mesa.git
13 years agost/mesa: Unreference the sampler view in st_bind_surface.
Chia-I Wu [Sat, 30 Oct 2010 17:13:27 +0000 (01:13 +0800)]
st/mesa: Unreference the sampler view in st_bind_surface.

Without this, update_textures may not pick up the new pipe_resource.

It is actually update_textures that should check
stObj->sampler_view->texture != stObj->pt, but let's follow st_TexImage
and others for now.

13 years agoosmesa: fix renderbuffer memleak in OSMesaMakeCurrent()
Brian Paul [Sat, 30 Oct 2010 16:10:52 +0000 (10:10 -0600)]
osmesa: fix renderbuffer memleak in OSMesaMakeCurrent()

Fixes fd.o bug 31128.

13 years agoautoconf: st/vega requires --enable-openvg.
Chia-I Wu [Sat, 30 Oct 2010 06:26:01 +0000 (14:26 +0800)]
autoconf: st/vega requires --enable-openvg.

Make it a warning for now to smooth the transition.

13 years agoglsl: Remove unused ARRAY_SIZE macro.
Kenneth Graunke [Fri, 29 Oct 2010 18:42:22 +0000 (11:42 -0700)]
glsl: Remove unused ARRAY_SIZE macro.

It's also equivalent to Elements(...) which is already used elsewhere.

13 years agomesa: Make metaops use program refcounts instead of names.
Eric Anholt [Fri, 29 Oct 2010 18:28:38 +0000 (11:28 -0700)]
mesa: Make metaops use program refcounts instead of names.

Fixes failure on restoring state when the program was active but
deleted, and the name no longer exists.

Bug #31194

13 years agomesa: remove dead code
Brian Paul [Fri, 29 Oct 2010 14:13:31 +0000 (08:13 -0600)]
mesa: remove dead code

13 years agomesa: Fix windows build (uint -> GLuint).
José Fonseca [Fri, 29 Oct 2010 12:05:31 +0000 (13:05 +0100)]
mesa: Fix windows build (uint -> GLuint).

13 years agotargets: Add missing quotes to Makefile.xorg.
Chia-I Wu [Fri, 29 Oct 2010 04:59:45 +0000 (12:59 +0800)]
targets: Add missing quotes to Makefile.xorg.

Fix

  $ make CC="ccache gcc"

13 years agoMerge branch 'glapi-reorg'
Chia-I Wu [Fri, 29 Oct 2010 04:45:04 +0000 (12:45 +0800)]
Merge branch 'glapi-reorg'

Conflicts:
src/mapi/glapi/glapi_sparc.S
src/mapi/glapi/glapi_x86.S
src/mapi/glapi/glapidispatch.h
src/mapi/glapi/glapioffsets.h
src/mapi/glapi/glprocs.h

13 years agoautoconf: Update configuration info.
Chia-I Wu [Fri, 29 Oct 2010 04:34:44 +0000 (12:34 +0800)]
autoconf: Update configuration info.

Output API info first.  Move GLU/GLw/GLUT and EGL near driver info.

13 years agodocs: Update egl and openvg docs.
Chia-I Wu [Thu, 28 Oct 2010 09:25:07 +0000 (17:25 +0800)]
docs: Update egl and openvg docs.

13 years agoautoconf: Better client API selection.
Chia-I Wu [Wed, 27 Oct 2010 08:14:27 +0000 (16:14 +0800)]
autoconf: Better client API selection.

Make autoconf decide the client APIs enabled first.  Then when OpenGL
and OpenGL ES are disabled, there is no need to build src/mesa/;  when
OpenGL is disabled, no $mesa_driver should be built.  Finally, add
--enable-openvg to enable OpenVG.

With these changes, an OpenVG only build can be configured with

  $ ./configure --disable-opengl --enable-openvg

src/mesa, src/glsl, and src/glx will be skipped, which saves a great
deal of compilation time.

