Brian Paul [Thu, 1 Jul 2010 17:43:18 +0000 (11:43 -0600)]
gallium/cso: unbind sampler views in cso_release_all()
Brian Paul [Thu, 1 Jul 2010 17:07:32 +0000 (11:07 -0600)]
mesa: free xform feedback hash table
Brian Paul [Thu, 1 Jul 2010 17:07:12 +0000 (11:07 -0600)]
st/mesa: fix comment
Brian Paul [Thu, 1 Jul 2010 15:15:31 +0000 (09:15 -0600)]
llvmpipe: silence pointer type warnings
José Fonseca [Thu, 1 Jul 2010 12:57:48 +0000 (13:57 +0100)]
gallivm: Support 4 x unorm8 in lp_build_fetch_rgba_aos().
Uses code and ideas from Brian Paul.
José Fonseca [Thu, 1 Jul 2010 11:33:34 +0000 (12:33 +0100)]
gallivm: Allow to conversions to/from registers of different sizes.
Allow for example to convert from 4 x float32 to 4 x unorm8 and vice versa.
Uses code and ideas from Brian Paul.
José Fonseca [Thu, 1 Jul 2010 11:16:09 +0000 (12:16 +0100)]
gallivm: Support extended swizzles in lp_build_swizzle1_aos().
And rename to lp_build_swizzle_aos().
José Fonseca [Thu, 10 Jun 2010 15:25:39 +0000 (16:25 +0100)]
gallivm: Setup a global optimization pass.
Modules are still free to setup their own optimization passes, but for the
normal case it should not be necessary.
José Fonseca [Thu, 1 Jul 2010 09:14:46 +0000 (10:14 +0100)]
llvmpipe: Remove lp_build_swizzle2_aos().
Unnecessary special case.
Brian Paul [Wed, 30 Jun 2010 17:56:56 +0000 (11:56 -0600)]
llvmpipe: another null pointer check
Brian Paul [Wed, 30 Jun 2010 17:57:14 +0000 (11:57 -0600)]
llvmpipe: use dummy tile when out of memory
Brian Paul [Wed, 30 Jun 2010 17:49:40 +0000 (11:49 -0600)]
llvmpipe: added new lp_memory.[ch] files
Functions for using dummy tiles when we detect OOM conditions.
Brian Paul [Wed, 30 Jun 2010 17:46:17 +0000 (11:46 -0600)]
draw: fix out of memory handling in polygon stipple stage
Neil Roberts [Wed, 30 Jun 2010 11:41:11 +0000 (12:41 +0100)]
glu: Fix some compiler warnings in libtess
When compiled with the more aggressive compiler warnings such as
-Wshadow and -Wempty-body the libtess code gives a lot more
warnings. This fixes the following issues:
* The 'Swap' macro tries to combine multiple statements into one and
then consume the trailing semicolon by using if(1){/*...*/}else.
This gives warnings because the else part ends up with an empty
statement. It also seems a bit dangerous because if the semicolon
were missed then it would still be valid syntax but it would just
ignore the following statement. This patch replaces it with the more
common idiom do { /*...*/ } while(0).
* 'free' was being used as a local variable name but this shadows the
global function. This has been renamed to 'free_handle'
* TRUE and FALSE were being unconditionally defined. Although this
isn't currently a problem it seems better to guard them with #ifndef
because it's quite common for them to be defined in other headers.
https://bugs.freedesktop.org/show_bug.cgi?id=28845
Signed-off-by: Brian Paul <brianp@vmware.com>
Chia-I Wu [Wed, 30 Jun 2010 10:49:36 +0000 (18:49 +0800)]
egl: Update MaxSize when a dynamic array is grown.
Chia-I Wu [Wed, 30 Jun 2010 10:27:22 +0000 (18:27 +0800)]
egl: Make _eglUnloadDrivers no-op on Windows.
Windows unloads DLLs before atexit. Make _eglUnloadDrivers no-op on
Windows for now.
