i965: Combine src/dest tex vs. rb checks in intel_copy_image_sub_data.

This simplifies things a little - now we only have one (tex or rb?)
if-ladder for src, and a second for dst, rather than four.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>

i965: Account for MinLayer in CopyImageSubData's blitter/CPU paths.

Fixes Piglit's arb_copy_image-texview test with the Meta path disabled
(so we hit the blitter/CPU fallback paths).

v2: Add MinLayer even for cube maps (suggested by Ilia).

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>

freedreno/ir3: cmdline compiler for glsl

Use glsl/libstandalone.la to add support for taking glsl src files (in
addition to .tgsi) as input.  Then glsl->nir and feed the result into
the ir3 backend as normal.

Signed-off-by: Rob Clark <robclark@freedesktop.org>

glsl: split out libstandalone

Split standalone glsl_compiler into a libstandalone.la and a thin
main.cpp.  This way drivers can re-use the glsl standalone frontend in
their own standalone compilers.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

android: drop build of standalone glsl_compiler

It's only a tool for debugging the glsl compiler, and should not be
installed.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Tested-by: Rob Herring <robh@kernel.org>
Acked-by: Emil Velikov <emil.l.velikov@gmail.com>

i965: Mark fallthrough in switch statement.

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

i965: Assert that a depth_mt exists when using HiZ.

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

nir: Strengthen assertion that 'out' is nonnull.

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

spirv: Mark default cases unreachable().

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

isl: Mark default cases unreachable.

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

isl: Remove useless qualifier from return type.

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

nvc0: add descriptions for hardware perf counters/metrics

The GALLIUM_HUD does not yet expose a description for each events, but
this might be useful for developers who want to have a long description
of hw perf counters directly in the source code.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>

mesa: 80-column wrapping for _context_lost_GetSynciv()

Reviewed-by: Kristian Høgsberg Kristensen <krh@bitplanet.net>

mesa: add GLAPIENTRY to new _context_lost_X functions

To fix MSVC build.  Any function which goes into the dispatch table
needs to have the GLAPIENTRY (__stdcall) tag.

Reviewed-by: Kristian Høgsberg Kristensen <krh@bitplanet.net>

scons: support 2.5.0

The get_implicit_deps changed in SCons 2.5, expecting a callable rather
than a path as third argument. Detect the SCons versions and set the
argument appropriately to support both 2.5 and earlier versions.

This closes #95211.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95211
Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com>
Cc: mesa-stable@lists.freedesktop.org
Acked-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

scons: whitespace cleanup

This text transformation was done automatically via the following shell
command:

$ find -name SCons\* -exec sed -i s/\\s\\+$// '{}' \;

Signed-off-by: Giuseppe Bilotta <giuseppe.bilotta@gmail.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

i965/fs: take into account doubles when emitting system values

Fixes the following cts test:
GL42-CTS.vertex_attrib_64bit.limits_test

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965: Fix shadowing of 'height' parameter

The nested declaration of 'height' shadows a parameter and uses
uninitialized memory. Fix by renaming to 'plane_height' which also makes
the code clearer.

This would typically break the bo size computation, but we don't use
that except when mmaping, and we don't mmap YUV buffers much.

Signed-off-by: Kristian Høgsberg Kristensen <krh@bitplanet.net>
Reported-by: Mathias Fröhlich <Mathias.Froehlich@gmx.net>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

mesa: Add .gitignore entries for make check binaries

Signed-off-by: Kristian Høgsberg Kristensen <krh@bitplanet.net>
Acked-by: Matt Turner <mattst88@gmail.com>

i965: Enable GL_KHR_robustness

GL_KHR_robustness adds the GL_CONTEXT_LOST error and five new entry
points that we already implement.  This patch adds a new dispatch table
that returns GL_CONTEXT_LOST from all entry points and implements the
GL_LOSE_CONTEXT_ON_RESET strategy by setting that table when we learn
that we've lost the context.

With the GL_CONTEXT_LOST reporting in place and dispatch for the new
entry points we can turn on GL_KHR_robustness.

Signed-off-by: Kristian Høgsberg Kristensen <krh@bitplanet.net>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Acked-by: Ilia Mirkin <imirkin@alum.mit.edu>

.mailmap: Use Chia-I Wu personal e-mail.