And an OpenGL ES only build can be configured with

  $ ./configure --disable-opengl --enable-gles-overlay

13 years agomesa: pixel transfer ops do not apply to integer-valued textures
Brian Paul [Fri, 29 Oct 2010 03:17:42 +0000 (21:17 -0600)]
mesa: pixel transfer ops do not apply to integer-valued textures

13 years agomesa: additional integer formats in _mesa_bytes_per_pixel()
Brian Paul [Fri, 29 Oct 2010 03:17:42 +0000 (21:17 -0600)]
mesa: additional integer formats in _mesa_bytes_per_pixel()

13 years agomesa: add const qualifier to _mesa_is_legal_format_and_type()
Brian Paul [Fri, 29 Oct 2010 03:17:42 +0000 (21:17 -0600)]
mesa: add const qualifier to _mesa_is_legal_format_and_type()

13 years agomesa: fix integer cases in _mesa_is_legal_format_and_type()
Brian Paul [Fri, 29 Oct 2010 03:17:42 +0000 (21:17 -0600)]
mesa: fix integer cases in _mesa_is_legal_format_and_type()

Some integer formats work with some packed datatypes.

13 years agomesa: fix incorrect type in _mesa_texstore_rgba_int16()
Brian Paul [Fri, 29 Oct 2010 03:17:42 +0000 (21:17 -0600)]
mesa: fix incorrect type in _mesa_texstore_rgba_int16()

13 years agomesa: remove obsolete comment
Brian Paul [Fri, 29 Oct 2010 03:17:42 +0000 (21:17 -0600)]
mesa: remove obsolete comment

13 years agomesa: add extension table entry for GL_EXT_gpu_shader4
Brian Paul [Fri, 29 Oct 2010 03:17:42 +0000 (21:17 -0600)]
mesa: add extension table entry for GL_EXT_gpu_shader4

13 years agomesa: clean-up array element code
Brian Paul [Fri, 29 Oct 2010 03:17:42 +0000 (21:17 -0600)]
mesa: clean-up array element code

Remove unnecessary GLAPIENTRY keywords, update comments, re-indent.

13 years agomesa: glArrayElement support for integer-valued arrays
Brian Paul [Fri, 29 Oct 2010 03:17:42 +0000 (21:17 -0600)]
mesa: glArrayElement support for integer-valued arrays

13 years agomesa: state/queries for GL_MIN/MAX_PROGRAM_TEXEL_OFFSET_EXT
Brian Paul [Fri, 29 Oct 2010 03:17:42 +0000 (21:17 -0600)]
mesa: state/queries for GL_MIN/MAX_PROGRAM_TEXEL_OFFSET_EXT

13 years agomesa: consolidate glVertex/Color/etcPointer() code
Brian Paul [Fri, 29 Oct 2010 03:17:42 +0000 (21:17 -0600)]
mesa: consolidate glVertex/Color/etcPointer() code

This removes a bunch of similar error checking code in all the vertex
pointer functions and puts nearly all the error checking in update_array().

13 years agomesa: add gl_client_array::Integer field and related vertex array state code
Brian Paul [Fri, 29 Oct 2010 03:17:41 +0000 (21:17 -0600)]
mesa: add gl_client_array::Integer field and related vertex array state code

13 years agomesa: implement integer-valued vertex attribute functions
Brian Paul [Fri, 29 Oct 2010 03:17:41 +0000 (21:17 -0600)]
mesa: implement integer-valued vertex attribute functions

The integers still get converted to floats.  That'll have to change someday.

13 years agomesa: add new GLvertexformat entries for integer-valued attributes
Brian Paul [Fri, 29 Oct 2010 03:17:41 +0000 (21:17 -0600)]
mesa: add new GLvertexformat entries for integer-valued attributes

13 years agomesa: plug in more GL_EXT_gpu_shader4 functions
Brian Paul [Fri, 29 Oct 2010 03:17:41 +0000 (21:17 -0600)]
mesa: plug in more GL_EXT_gpu_shader4 functions

13 years agomesa: add glGetUniformuiv(), plug in uint glUniform funcs
Brian Paul [Fri, 29 Oct 2010 03:17:41 +0000 (21:17 -0600)]
mesa: add glGetUniformuiv(), plug in uint glUniform funcs

13 years agomesa: plug in stubs for glBindFragDataLocation(), glGetFragDataLocation()
Brian Paul [Fri, 29 Oct 2010 03:17:41 +0000 (21:17 -0600)]
mesa: plug in stubs for glBindFragDataLocation(), glGetFragDataLocation()

13 years agoglapi: regenerated API files
Brian Paul [Fri, 29 Oct 2010 03:17:41 +0000 (21:17 -0600)]
glapi: regenerated API files