Chia-I Wu [Wed, 30 Jun 2010 10:21:09 +0000 (18:21 +0800)]
st/egl: Manually free configs on terminate.
The configs should be FREE()ed, not free()ed. We cannot rely on
_eglCleanupDisplay here.
Chia-I Wu [Wed, 30 Jun 2010 10:02:23 +0000 (18:02 +0800)]
egl: Store screens in a dynamic array.
Chia-I Wu [Wed, 30 Jun 2010 09:10:10 +0000 (17:10 +0800)]
egl: Store configs in a dynamic array.
Chia-I Wu [Wed, 30 Jun 2010 08:08:52 +0000 (16:08 +0800)]
egl: Add dynamic array.
Dynamic arrays will be used to store configs and screens of a display.
Chia-I Wu [Wed, 30 Jun 2010 06:42:00 +0000 (14:42 +0800)]
st/vega: Match MALLOC/FREE for vg_shader.
A vg_shader is destroyed with FREE.
José Fonseca [Wed, 30 Jun 2010 10:10:33 +0000 (11:10 +0100)]
llvmpipe: Add a new scene state to describe scenes which only have state changes.
It's a rare condition, but it may happen if all primitives are
clipped/culled.
For now we just do a no-op rasterization, but we could bypass it.
José Fonseca [Thu, 24 Jun 2010 19:57:21 +0000 (20:57 +0100)]
llvmpipe: Don't reset the bin when there's a zsbuf bound.
The previous rendering may have secondary effects on the zsbuf.
Fixes the missing tiles on gearbox.
nobled [Wed, 30 Jun 2010 06:24:13 +0000 (14:24 +0800)]
st/vega: s/free/FREE for matching MALLOC/CALLOC
[Manually fix a conflict in vg_context.c by Chia-I Wu]
nobled [Thu, 24 Jun 2010 01:32:41 +0000 (21:32 -0400)]
st/xorg: s/free/FREE for matching MALLOC/CALLOC
Chia-I Wu [Wed, 30 Jun 2010 05:15:18 +0000 (13:15 +0800)]
st/egl: Add egl-gdi target.
The target supports OpenVG on Windows with software rasterizer. The
egl_g3d_loader defined by the target supports arbitrary client APIs and
window systems. It is the SConscript that limits the support to OpenVG
and GDI.
This commit also fixes a typo in gdi backend.
Chia-I Wu [Wed, 30 Jun 2010 04:58:36 +0000 (12:58 +0800)]
mesa: Use fpclassify for GL_OES_query_matrix on OpenBSD and NetBSD.
Patch from Brad Smith <brad@comstyle.com>
The attached patch allows the GL_OES_query_matrix function to use the
systems fpclassify() for OpenBSD and NetBSD.
Zack Rusin [Wed, 30 Jun 2010 01:45:41 +0000 (21:45 -0400)]
mesa: make the arguments in the asm statemants optional
geometry shaders emit/end functions don't take any arguments
Marek Olšák [Tue, 29 Jun 2010 22:02:27 +0000 (00:02 +0200)]
r300g: add workaround for multiple contexts
Marek Olšák [Tue, 29 Jun 2010 21:34:36 +0000 (23:34 +0200)]
r300g: move one flush from winsys to the context
This flush happens when changing the tiling flags, and it should really be
done in the context.
I hope this fixes FDO bug #28630.
Brian Paul [Tue, 29 Jun 2010 21:40:15 +0000 (15:40 -0600)]
llvmpipe: don't crash/assert on out of memory
Check for null pointers and return early, etc.
Brian Paul [Tue, 29 Jun 2010 21:18:58 +0000 (15:18 -0600)]
llvmpipe: restore call to lp_setup_update_state()
This undoes part of commit
8be645d53a0d5d0ca50e4e9597043225e2231b6d
and fixes fd.o bug 28822 as well as other regressions.
The 'draw' module may issue additional state-change commands while
we're inside the draw_arrays/elements() call so it's important to
check for updated state at this point.
