glsl_to_tgsi: fix a reference leak on an error path

i965: Avoid generating MOVs for most ir_assignment handling.

This is a port of vec4_visitor::try_rewrite_rhs_to_dst to fs_visitor.

Not only is this technique less invasive and more robust, it also
generates better code.  Over and above the previous technique, this
reduced instruction count in shader-db by 0.28% on average and 1.4% in
the best case.

In no case did this technique result in more code than the prior method.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Eric Anholt <eric@anholt.net>

i965/fs: Revert "Avoid generating MOVs for assignments for expressions."

This reverts commit 53c89c67f33639afef951e178f93f4e29acc5d53, along with
the subsequent this->result = reg_undef additions it required.

Both Eric and I agree that the way he did this is really fragile; if you
forget to add this->result = reg_undef before calling accept(), it may
end up using the same register for two separate things, breaking things
in strange and mysterious ways.

The next commit will port over the new VS backend's method for solving
this problem, which is simpler, less intrusive, and still manages to
avoid MOVs in the common case.

mesa: Remove all mention of MESA_FORMAT_CI8

Nothing in Mesa supports color-index textures, and most of the other
infrastructure that could allow such support has already been removed.
This puts the final nail in the coffin.

Also clean out some GL_COLOR_INDEX comments in formats.c.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

dri: Remove MESA_FORMAT_CI8 bits from drivers that don't do paletted textures

This came from the "kill it with fire" discussion at XDS 2010.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

scons: add swrast/s_texture.c to the build

intel: Rely on Mesa core for the non-blit glTexSubImage* implementation.

It uses MapTextureImage() now, so we don't need our own mapping.

Reviewed-by: Brian Paul <brianp@vmware.com>

mesa: Convert texture debug dump function to using MapTextureImage().

Reviewed-by: Brian Paul <brianp@vmware.com>

intel: Remove our custom _mesa_store_compressed_texsubimage2d().

Now that Mesa core knows how to map teximages, we no longer needed the
compressed paths here.

Reviewed-by: Brian Paul <brianp@vmware.com>

mesa: Convert texstore.c to accessing textures using MapTextureImage.

This continues to allocate texImage->Data as before, so
drivers calling these functions need to use that when present.

Reviewed-by: Brian Paul <brianp@vmware.com>

nouveau: Add MapTextureImage() implementation.

This is untested, but should be close to working since it's basically
a copy of nouveau_teximage_map().

Reviewed-by: Brian Paul <brianp@vmware.com>

radeon: Add MapTextureImage() implementation.

Reviewed-by: Brian Paul <brianp@vmware.com>

radeon: Refactor the common texture hook setup to common code.

Reviewed-by: Brian Paul <brianp@vmware.com>

swrast: Add implementation of MapTextureImage/UnmapTextureImage.

Reviewed-by: Brian Paul <brianp@vmware.com>

intel: Add implementation of MapTextureImage/UnmapTextureImage.

Reviewed-by: Brian Paul <brianp@vmware.com>

st/mesa: Add implementation of MapTextureImage.

Reviewed-by: Brian Paul <brianp@vmware.com>

mesa: Add driver hooks for texture image mapping/unmapping.

ctx->Driver.MapTextureImage() / UnmapTextureImage() will be called by
the glTex[Sub]Image(), glGetTexImage() functions, etc. when we're
accessing texture data, and also for software rendering when accessing
texture data.

Reviewed-by: Brian Paul <brianp@vmware.com>

mesa: Don't check for image->Data when freeing an image's contents.

All driver implementations of FreeTextureImageBuffer already check
that Data != NULL and free it.  However, this means that we will also
free driver storage if the driver storage wasn't in the form of a Data
pointer.

This was produced by the following semantic patch:

@@
expression C;
expression T;
@@
- if (T->Data) {
- C->Driver.FreeTextureImageBuffer(C, T);
+ C->Driver.FreeTextureImageBuffer(C, T);
- }

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

Rename some driver FreeTextureImageData functions to FreeTextureImageBuffer.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

mesa: Rename FreeTexImageData to FreeTextureImageBuffer.

This was produced by sed, except for one hunk in driverfuncs.c where
trailing whitespace was dropped.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

i915g: remove unused var in i915_flush_heuristically()

winsys/g3dvl: Fix include style

As explained in the thread starting at [0], the internal include style
should be »#include "path/to/header.h"« for non-system includes.