13 years agoglapi: include EXT_gpu_shader4.xml
Brian Paul [Fri, 29 Oct 2010 03:17:41 +0000 (21:17 -0600)]
glapi: include EXT_gpu_shader4.xml

13 years agoglapi: xml spec file for GL_EXT_gpu_shader4
Brian Paul [Fri, 29 Oct 2010 03:17:41 +0000 (21:17 -0600)]
glapi: xml spec file for GL_EXT_gpu_shader4

13 years agovbo: re-indent file
Brian Paul [Fri, 29 Oct 2010 03:17:41 +0000 (21:17 -0600)]
vbo: re-indent file

13 years agomesa: remove 'normalized' parameter from _mesa_VertexAttribIPointer()
Brian Paul [Fri, 29 Oct 2010 03:17:41 +0000 (21:17 -0600)]
mesa: remove 'normalized' parameter from _mesa_VertexAttribIPointer()

13 years agoi965: Update the gen6 stencil ref state when stencil state changes.
Eric Anholt [Thu, 28 Oct 2010 23:09:41 +0000 (16:09 -0700)]
i965: Update the gen6 stencil ref state when stencil state changes.

Fixes 6 piglit tests about stencil operations.

13 years agoi965: Upload required gen6 VS push constants even when using pull constants.
Eric Anholt [Thu, 28 Oct 2010 22:30:09 +0000 (15:30 -0700)]
i965: Upload required gen6 VS push constants even when using pull constants.

Matches pre-gen6, and fixes glsl-vs-large-uniform-array.

13 years agoi965: Update gen6 SF state when point state (sprite or attenuation) changes.
Eric Anholt [Thu, 28 Oct 2010 22:23:23 +0000 (15:23 -0700)]
i965: Update gen6 SF state when point state (sprite or attenuation) changes.

13 years agoi965: Add user clip planes support to gen6.
Eric Anholt [Wed, 27 Oct 2010 18:00:14 +0000 (11:00 -0700)]
i965: Add user clip planes support to gen6.

Fixes piglit user-clip, and compiz desktop switching when dragging a
window and using just 2 desktops.  Bug #30446.

13 years agogallivm: Remove the EMMS opcodes.
José Fonseca [Mon, 18 Oct 2010 13:24:30 +0000 (06:24 -0700)]
gallivm: Remove the EMMS opcodes.

Unnecessary now that lp_set_target_options() successful disables MMX code
emission.

13 years agogallivm: always enable LLVMAddInstructionCombiningPass()
José Fonseca [Mon, 25 Oct 2010 17:42:03 +0000 (18:42 +0100)]
gallivm: always enable LLVMAddInstructionCombiningPass()

13 years agogallium: Avoid using __doc__ in python scripts.
José Fonseca [Thu, 28 Oct 2010 16:38:18 +0000 (17:38 +0100)]
gallium: Avoid using __doc__ in python scripts.

13 years agost/mesa: Silence uninitialized variable warning.
Vinson Lee [Thu, 28 Oct 2010 13:08:19 +0000 (06:08 -0700)]
st/mesa: Silence uninitialized variable warning.

Fixes this GCC warning.
state_tracker/st_program.c: In function 'st_print_shaders':
state_tracker/st_program.c:735: warning: 'sh' may be used uninitialized in this function

13 years agor300/compiler: Use rc_get_readers_normal() for presubtract optimizations
Tom Stellard [Fri, 22 Oct 2010 01:55:12 +0000 (18:55 -0700)]
r300/compiler: Use rc_get_readers_normal() for presubtract optimizations

13 years agoi965: Add bit operation support to the fragment shader backend.
Kenneth Graunke [Tue, 19 Oct 2010 00:28:56 +0000 (17:28 -0700)]
i965: Add bit operation support to the fragment shader backend.

13 years agoi965: Make FS uniforms be the actual type of the uniform at upload time.
Eric Anholt [Mon, 25 Oct 2010 19:52:29 +0000 (12:52 -0700)]
i965: Make FS uniforms be the actual type of the uniform at upload time.

This fixes some insanity that would otherwise be required for GLSL
1.30 bit ops or gen6 integer uniform operations in general, at the
cost of upload-time pain.  Given that we only have that pain because
mesa's mangling our integer uniforms to be floats, this something that
should be fixed outside of the shader codegen.