Jakob Bornecrantz [Tue, 29 Jun 2010 17:21:00 +0000 (18:21 +0100)]
egl: Check for drm winsys not just the driver name
Also fix pipe_radeon.so not building
José Fonseca [Tue, 29 Jun 2010 13:30:44 +0000 (14:30 +0100)]
draw: Add inline keyword to inline function.
Otherwise gcc will warn about unusued functions.
Jakob Bornecrantz [Tue, 29 Jun 2010 11:36:40 +0000 (12:36 +0100)]
i915g: Fix copy pasto
Jakob Bornecrantz [Tue, 29 Jun 2010 11:28:41 +0000 (12:28 +0100)]
egl: Fix gallium build when EGL is not installed on system
Jakob Bornecrantz [Tue, 29 Jun 2010 11:28:15 +0000 (12:28 +0100)]
i915g: Add missing egl pipe file
Chia-I Wu [Thu, 24 Jun 2010 16:41:56 +0000 (00:41 +0800)]
egl: Remove st/egl probe code.
It is no longer needed.
Chia-I Wu [Thu, 24 Jun 2010 08:49:40 +0000 (16:49 +0800)]
egl: Do not call drv->Initialize with global mutex locked.
Chia-I Wu [Wed, 23 Jun 2010 13:36:20 +0000 (21:36 +0800)]
st/egl: Build a single EGL driver.
This change makes st/egl build a single egl_gallium.so and multiple
st_<API>.so and pipe_<HW>.so. When a display is initialized, the
corresponding pipe driver will be loaded. When a context is created,
the corresponding state tracker will be loaded.
Unlike DRI drivers, no ABI compatibility is maintained. egl_gallium,
pipe drivers and state trackers should always be distributed as a single
package. As such, there is only a single src/gallium/targets/egl/ that
builds everything for the package.
Chia-I Wu [Tue, 29 Jun 2010 06:58:33 +0000 (14:58 +0800)]
st/egl: Move module loading code to targets.
Several changes are made. libegl.a no longer defines _eglMain. It
defines functions to create and destroy a _EGLDriver instead. The
creation function is called by the targets. It takes an egl_g3d_loader
as its argument. The loader is defined by the targets and is in charge
of creating st_api and pipe_screen. This allows us to move the module
loading code to targets. Lastly, the modules are now loaded as the
respective contexts are created.
Chia-I Wu [Tue, 29 Jun 2010 06:04:27 +0000 (14:04 +0800)]
st/egl: Reorganize targets.
Merge all targets into targets/egl/. The target produces
egl_gallium_<HW>.so for each pipe driver and st_<API>.so for each client
APIs. This enables us to further merge egl_gallium_<HW>.so into
egl_gallium.so later.
Chia-I Wu [Thu, 17 Jun 2010 15:45:41 +0000 (23:45 +0800)]
st/egl: One driver per hardware.
Merge multiple egl_<platform>_<pipe>.so into a single
egl_gallium_<pipe>.so. The environment variable EGL_PLATFORM is now
used to modify the return value of _eglGetNativePlatform.
Chia-I Wu [Thu, 17 Jun 2010 15:21:43 +0000 (23:21 +0800)]
st/egl: Introduce native_platform.
Move native_get_name, native_create_probe, native_get_probe_result, and
native_create_display into struct native_platform, and add
native_get_platform to get a handle to the struct.
Chia-I Wu [Wed, 23 Jun 2010 09:45:06 +0000 (17:45 +0800)]
st/egl: Make api_OpenGL support OpenGL ES.
This allows api_OpenGL.so to support OpenGL ES.
Chia-I Wu [Wed, 23 Jun 2010 09:40:49 +0000 (17:40 +0800)]
st/mesa: Add support for multiple APIs.
Add st_gl_api_create_es1 and st_gl_api_create_es2 to create OpeGL ES 1.1
and OpenGL ES 2.0 contexts respectively.