[0]
<http://news.gmane.org/find-root.php?message_id=%3c4E5802BE.6020206%40vmware.com%3e>

Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Signed-off-by: Brian Paul <brianp@vmware.com>

tests/unit: Fix include style

As explained in the thread starting at [0], the internal include style
should be »#include "path/to/header.h"« for non-system includes.

[0]
<http://news.gmane.org/find-root.php?message_id=%3c4E5802BE.6020206%40vmware.com%3e>

Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Signed-off-by: Brian Paul <brianp@vmware.com>

targets/xorg-vmwgfx: Fix include style

As explained in the thread starting at [0], the internal include style
should be »#include "path/to/header.h"« for non-system includes.

[0]
<http://news.gmane.org/find-root.php?message_id=%3c4E5802BE.6020206%40vmware.com%3e>

Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Signed-off-by: Brian Paul <brianp@vmware.com>

st/xorg: Fix include style

As explained in the thread starting at [0], the internal include style
should be »#include "path/to/header.h"« for non-system includes.

[0]
<http://news.gmane.org/find-root.php?message_id=%3c4E5802BE.6020206%40vmware.com%3e>

Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Signed-off-by: Brian Paul <brianp@vmware.com>

vdpau: Fix include style

As explained in the thread starting at [0], the internal include style
should be »#include "path/to/header.h"« for non-system includes.

[0]
<http://news.gmane.org/find-root.php?message_id=%3c4E5802BE.6020206%40vmware.com%3e>

Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Signed-off-by: Brian Paul <brianp@vmware.com>

va: Fix include style

As explained in the thread starting at [0], the internal include style
should be »#include "path/to/header.h"« for non-system includes.

[0]
<http://news.gmane.org/find-root.php?message_id=%3c4E5802BE.6020206%40vmware.com%3e>

Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Signed-off-by: Brian Paul <brianp@vmware.com>

d3d1x: Fix include style

As explained in the thread starting at [0], the internal include style
should be »#include "path/to/header.h"« for non-system includes.

[0]
<http://news.gmane.org/find-root.php?message_id=%3c4E5802BE.6020206%40vmware.com%3e>

Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Signed-off-by: Brian Paul <brianp@vmware.com>

include/pipe: Fix include style

As explained in the thread starting at [0], the internal include style
should be »#include "path/to/header.h"« for non-system includes.

[0]
<http://news.gmane.org/find-root.php?message_id=%3c4E5802BE.6020206%40vmware.com%3e>

Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Signed-off-by: Brian Paul <brianp@vmware.com>

r600g: Fix include style

As explained in the thread starting at [0], the internal include style
should be »#include "path/to/header.h"« for non-system includes.

[0]
<http://news.gmane.org/find-root.php?message_id=%3c4E5802BE.6020206%40vmware.com%3e>

Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Signed-off-by: Brian Paul <brianp@vmware.com>

noop: Fix include style

As explained in the thread starting at [0], the internal include style
should be »#include "path/to/header.h"« for non-system includes.

[0]
<http://news.gmane.org/find-root.php?message_id=%3c4E5802BE.6020206%40vmware.com%3e>

Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Signed-off-by: Brian Paul <brianp@vmware.com>

vl: Fix include style

As explained in the thread starting at [0], the internal include style
should be »#include "path/to/header.h"« for non-system includes.

[0]
<http://news.gmane.org/find-root.php?message_id=%3c4E5802BE.6020206%40vmware.com%3e>

Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Signed-off-by: Brian Paul <brianp@vmware.com>

util: Fix include style

As explained in the thread starting at [0], the internal include style
should be »#include "path/to/header.h"« for non-system includes.

[0]
<http://news.gmane.org/find-root.php?message_id=%3c4E5802BE.6020206%40vmware.com%3e>

Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Signed-off-by: Brian Paul <brianp@vmware.com>

docs: Fix minor typos.

dbec3a5d introduced minor typos, this should fix them.

Signed-off-by: Kai Wasserbäch <kai@dev.carbon-project.org>
Signed-off-by: Brian Paul <brianp@vmware.com>

g3dvl: Fix a bug not decoding the last 32-64 bits of an mpeg2 bitstream.