The LunarG one is bouncing.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

.mailmap: Use my (Emil Velikov) personal e-mail.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>

docs: add missing GL_OES/EXT_gpu_shader5 enablement note

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>

glsl: add GL_EXT_clip_cull_distance define, add helpers

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Tobias Klausmann <tobias.johannes.klausmann@mni.thm.de>

tgsi: print TGSI_PROPERTY_NEXT_SHADER value as string, not an integer

Print "GEOM" instead of "2", for example.

v2: also update the text parsing code, per Ilia.

Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

tgsi: s/6/PIPE_SHADER_TYPES/ for tgsi_processor_type_names array size

Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

nir/spirv: Handle location decorations on structure members

nir/spirv: Add explicit handling for all decorations

From time to time we have had cases where glslang has added a decoration we
don't handle and it has caused problems.  This audit ensures that, for
every decoration, we either handle it or hit an unreachable() with an
accurate description of why we don't have to.

i965/draw: Use the correct buffer index for interleaved VBO sizes

The buffer_range_* arrays are indexed by buffer index not element index.

Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>

i965/gen7: Fix gl_HelperInvocation

It appears that UV immediates aren't working on Ivy Bridge. In this
case, a signed version will work, and this fixes the piglit
tests/spec/glsl-4.50/execution/helper-invocation.shader_test test.

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

mesa_glinterop: make GL interop version field bidirectional

This allows clear and easy communication between the two.

Caller: Requesting information (struct vN)
Callee: I know how to deal with older version (vN-1) only. Here is your
data and the version I support.
Caller: Older version ? Sure I'll cap all access to the fields provided
by the older version (vN-1)

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>

mesa_glinterop: drop mesa_glinterop_device_info::interop_version

One cannot use a single version to control both export_in and export_out
versions. Using this forces us to always extend/bump both structs at the
same time.

An alternative scheme is coming with next patch.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>

st/dri: add note about GL interop version checks

... and make them more explicit.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>

mesa_glinterop: rename MESA_GLINTEROP_INVALID_{VALUE,VERSION}

Be more explicit what it actually does.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>

mesa_glinterop: s/struct_version/version/

OCD polish for consistency with other mesa interfaces.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>

mesa_glinterop: fix GL interop *_VERSION comments

Using the macro to set the version is wrong and ill-advised. Please don't
do it.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>

mesa_glinterop: remove inclusion of EGL header

Analogous to previous commit, but for EGL.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>

mesa_glinterop: remove inclusion of GLX header

Since we only need partial information about the GLX symbols we can
forward declare them and drop the include. Obviously each user of the
said API will needs more than what's provides, so they'll include the
GLX header.

If they don't, the compiler will give us a nice warning ;-)

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>

mesa_glinterop: remove unneeded GLAPI/GLAPIENTRY/APIENTRYP symbols

These come from windows.h, gl.h, glcorearb.h and/or glext.h.

The interop interface is aimed at non-Windows platforms while the macros
are used/derived due to Windows specifics. Thus we can safely remove
them.

Strictly speaking there should be GLXAPIENTRY/EGLAPIENTRY and alike
macros, although a) there is no GLX ones and b) this brings us even
further from decoupling the file from the GLX/EGL header dependency.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>

mesa_glinterop: replace GL types with their native counterpart.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>

mesa_glinterop: use generic variable types for the GL interop

Thus we can preserve the ABI, while avoiding the inclusion of some/all
of the following:

  EGL/egl.h
  GL/gl.h
  GL/glcorearb.h
  GLES/gl.h
  GLES2/gl2.h
  GLES3/gl3.h
  GLES3/gl31.h

This will allow us to build/use it alongside any combination of APIs.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>

mesa_glinterop: use consistent naming scheme for GL interop

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>

Revert "mesa: Build EGL without X11 headers after interop patchset"

This reverts commit 4e2c9a04354b6b133845b8b93c0c5d34261a91d0.

The solution was incomplete and fragile. An alternative one is coming
shortly.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>

docs: Note that GL_OES_geometry_shader and GL_OES_tessellation_shader are started

The GL_OES_geometry_shader work is on the oes_shader_io_blocks branch
of idr's fd.o repository.

The GL_OES_tessellation_shader work is on the tess-gles branch
of kwg's fd.o repository.

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

c11/threads: resolve link issues with -O0

Add weak symbol notation for the pthread_mutexattr* symbols, thus making
the linker happy. When building with -O1 or greater the optimiser will
kick in and remove the said functions as they are dead/unreachable code.