Chia-I Wu [Wed, 23 Jun 2010 08:14:49 +0000 (16:14 +0800)]
st_api: Remove st_context::is_visual_supported.
The callback is used by st/vega to check if a visual specifies the
depth/stencil format. It forces st/vega to be loaded by st/egl to
perform the check. As noted in EGL spec, the depth/stencil format of a
visual should not affect OpenVG. It should be better to ignore the
field and always allocate the depth/stencil texture.
Zack Rusin [Mon, 28 Jun 2010 21:31:21 +0000 (17:31 -0400)]
mesa: initial support for ARB_geometry_shader4
laying down the foundation for everything and implementing most of the
stuff.
linking, gl_VerticesIn and multidimensional inputs are left.
Marek Olšák [Mon, 28 Jun 2010 23:52:09 +0000 (01:52 +0200)]
util: reference surfaces and sampler views in blitter when saving them
Ooops. This should possibly fix some bugs...
Jakob Bornecrantz [Mon, 28 Jun 2010 20:33:39 +0000 (22:33 +0200)]
i915g: Add Galahad to targets
Jakob Bornecrantz [Mon, 28 Jun 2010 20:27:51 +0000 (22:27 +0200)]
r300g: Galahad for scons as well
Jakob Bornecrantz [Mon, 28 Jun 2010 19:54:38 +0000 (21:54 +0200)]
r300g: Add target defines for Galahad
Jakob Bornecrantz [Mon, 28 Jun 2010 20:22:53 +0000 (22:22 +0200)]
glhd: Build with scons
Jakob Bornecrantz [Mon, 28 Jun 2010 19:50:51 +0000 (21:50 +0200)]
glhd: Re-integrate with the debug system
Jakob Bornecrantz [Mon, 28 Jun 2010 19:37:03 +0000 (21:37 +0200)]
glhd: Remove leftover file
Jakob Bornecrantz [Mon, 28 Jun 2010 18:59:09 +0000 (20:59 +0200)]
Merge branch 'gallium-drm-driver-drescriptor'
Conflicts:
src/gallium/state_trackers/egl/x11/native_dri2.c
src/gallium/state_trackers/egl/x11/native_x11.c
src/gallium/state_trackers/egl/x11/native_x11.h
src/gallium/state_trackers/xorg/xorg_driver.c
src/gallium/winsys/radeon/drm/radeon_drm.c
José Fonseca [Mon, 28 Jun 2010 15:33:52 +0000 (16:33 +0100)]
llvmpipe: set WRITE_ALL only a per-tile basis in lp_resource_copy().
José Fonseca [Mon, 28 Jun 2010 14:39:54 +0000 (15:39 +0100)]
llvmpipe: Actually flush in lp_resource_copy()
The cpu_access is redundant in a software rasterizer.
José Fonseca [Mon, 28 Jun 2010 15:28:55 +0000 (16:28 +0100)]
llvmpipe: Ensure outdated framebuffer state is not reused in lp_setup_bind_framebuffer().
We were starting a scene whenever lp_setup_get_vertex_info() was called by
the draw module. So when when all primitives were culled/clipped, not only
did we create a new scene for nothing, but we end up using the old scene
with the old framebuffer state instead of a new one.