Another bug found by Andy Furniss.

st/vdpau: Respect source_rect in VideoMixerRender

Fixing a bug reported by Andy Furniss.

egl: Use gbm/wayland flags regardless of egl_dri2

Since they are needed for display autodetection.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=40443

xvmc tests: Clean up test_rendering slightly

Signed-off-by: Maarten Lankhorst <m.b.lankhorst@gmail.com>

android: add support for egl_dri2

Add rules to build egl_dri2 and make it a built-in EGL driver of
libGLES_mesa.

Reviewed-by: Chad Versace <chad@chad-versace.us>

egl_dri2: add support for Android

Add platform_android.c that supports _EGL_PLAFORM_ANDROID.  It works
with drm_gralloc, where back buffers of windows are backed by GEM
objects.

In Android a native window has a queue of back buffers allocated by the
server, through drm_gralloc.  For each frame, EGL needs to

  dequeue the next back buffer
  render to the buffer
  enqueue the buffer

After enqueuing, the buffer is no longer valid to EGL.  A window has no
depth buffer or other aux buffers.  They need to be allocated locally by
EGL.

Reviewed-by: Benjamin Franzke <benjaminfranzke@googlemail.com>
Reviewed-by: Chad Versace <chad@chad-versace.us>
[olv: with assorted minor changes, mostly suggested during the review]

egl_dri2: allow RGBA masks to be specified for matching

Add rgba_masks to dri2_add_config.  When it is non-NULL, the DRI config
is accepted only when the offsets and sizes of the its channels match
rgba_mask.

Reviewed-by: Chad Versace <chad@chad-versace.us>

glsl_to_tgsi: remove unused code

android: make DRM optional

For BOARD_GPU_DRIVERS=swrast build, DRM is not needed.

android: add support for nouveau

Compile tested only.

android: add support for r300g

Compile tested only.

android: add support for i915g

Quickly tested with 945GME.  SurfaceFlinger (the display server and
compositor) works.  2D apps with RGB or RGBA visuals work.  As for 3D
apps, some work and some do not.

android: add support for vmwgfx

Quickly tested with VMWare Workstation 7.1.4 on Linux with GeForce
GT220.  SurfaceFlinger (the display server and compositor) works.  2D
apps with RGB visual works.  However, due to missing
PIPE_FORMAT_R8G8B8A8_UNORM support, those with RGBA visual do not.

winsys/i915: share the source list

Factor out C_SOURCES from Makefile to Makefile.sources, and let Makefile
and SConscript share it.

winsys/svga: share the source list

Factor out C_SOURCES from Makefile to Makefile.sources, and let Makefile
and SConscript share it.

nouveau: share the source lists

For each driver, factor out C_SOURCES from Makefile to Makefile.sources,
and let Makefile and SConscript share it.

r300g: share the source list

Factor out C_SOURCES from Makefile to Makefile.sources, and let Makefile
and SConscript share it.

Note that

  $(TOP)/src/glsl/ralloc.c and
  $(TOP)/src/mesa/program/register_allocate.c

are removed from C_SOURCES in Makefile.sources and added back in
Makefile and SConscript.  The idea is that they are not part of r300g.
But having them in libr300.a makes build non-GL targets such as the
compiler tests or g3dvl much easier.  Also, for practical reason, TOP
would be an undefined variable in Makefile.sources.

i915g: share the source list

Factor out C_SOURCES from Makefile to Makefile.sources, and let Makefile
and SConscript share it.

svga: share the source list

Factor out C_SOURCES from Makefile to Makefile.sources, and let Makefile
and SConscript share it.

targets/egl-static: fix nouveau and vmwgfx for Android

drmVersion and driver specific ioctls are used to get the PCI ID from a
DRM fd.  Eexpand the mechanism to nouveau and vmwgfx, except that for
nouveau, only the vendor ID is needed, and for vmwgfx, always assume
SVGA II.

pci_ids: add vmwgfx pci id list

There is only one chipset

 15ad:0405 VMware SVGA II Adapter

mapi: Commit generated files modified by previous commit

Some of the changes are spurious because somebody forgot to do this
when adding glFramebufferTextureLayerARB.

mapi: Silence many "warning: unused parameter"

When generating dispatch templates, emit the '(void) blah;' magic to
make GCC happy.  This reduces a lot of warning spam if you build with
-Wunused-parameter or -Wextra.