13 years agodocs: add GL_EXT_separate_shader_objects to release notes
Ian Romanick [Wed, 27 Oct 2010 20:45:29 +0000 (13:45 -0700)]
docs: add GL_EXT_separate_shader_objects to release notes

13 years agointel: Enable GL_EXT_separate_shader_objects in Intel drivers
Ian Romanick [Wed, 13 Oct 2010 21:00:56 +0000 (14:00 -0700)]
intel: Enable GL_EXT_separate_shader_objects in Intel drivers

13 years agoswrast: Enable GL_EXT_separate_shader_objects in software paths
Ian Romanick [Wed, 13 Oct 2010 21:00:24 +0000 (14:00 -0700)]
swrast: Enable GL_EXT_separate_shader_objects in software paths

13 years agoTrack separate programs for each stage
Ian Romanick [Wed, 13 Oct 2010 20:58:44 +0000 (13:58 -0700)]
Track separate programs for each stage

The assumption is that all stages are the same program or that
varyings are passed between stages using built-in varyings.

13 years agomesa: Track an ActiveProgram distinct from CurrentProgram
Ian Romanick [Mon, 11 Oct 2010 23:07:08 +0000 (16:07 -0700)]
mesa: Track an ActiveProgram distinct from CurrentProgram

ActiveProgram is the GL_EXT_separate_shader_objects state variable
used for glUniform calls.  glUseProgram also sets this.

13 years agomesa: Add display list support for GL_EXT_separate_shader_objects functions
Ian Romanick [Wed, 13 Oct 2010 20:54:13 +0000 (13:54 -0700)]
mesa: Add display list support for GL_EXT_separate_shader_objects functions

13 years agomesa: Skeletal support for GL_EXT_separate_shader_objects
Ian Romanick [Mon, 11 Oct 2010 22:50:04 +0000 (15:50 -0700)]
mesa: Skeletal support for GL_EXT_separate_shader_objects

Really just filling in the entry points.  None of them do anything
other than validate their inputs.

13 years agomesa: Add infrastructure to track GL_EXT_separate_shader_objects
Ian Romanick [Mon, 11 Oct 2010 22:09:41 +0000 (15:09 -0700)]
mesa: Add infrastructure to track GL_EXT_separate_shader_objects

13 years agoglapi: Commit files changed by previous commit
Ian Romanick [Mon, 11 Oct 2010 22:07:43 +0000 (15:07 -0700)]
glapi: Commit files changed by previous commit

13 years agoglapi: Add GL_EXT_separate_shader_objects
Ian Romanick [Mon, 11 Oct 2010 22:04:59 +0000 (15:04 -0700)]
glapi: Add GL_EXT_separate_shader_objects

13 years agoFix build on systems where "python" is python 3.
Kenneth Graunke [Mon, 25 Oct 2010 20:52:58 +0000 (13:52 -0700)]
Fix build on systems where "python" is python 3.

First, it changes autoconf to use a "python2" binary when available,
rather than plain "python" (which is ambiguous).  Secondly, it changes
the Makefiles to use $(PYTHON) $(PYTHON_FLAGS) rather than calling
python directly.

Signed-off-by: Xavier Chantry <chantry.xavier@gmail.com>
Signed-off-by: Matthew William Cox <matt@mattcox.ca>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
13 years agor300g: add a default channel ordering of texture border for unhandled formats
Marek Olšák [Wed, 27 Oct 2010 19:13:41 +0000 (21:13 +0200)]
r300g: add a default channel ordering of texture border for unhandled formats

It should fix the texture border for compressed textures.
Broken since 8449a4772a73f613d9425b691cffba6a261df813.

13 years agor600c: add missing radeon_prepare_render() call on evergreen
Alex Deucher [Wed, 27 Oct 2010 18:30:50 +0000 (14:30 -0400)]
r600c: add missing radeon_prepare_render() call on evergreen

13 years agor100: revalidate after radeon_update_renderbuffers
Alex Deucher [Wed, 27 Oct 2010 17:53:29 +0000 (13:53 -0400)]
r100: revalidate after radeon_update_renderbuffers

This is a port of 603741a86df0e43c0b52e8c202a35c7fe2fc1d9c
to r100.

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
13 years agoswrast: Print out format on unexpected failure in _swrast_ReadPixels.
Vinson Lee [Wed, 27 Oct 2010 17:16:18 +0000 (10:16 -0700)]
swrast: Print out format on unexpected failure in _swrast_ReadPixels.