Fix consists in:
- don't call lp_setup_update_state() in lp_setup_get_vertex_info() -- no
longer necessary
- always setting the scene state before binning a command -- query
commands were bypassing it
- assert no old scene is reused in lp_setup_bind_framebuffer()
Marek Olšák [Sun, 27 Jun 2010 15:15:23 +0000 (17:15 +0200)]
r300g: fix a compile error on non-debug builds
Marek Olšák [Sun, 27 Jun 2010 13:02:08 +0000 (15:02 +0200)]
util: fix a memory leak in blitter
Marek Olšák [Sun, 27 Jun 2010 11:55:59 +0000 (13:55 +0200)]
r300g: separate the hyperz state and pipelined FB regs out of the FB state
Marek Olšák [Sat, 26 Jun 2010 23:07:34 +0000 (01:07 +0200)]
r300g: release referenced objects in destroy_context
Marek Olšák [Sat, 26 Jun 2010 22:39:27 +0000 (00:39 +0200)]
r300g: reference surfaces in set_framebuffer_state
Marek Olšák [Sat, 26 Jun 2010 17:09:19 +0000 (19:09 +0200)]
r300g: group debug messages
Marek Olšák [Fri, 25 Jun 2010 02:48:02 +0000 (04:48 +0200)]
r300g: turn rasterizer state into a command buffer
Marek Olšák [Fri, 25 Jun 2010 01:28:09 +0000 (03:28 +0200)]
r300g: move emission of the MSPOS regs into the framebuffer state
Now the question is whether we are allowed to ignore gl_rasterization_rules and
pipe_rasterizer_state::multisample. The former is invariant anyway and
I think the latter would need re-emitting the AA state which is quite costly,
considering that it implicitly flushes the whole pipeline (all AA regs
in the AA state are *unpipelined*).
Marek Olšák [Fri, 25 Jun 2010 00:55:14 +0000 (02:55 +0200)]
r300g: turn invariant state into a command buffer
Marek Olšák [Fri, 25 Jun 2010 00:37:06 +0000 (02:37 +0200)]
r300g: initialize some crucial state in the first CS
Marek Olšák [Fri, 25 Jun 2010 00:16:43 +0000 (02:16 +0200)]
r300g: introduce VAP invariant state
Unlike other invariant states, this one must be emitted after VAP flush.
Marek Olšák [Tue, 22 Jun 2010 05:54:21 +0000 (07:54 +0200)]
r300g: immediate mode cleanup
Marek Olšák [Sat, 26 Jun 2010 15:41:42 +0000 (17:41 +0200)]
r300g: update my notes about fastfill and zbuffer compression
Marek Olšák [Sat, 26 Jun 2010 15:35:56 +0000 (17:35 +0200)]
r300g: enum r300_blitter_op is a bitmask
Marek Olšák [Sat, 26 Jun 2010 12:00:24 +0000 (14:00 +0200)]
draw: fix build with llvm and make
Vinson Lee [Sat, 26 Jun 2010 07:47:52 +0000 (00:47 -0700)]
mesa: Remove unnecessary header.
Vinson Lee [Sat, 26 Jun 2010 07:38:05 +0000 (00:38 -0700)]
st/mesa: Remove unnecessary headers.
Vinson Lee [Sat, 26 Jun 2010 07:31:57 +0000 (00:31 -0700)]
draw: Remove unnecessary header.
Zack Rusin [Sat, 26 Jun 2010 00:06:53 +0000 (20:06 -0400)]
draw: deleted by mistake
Zack Rusin [Fri, 25 Jun 2010 23:58:09 +0000 (19:58 -0400)]
draw: initialize headers in the none post paths as well
Zack Rusin [Fri, 25 Jun 2010 23:31:09 +0000 (19:31 -0400)]
draw: limit the number of vertex shader variants kept around
we used to create and cache unltimited number of variant, this
change limits the number of variants kept around to a fixed number.
the change is based on a similar patch by Roland for llvmpipe fragment
shaders.
Zack Rusin [Wed, 23 Jun 2010 22:13:25 +0000 (18:13 -0400)]
draw: initialize vertex header
Brian Paul [Fri, 25 Jun 2010 20:12:30 +0000 (14:12 -0600)]
gallium/util: fix pipe_sampler_view_reference() calls
The conditionals aren't needed.
Brian Paul [Fri, 25 Jun 2010 15:39:43 +0000 (09:39 -0600)]
softpipe: rename a var
Chia-I Wu [Fri, 25 Jun 2010 10:32:41 +0000 (18:32 +0800)]
draw: Work around an invalid write.
SSE vertex shader does not seem to honor the execution mask. Pad the
output array as a workaround.