Reviewed-by: Chia-I Wu <olv@lunarg.com>

mesa/tnl_dd: Remove unused source tree mesa/tnl_dd/imm

Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Acked-by: Marek Olšák <maraeo@gmail.com>
Acked-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Acked-by: Jakob Bornecrantz <jakob@vmware.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Build-Tested-by: Jakob Bornecrantz <jakob@vmware.com>
Tested-by: Eugeni Dodonov <eugeni.dodonov@intel.com>

mesa/tnl_dd: Remove unused header file t_dd_vbtmp.h

Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Acked-by: Marek Olšák <maraeo@gmail.com>
Acked-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Acked-by: Jakob Bornecrantz <jakob@vmware.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Build-Tested-by: Jakob Bornecrantz <jakob@vmware.com>
Tested-by: Eugeni Dodonov <eugeni.dodonov@intel.com>

mesa/tnl_dd: Remove unused header file t_dd_rendertmp.h

Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Acked-by: Marek Olšák <maraeo@gmail.com>
Acked-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Acked-by: Jakob Bornecrantz <jakob@vmware.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Build-Tested-by: Jakob Bornecrantz <jakob@vmware.com>
Tested-by: Eugeni Dodonov <eugeni.dodonov@intel.com>

mesa: Remove support for BeOS

Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Acked-by: Marek Olšák <maraeo@gmail.com>
Acked-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Acked-by: Jakob Bornecrantz <jakob@vmware.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Build-Tested-by: Jakob Bornecrantz <jakob@vmware.com>
Tested-by: Eugeni Dodonov <eugeni.dodonov@intel.com>

dri: Remove unused header files mmx.h and spantmp.h

These header files were only used by drivers removed in a previous commit.

Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Acked-by: Marek Olšák <maraeo@gmail.com>
Acked-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Acked-by: Jakob Bornecrantz <jakob@vmware.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Build-Tested-by: Jakob Bornecrantz <jakob@vmware.com>
Tested-by: Eugeni Dodonov <eugeni.dodonov@intel.com>

dri: Remove driRenderbuffer::backBuffer field

The tdfx driver was the only user.

Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Acked-by: Marek Olšák <maraeo@gmail.com>
Acked-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Acked-by: Jakob Bornecrantz <jakob@vmware.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Build-Tested-by: Jakob Bornecrantz <jakob@vmware.com>
Tested-by: Eugeni Dodonov <eugeni.dodonov@intel.com>

dri: Remove all DRI1 drivers

Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Acked-by: Marek Olšák <maraeo@gmail.com>
Acked-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Acked-by: Jakob Bornecrantz <jakob@vmware.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Build-Tested-by: Jakob Bornecrantz <jakob@vmware.com>
Tested-by: Eugeni Dodonov <eugeni.dodonov@intel.com>

mesa: Remove obsolete Windows gldirect and ICD drivers

Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Acked-by: Marek Olšák <maraeo@gmail.com>
Acked-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Acked-by: Jakob Bornecrantz <jakob@vmware.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Build-Tested-by: Jakob Bornecrantz <jakob@vmware.com>
Tested-by: Eugeni Dodonov <eugeni.dodonov@intel.com>

mesa: Remove obsolete linux-fbdev software driver

Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Acked-by: Marek Olšák <maraeo@gmail.com>
Acked-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Acked-by: Jakob Bornecrantz <jakob@vmware.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Build-Tested-by: Jakob Bornecrantz <jakob@vmware.com>
Tested-by: Eugeni Dodonov <eugeni.dodonov@intel.com>

mesa: Remove stray, unused file

Acked-by: Kristian Høgsberg <krh@bitplanet.net>
Acked-by: Marek Olšák <maraeo@gmail.com>
Acked-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Acked-by: Jakob Bornecrantz <jakob@vmware.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Build-Tested-by: Jakob Bornecrantz <jakob@vmware.com>
Tested-by: Eugeni Dodonov <eugeni.dodonov@intel.com>

i965: Factor our source lists into Makefile.sources

In preparation for porting i965 to Android, factor its source lists into
a shared makefile. This prevents duplication of source lists, and hence
prevents the Android from breaking as often.

Acked-by: Chia-I Wu <olv@lunarg.com>
Signed-off-by: Chad Versace <chad@chad-versace.us>

Merge branch 'master' of git://anongit.freedesktop.org/mesa/mesa

g3dvl: use pointer_to_uintptr() to silence a cast warning

docs: Add a page on post-processing

With edits by Brian.