13 years agoegl: Remove unnecessary headers.
Vinson Lee [Wed, 27 Oct 2010 16:51:11 +0000 (09:51 -0700)]
egl: Remove unnecessary headers.

13 years agomesa: Remove unnecessary header.
Vinson Lee [Wed, 27 Oct 2010 16:38:33 +0000 (09:38 -0700)]
mesa: Remove unnecessary header.

13 years agost/mesa: Remove unnecessary header.
Vinson Lee [Wed, 27 Oct 2010 16:33:13 +0000 (09:33 -0700)]
st/mesa: Remove unnecessary header.

13 years agor600g: Silence uninitialized variable warnings.
Vinson Lee [Wed, 27 Oct 2010 16:26:27 +0000 (09:26 -0700)]
r600g: Silence uninitialized variable warnings.

13 years agomesa: Remove unnecessary headers.
Vinson Lee [Wed, 27 Oct 2010 16:09:47 +0000 (09:09 -0700)]
mesa: Remove unnecessary headers.

13 years agor300g: Silence uninitialized variable warning.
Vinson Lee [Wed, 27 Oct 2010 16:02:00 +0000 (09:02 -0700)]
r300g: Silence uninitialized variable warning.

Fixes this GCC warning.
r300_state_derived.c: In function 'r300_update_derived_state':
r300_state_derived.c:593: warning: 'r' may be used uninitialized in this function
r300_state_derived.c:593: note: 'r' was declared here

13 years agor600g: Destroy the blitter.
Tilman Sauerbeck [Sun, 24 Oct 2010 12:52:14 +0000 (14:52 +0200)]
r600g: Destroy the blitter.

This fix got lost in the state rework merge.

Signed-off-by: Tilman Sauerbeck <tilman@code-monkey.de>
13 years agor600g: In radeon_bo(), call LIST_INITHEAD early.
Tilman Sauerbeck [Sat, 23 Oct 2010 17:05:26 +0000 (19:05 +0200)]
r600g: In radeon_bo(), call LIST_INITHEAD early.

radeon_bo_destroy() will want to read the list field. Without this patch,
we'd end up evaluating the list pointers before they have been properly
set up when we destroyed the newly created bo if it cannot be mapped.

Signed-off-by: Tilman Sauerbeck <tilman@code-monkey.de>
13 years agomesa: Remove unnecessary glapitable.h includes.
Chia-I Wu [Mon, 25 Oct 2010 20:25:45 +0000 (04:25 +0800)]
mesa: Remove unnecessary glapitable.h includes.

With 07b85457d95bcc70588584e9380c51cd63aa3a2b, glapitable.h is included
by core mesa only to know the size of _glapi_table.  It is not necessary
as the same info is given by _gloffset_COUNT.

This change makes _glapi_table opaque to core mesa.  All operations on
it are supposed to go through one of the SET/GET/CALL macros.

13 years agovbo: Use CALL_* macros.
Chia-I Wu [Mon, 25 Oct 2010 16:35:35 +0000 (00:35 +0800)]
vbo: Use CALL_* macros.

Use macros to access _glapi_table consistently.  There is no functional
change.

13 years agoglapi: Merge glapioffsets.h into glapidispath.h.
Chia-I Wu [Mon, 25 Oct 2010 16:16:49 +0000 (00:16 +0800)]
glapi: Merge glapioffsets.h into glapidispath.h.

Move defines in glapioffsets.h to glapidispatch.h.  Rename
_gloffset_FIRST_DYNAMIC to _gloffset_COUNT, which is equal to the number
of entries in _glapi_table.

Consistently use SET_by_offset, GET_by_offset, CALL_by_offset, and
_gloffset_* to recursively define all SET/GET/CALL macros.

13 years agoglapi: Do not use glapioffsets.h.
Chia-I Wu [Mon, 25 Oct 2010 14:27:17 +0000 (22:27 +0800)]
glapi: Do not use glapioffsets.h.

glapioffsets.h exists for the same reason as glapidispatch.h does.  It
is of no use to glapi.  This commit also drops the use of glapioffsets.h
in glx as glx is considered an extension to glapi when it comes to
defining public GL entries.

