Alex Deucher [Thu, 22 Oct 2009 21:18:01 +0000 (17:18 -0400)]
r600: clean up context creation
Make it more consistent with other radeon drivers.
Ian Romanick [Fri, 23 Oct 2009 02:21:21 +0000 (19:21 -0700)]
Revert "Store clipping distance for user clip planes as part of vertex processing"
This reverts commit
f058b25881e08c9d89a33345e5c84e1357396932.
This change is completely wrong in so many ways. When clip distances
are generated as part of vertex processing, they must be interpolated
to perform clipping. Geometric clipping goes right out the window.
Ian Romanick [Tue, 20 Oct 2009 17:58:14 +0000 (10:58 -0700)]
ARB prog parser: Fix parameter array size comparison
Array indexes are invalid when >= the maximum, but array sizes are
only in valid when > the maximum. This prevented programs from
declaring a single maximum size array.
See the piglit vp-max-array test.
Brian Paul [Fri, 23 Oct 2009 00:38:19 +0000 (18:38 -0600)]
Merge branch 'mesa_7_6_branch'
Brian Paul [Thu, 22 Oct 2009 15:38:39 +0000 (09:38 -0600)]
docs: document 'useprog' debug option
Brian Paul [Thu, 22 Oct 2009 15:37:56 +0000 (09:37 -0600)]
mesa: fix up vbo comments
Brian Paul [Thu, 22 Oct 2009 15:37:22 +0000 (09:37 -0600)]
mesa: added _mesa_dump_texture()
Brian Paul [Fri, 23 Oct 2009 00:32:48 +0000 (18:32 -0600)]
intel: flush old context before binding new context
Per the GLX spec, when changing rendering contexts, the old context
should first be flushed.
Brian Paul [Fri, 23 Oct 2009 00:19:01 +0000 (18:19 -0600)]
glx: don't destroy context immediately if it's currently bound
According to the GLXDestroyContext() man page, the context should not
immediately be destroyed if it's bound to some thread. Wait until it's
unbound to really delete it. The code for doing the later part is
already present in MakeContextCurrent() so no change was needed there.
Brian Paul [Fri, 23 Oct 2009 00:16:10 +0000 (18:16 -0600)]
mesa: code refactoring- new _mesa_finish(), _mesa_flush()
Zack Rusin [Thu, 22 Oct 2009 23:31:44 +0000 (19:31 -0400)]
st/xorg: fix the makefile when used with new xextproto
Brian Paul [Thu, 22 Oct 2009 23:11:59 +0000 (17:11 -0600)]
progs/xdemos: fix command line flag typo
Brian Paul [Thu, 22 Oct 2009 22:45:50 +0000 (16:45 -0600)]
i965: fix hacked Fallback usage in brw_prepare_vertices()
Setting intel->Fallback = 1 clobbered any fallback state that was already
set. Not sure where this hack originated (the git history is a little
convoluted). Define and use a new BRW_FALLBACK_DRAW bit instead. This
shouldn't break anything and could potentially fix some bugs (but no
specific ones are known).
Brian Paul [Thu, 22 Oct 2009 22:45:03 +0000 (16:45 -0600)]
intel: define INTEL_FALLBACK_DRIVER for drivers
Brian Paul [Thu, 22 Oct 2009 22:33:36 +0000 (16:33 -0600)]
intel: Fallback field is a bitmask, use GLbitfield
Brian Paul [Thu, 22 Oct 2009 22:32:08 +0000 (16:32 -0600)]
i965: remove unused brw_context::tmp_fallback field
Brian Paul [Thu, 22 Oct 2009 22:26:54 +0000 (16:26 -0600)]
i965: remove unused BRW_FALLBACK_TEXTURE bit
The value was probably wrong too.
It was the same as INTEL_FALLBACK_DRAW_BUFFER.
Corbin Simpson [Thu, 22 Oct 2009 21:28:47 +0000 (14:28 -0700)]
r300g: Enable more stuff in r300_screen, cleanup comments.
Also enable 24-bit depth buffers without stencil.
Corbin Simpson [Thu, 22 Oct 2009 20:42:03 +0000 (13:42 -0700)]
r300g: Cleanup PSC setup math a bit and stop using Draw formats.
Christoph Bumiller [Thu, 22 Oct 2009 20:57:30 +0000 (22:57 +0200)]
nv50: handle PIPE_TEX_FILTER_ANISO case
Set the same bits as for linear filtering (in addition
to max anisotropy), and 2 unknown bits I've seen set.
Patrice Mandin [Thu, 22 Oct 2009 20:01:53 +0000 (22:01 +0200)]
nouveau: nv30: rewrite so we can render only in depth buffer
Patrice Mandin [Thu, 22 Oct 2009 19:55:09 +0000 (21:55 +0200)]
nouveau: nv30: use a8r8g8b8 as depth texture format for z24s8
Corbin Simpson [Thu, 22 Oct 2009 19:57:13 +0000 (12:57 -0700)]
Kill off trident.
Hm. I could have said "chew trident and spit it out," or perhaps
"spear trident," instead. Dohoho.
Corbin Simpson [Thu, 22 Oct 2009 19:29:30 +0000 (12:29 -0700)]
Nuke s3v.
As per FDO #17889.
Corbin Simpson [Thu, 22 Oct 2009 18:45:36 +0000 (11:45 -0700)]
r300g: Clean up duplicate code in r300_render.
Corbin Simpson [Thu, 22 Oct 2009 18:39:59 +0000 (11:39 -0700)]
r300g: Move render functions to r300_render.
Part of the fastpath cleanup.