Signed-off-by: Lauri Kasanen <cand@gmx.com>
Signed-off-by: Brian Paul <brianp@vmware.com>

svga: include LLVM in name string in debug builds

pp: initialize the sample mask

We cannot rely on pipe drivers to default to non-zero.

Fixes pp being a no-op on nv50.

Reviewed-by: Lauri Kasanen <cand@gmx.com>

st/xorg: Fix solid fills for formats other than PICT_a8r8g8b8.

Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>

r600g: Hook up xorg state tracker.

Mostly copied from r300g.

Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

r600g: Handle PIPE_TRANSFER_MAP_DIRECTLY.

If the state tracker tries to map the resource directly but we can't or don't
want to do that, fail to create a transfer.

Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

st/xorg: Disable dirty throttling by default.

Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

st/xorg: Only damage non-front source in DRI2 CopyRegion hook.

Based on a vmwgfx xa/saa fix.

Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>

g3dvl: s/inline/INLINE/ to fix MSVC build

g3dvl: fix compilation failure on MSVC

I assume the intention of "mb = {}" was to zero-initialize it.

scons: don't compile some files with -gstabs if using mingw32

Compiling some (large) files with i686-pc-mingw32-gcc 4.2.2 (at least)
and the -gstabs option triggers a compiler error.  Use this work-around
to simply compile the effected files without -gstabs.

scons: add more LIBS for compiling with LLVM 2.9 on Windows

These extra libs shouldn't hurt with LLVM 2.8 or older.

g3dvl: Rewrite the mpeg 1&2 bitstream parser

Based on work of Maarten Lankhorst this time.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Younes Manton <younes.m@gmail.com>

g3dvl: Use a single texture for luma and chroma data

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Younes Manton <younes.m@gmail.com>

g3dvl: Rework the decoder interface part 5/5

Make setting the quant matrixes a generic interface.
Also removes setting the quant matrix from the XvMC interface

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Younes Manton <younes.m@gmail.com>

g3dvl: Rework the decoder interface part 4/5

Make the picture_structure enum spec complient.
Also remove it from the compositor.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Younes Manton <younes.m@gmail.com>

g3dvl: Rework the decoder interface part 3/5

Revert back to a macroblock based interface. The structure used
tries to keep as close to the spec as possible.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Younes Manton <younes.m@gmail.com>

g3dvl: Rework the decoder interface part 2/5

Implement PIPE_CAP_NUM_BUFFERS_DESIRED giving the decoder control over
the number of buffers a state tracker should allocate.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Younes Manton <younes.m@gmail.com>

g3dvl: Rework the decoder interface part 1/5

First of all get ride of the decode_buffer structure, while still giving
the decoder the ability to organize it's buffers depending on the needs
of the state tracker.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Younes Manton <younes.m@gmail.com>

tgsi: update tgsi.rst for TXQ

add some info on the TXQ opcode.

Signed-off-by: Dave Airlie <airlied@redhat.com>

i915g: Fix off-by-one in scissors.

i915g: Fix case where texcoords can overlap with fragpos/frontface.

i915g: Improve the flush heuristic by using the previous frame's number of vertices.

r600g: fix replace_gpr_with_pv_ps

Instructions with 3 source operands have no write mask, so we may replace their
destinations with PV/PS in the next group even if their dst.write is 0.

Note: This is a candidate for the 7.11 branch.

Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

r600g: fix check_and_set_bank_swizzle

Need to do full check when not all bank swizzles in the group are forced
(e.g. when trying to merge interp_* group with the next instruction)

Note: This is a candidate for the 7.11 branch.

Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

glcpp: Add GL_ARB_conservative_depth #define.

Forgotten in the patch that enabled the extension.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>

i965: Remove all bits of NRM3 and NRM4 code

Nothing in Mesa generates these opcodes, and i965 hardware cannot
support it natively.  If support were ever added for this opcode in
Mesa, there had better be a lowering pass for hardware that doesn't
support it natively.

glsl: fix crash when a const is passed to texelFetchOffset

while debugging texelFetchOffset we kept hitting the assert.

Signed-off-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

glsl: Bail after reporting an error for non-constant const_in parameters.

Otherwise we continue and hit the "Illegal formal parameter mode"
assertion.

Fixes negative compile test texelFetchOffset.frag in piglit.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>

pp: add files to Makefile.sources