13 years agomesa: rename function to _mesa_is_format_integer_color()
Brian Paul [Wed, 27 Oct 2010 02:30:40 +0000 (20:30 -0600)]
mesa: rename function to _mesa_is_format_integer_color()

Be a bit more clear about its operation.

13 years agomesa: fix bug in _mesa_is_format_integer()
Brian Paul [Wed, 27 Oct 2010 02:23:57 +0000 (20:23 -0600)]
mesa: fix bug in _mesa_is_format_integer()

We only want to return true if it's an integer _color_ format, not a
depth and/or stencil format.
Fixes http://bugs.freedesktop.org/show_bug.cgi?id=31143

13 years agoglapi: Move glapidispatch.h to core mesa.
Chia-I Wu [Mon, 25 Oct 2010 19:42:09 +0000 (03:42 +0800)]
glapi: Move glapidispatch.h to core mesa.

It is a core mesa header, not a glapi header.

13 years agoglapi: Do not use glapidispatch.h.
Chia-I Wu [Mon, 25 Oct 2010 14:05:20 +0000 (22:05 +0800)]
glapi: Do not use glapidispatch.h.

glapidispatch.h exists so that core mesa (libmesa.a) can be built for
DRI drivers or for non-DRI drivers as a compile time decision (whether
IN_DRI_DRIVER is defined).  It is of no use to glapi.  This commit also
drops the use of glapidispatch.h in glx and libgl-xlib as they are
considered extensions to glapi when it comes to defining public GL
entries.

13 years agomesa: remove the unused _mesa_is_fragment_shader_active() function
Brian Paul [Wed, 27 Oct 2010 00:04:28 +0000 (18:04 -0600)]
mesa: remove the unused _mesa_is_fragment_shader_active() function

This reverts commit 013d5ffeec3af5665c81c6a7a8370d21699ca609.

13 years agomesa: call _mesa_valid_to_render() in glDrawPixels, glCopyPixels, glBitmap
Brian Paul [Wed, 27 Oct 2010 00:01:50 +0000 (18:01 -0600)]
mesa: call _mesa_valid_to_render() in glDrawPixels, glCopyPixels, glBitmap

This lets us simplify and consolidate some state checking code.

This implements the GL_INVALID_OPERATION check for all drawing commands
required by GL_EXT_texture_integer.

13 years agomesa: do integer FB / shader validation check in _mesa_valid_to_render()
Brian Paul [Wed, 27 Oct 2010 00:01:06 +0000 (18:01 -0600)]
mesa: do integer FB / shader validation check in _mesa_valid_to_render()

13 years agoi965: Disable register spilling on gen6 until it's fixed.
Eric Anholt [Tue, 26 Oct 2010 21:49:38 +0000 (14:49 -0700)]
i965: Disable register spilling on gen6 until it's fixed.

Avoids GPU hang on glsl-fs-convolution-1.

13 years agoi965: Fix VS URB entry sizing.
Eric Anholt [Tue, 26 Oct 2010 21:36:18 +0000 (14:36 -0700)]
i965: Fix VS URB entry sizing.

I'm trying to clamp to a minimum of 1 URB row, not a maximum of 1.

Fixes:
glsl-kwin-blur
glsl-max-varying
glsl-routing

13 years agoi965: Drop the eot argument to read messages, which can never be set.
Eric Anholt [Tue, 26 Oct 2010 20:46:09 +0000 (13:46 -0700)]
i965: Drop the eot argument to read messages, which can never be set.

13 years agoi965: Add support for constant buffer loads on gen6.
Eric Anholt [Tue, 26 Oct 2010 20:17:54 +0000 (13:17 -0700)]
i965: Add support for constant buffer loads on gen6.

Fixes glsl-fs-uniform-array-5.

13 years agoi965: Set up the constant buffer on gen6 when it's needed.
Eric Anholt [Tue, 26 Oct 2010 20:15:01 +0000 (13:15 -0700)]
i965: Set up the constant buffer on gen6 when it's needed.

This was slightly confused because gen6_wm_constants does the push
constant buffer, while brw_wm_constants does pull constants.

13 years agoi965: Fix typo in comment about state flags.
Eric Anholt [Tue, 26 Oct 2010 19:19:46 +0000 (12:19 -0700)]
i965: Fix typo in comment about state flags.

13 years agoi965: Handle new ir_unop_round_even in channel expression splitting.
Eric Anholt [Tue, 26 Oct 2010 18:23:27 +0000 (11:23 -0700)]
i965: Handle new ir_unop_round_even in channel expression splitting.