Brian Paul [Thu, 24 Jun 2010 20:10:24 +0000 (14:10 -0600)]
mesa: initialize extension string when context is first bound
...instead of waiting until glGetString(GL_EXTENSIONS) is called.
This fixes a problem where the MESA_EXTENSION_OVERRIDE env var is
ignored if the app never calls glGetString(GL_EXTENSIONS).
NOTE: this is a candidate patch for the 7.8 branch.
Brian Paul [Thu, 24 Jun 2010 18:45:29 +0000 (12:45 -0600)]
llvmpipe: fix comment typo
Brian Paul [Thu, 24 Jun 2010 18:45:14 +0000 (12:45 -0600)]
softpipe: fix comment typo
Marek Olšák [Thu, 24 Jun 2010 20:19:44 +0000 (22:19 +0200)]
r300g: a couple more chipsets do not have HiZ RAM
Corbin Simpson [Thu, 24 Jun 2010 14:54:00 +0000 (07:54 -0700)]
gallium/docs: Vertex data formats.
I'm not sure if I really got it right. This seems like one of those
"Duh, of course it works that way" things, but I'd like the
documentation to be readable by people not acquainted with OGL/D3D.
Corbin Simpson [Thu, 24 Jun 2010 14:18:59 +0000 (07:18 -0700)]
glhd: Add test for logicop enable.
Only for first RT at the moment, as there is no trivial way in galahad
to look at framebuffer state and (sadly) people don't usually calloc
their CSOs, so flags could be wrongly set.
On the other hand, of course, galahad will hopefully encourage more
people to calloc their CSOs. :3
Corbin Simpson [Thu, 24 Jun 2010 14:13:10 +0000 (07:13 -0700)]
gallium/docs: Lops override the rest of the blending state when enabled.
Brian Paul [Thu, 24 Jun 2010 13:21:15 +0000 (07:21 -0600)]
gallium/docs: updated remaining semantic label docs
nobled [Thu, 24 Jun 2010 01:31:28 +0000 (21:31 -0400)]
pipe: Add PIPE_OS_HURD
One tiny step toward porting Gallium to the GNU/Hurd kernel
(and fixing Debian bug #585618).
Signed-off-by: Corbin Simpson <MostAwesomeDude@gmail.com>
Corbin Simpson [Wed, 23 Jun 2010 21:25:26 +0000 (14:25 -0700)]
gallium/docs: Add lop table.
Was feeling kind of weird without it.
Corbin Simpson [Wed, 23 Jun 2010 20:57:30 +0000 (13:57 -0700)]
gallium/docs: Slight clarification and formatting for Blend.
Brian Paul [Thu, 24 Jun 2010 04:30:26 +0000 (22:30 -0600)]
Merge branch 'shader-file-reorg'
1. Move all GL entrypoint functions and files into src/mesa/main/
This includes the ARB vp/vp, NV vp/fp, ATI fragshader and GLSL bits
that were in src/mesa/shader/
2. Move src/mesa/shader/slang/ to src/mesa/slang/ to reduce the tree depth
3. Rename src/mesa/shader/ to src/mesa/program/ since all the
remaining files are concerned with GPU programs.
4. Misc code refactoring. In particular, I got rid of most of the
GLSL-related ctx->Driver hook functions. None of the drivers used
them.
Conflicts:
src/mesa/drivers/dri/i965/brw_context.c
Marek Olšák [Thu, 24 Jun 2010 04:15:36 +0000 (06:15 +0200)]
r300g: remove an XXX comment
Marek Olšák [Thu, 24 Jun 2010 03:54:18 +0000 (05:54 +0200)]
r300g: move AA registers into the new AA state
Marek Olšák [Thu, 24 Jun 2010 02:43:40 +0000 (04:43 +0200)]
r300g: separate the cache flush from the framebuffer state
Marek Olšák [Thu, 24 Jun 2010 01:52:45 +0000 (03:52 +0200)]
r300g: reorder and cleanup register writes everywhere