13 years agoi965: Add support for discard instructions on gen6.
Eric Anholt [Mon, 25 Oct 2010 23:22:31 +0000 (16:22 -0700)]
i965: Add support for discard instructions on gen6.

It's a little more painful than before because we don't have the handy
mask register any more, and have to make do with cooking up a value
out of the flag register.

13 years agoi965: Add disasm for the flag register.
Eric Anholt [Tue, 26 Oct 2010 17:56:53 +0000 (10:56 -0700)]
i965: Add disasm for the flag register.

13 years agoi965: Clear some undefined fields of g0 when using them for gen6 FB writes.
Eric Anholt [Tue, 26 Oct 2010 17:31:08 +0000 (10:31 -0700)]
i965: Clear some undefined fields of g0 when using them for gen6 FB writes.

This doesn't appear to help any testcases I'm looking at, but it looks
like it's required.

13 years agoi965: Use SENDC on the first render target write on gen6.
Eric Anholt [Tue, 26 Oct 2010 16:35:34 +0000 (09:35 -0700)]
i965: Use SENDC on the first render target write on gen6.

This is apparently required, as the thread will be initiated while it
still has dependencies, and this is what waits for those to be
resolved before writing color.

13 years agoi965: Clarify an XXX comment in FB writes with real info.
Eric Anholt [Mon, 25 Oct 2010 23:22:04 +0000 (16:22 -0700)]
i965: Clarify an XXX comment in FB writes with real info.

13 years agoi965: Add EU code for dword scattered reads (constant buffer array indexing).
Eric Anholt [Mon, 25 Oct 2010 18:23:13 +0000 (11:23 -0700)]
i965: Add EU code for dword scattered reads (constant buffer array indexing).

13 years agoegl_dri2: Fix a typo that make glFlush be called at wrong time.
Chia-I Wu [Tue, 26 Oct 2010 07:00:56 +0000 (15:00 +0800)]
egl_dri2: Fix a typo that make glFlush be called at wrong time.

We want to call glFlush when there is a current context.  That is,
old_ctx.  This is a regression introduced by
d19afc57fe49816f3f3290410e0124d326577be2.

13 years agor600g: add assembler support for all the kcache fields.
Dave Airlie [Tue, 26 Oct 2010 02:08:00 +0000 (12:08 +1000)]
r600g: add assembler support for all the kcache fields.

13 years agomesa: additional teximage error checks for GL_EXT_texture_integer
Brian Paul [Tue, 26 Oct 2010 01:08:55 +0000 (19:08 -0600)]
mesa: additional teximage error checks for GL_EXT_texture_integer

13 years agomesa: additional switch cases for GL_EXT_texture_integer
Brian Paul [Tue, 26 Oct 2010 01:08:04 +0000 (19:08 -0600)]
mesa: additional switch cases for GL_EXT_texture_integer

13 years agomesa: additional glReadPixels error checks for GL_EXT_texture_integer
Brian Paul [Tue, 26 Oct 2010 01:07:33 +0000 (19:07 -0600)]
mesa: additional glReadPixels error checks for GL_EXT_texture_integer

13 years agor600g: fix magic 0x1 ->flat shade ena
Dave Airlie [Thu, 14 Oct 2010 01:15:37 +0000 (11:15 +1000)]
r600g: fix magic 0x1 ->flat shade ena

13 years agoglsl: Fix constant component count in vector constructor emitting.
Kenneth Graunke [Mon, 25 Oct 2010 19:44:55 +0000 (12:44 -0700)]
glsl: Fix constant component count in vector constructor emitting.

Fixes freedesktop.org bug #31101 as well as piglit test cases
assignment-type-mismatch.vert and constructor-28.vert.

13 years agoglsl: Fix ast-to-hir for ARB_fragment_coord_conventions
Chad Versace [Sat, 23 Oct 2010 17:40:40 +0000 (10:40 -0700)]
glsl: Fix ast-to-hir for ARB_fragment_coord_conventions

Function ast_declarator_list::hir(), when processing keywords added by
extension ARB_fragment_coord_conventions, made the mistake of checking only if
the extension was __supported by the driver__. The correct behavior is to check
if the extensi0n is __enabled in the parse state__.

NOTE: this is a candidate for the 7.9 branch.

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