Ideally we'll enable the optimisations locally, yet that does not seem
to work atm.

v2: Add the AX_GCC_FUNC_ATTRIBUTE([weak]) hunk in configure.

Cc: Alejandro Piñeiro <apinheiro@igalia.com>
Cc: Ben Widawsky <ben@bwidawsk.net>
Cc: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: Rob Herring <robh@kernel.org>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Rob Clark <robdclark@gmail.com>
Tested-by: Mark Janes <mark.a.janes@intel.com>

swr: [rasterizer] remove containers.hpp

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

swr: [rasterizer core] remove utility dead code

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

swr: [rasterizer core] buckets fixes

1. Don't clear bucket descriptions to fix issues with sim level
   buckets getting out of sync.

2. Close out threadviz file descriptors in ClearThreads().

3. Skip buckets for jitter based buckets when multithreaded. We need
   thread local storage through llvm jit functions to be fixed before
   we can enable this.

4. Fix buckets StopCapture to correctly detect capture complete.

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

swr: [rasterizer core] move centroid setup out of CalcCentroidBarycentrics

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

swr: [rasterizer jitter] implement InstanceID/VertexID in fetch jit

Reviewed-by: Bruce Cherniak <bruce.cherniak@intel.com>

mesa: Silence unused parameter warnings

Neither shProg nor name was used.  Remove them both.

main/shader_query.cpp:779:53: warning: unused parameter ‘shProg’ [-Wunused-parameter]
 program_resource_location(struct gl_shader_program *shProg,
                                                     ^
main/shader_query.cpp:780:72: warning: unused parameter ‘name’ [-Wunused-parameter]
                           struct gl_program_resource *res, const char *name,
                                                                        ^

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>

glsl/linker: Silence unused parameter warning

The parameter is required for the interface.

glsl/link_uniforms.cpp:689:61: warning: unused parameter ‘record_type’ [-Wunused-parameter]
                             bool row_major, const glsl_type *record_type,
                                                             ^

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>

dri: Add YVU formats

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

i965: Allow creating planar YUV __DRIimages

Lift the resctriction we had before and allow creation of images with
multiple planes. We still require all the planes to be within the same
bo.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Chad Versace <chad.versace@intel.com>

i965: Invoke lowering pass for YUV textures

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

i965: Support textures with multiple planes

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

i965: Create multiple miptrees for planar YUV images

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

i965: Refactor intel_set_texture_image_bo() to create_mt_for_dri_image()

This function now only creates the mt and we then call
intel_set_texture_image_mt() in intel_image_target_texture_2d() to set
it for the texture image.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

i965: Use intel_set_texture_image_mt() in intelSetTexBuffer2()

Create the mt for the drawable bo directly and call our new
intel_miptree_create_for_bo() helper instead.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Chad Versace <chad.versace@intel.com>

i965: Add new intel_set_texture_image_mt() helper

This factors out the work of setting up a miptree as the backing for a
texture image into a new helper.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Chad Versace <chad.versace@intel.com>

nir: Add a lowering pass for YUV textures

This lowers sampling from YUV textures to 1) one or more texture
instructions to sample each plane and 2) color space conversion to RGB.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

nir: Handle NULL in nir_copy_deref()

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

nir: Add new 'plane' texture source type

This will be used to select the plane to sample from for planar
textures.

Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

mesa: log buffer ID numbers in decimal, not hexadecimal

All the other error messages use decimal.  Let's be consistent.

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

mesa: use enum name in bind_buffer_object() error message

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

mesa: raise error for glEnable(GL_VERTEX_ARRAY), etc. in core profile

Otherwise, if the call executes normally we'll hit an assertion later
in the VBO code when we draw something.  Note that these cases were
already handled correctly for the glIsEnabled() function (and the API
checks were copied from there).

Tested with new piglit gl-3.1-enable-vertex-array test.

v2: fix compat/es mix-up, per Ilia.

Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Charmaine Lee <charmainel@vmware.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>

docs/egl: Android platform can also be build using autotools

We added support for Android build using autotools (configure),
update the documentation to reflect that.

Signed-off-by: Nicolas Boichat <drinkcat@google.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

i965: fix double-precision vertex inputs measurement

For double-precision vertex inputs we need to measure them in dvec4
terms, and for single-precision vertex inputs we need to measure them in
vec4 terms.

For the later case, we use type_size_vec4() function. For the former
case, we had a wrong implementation based on type_size_vec4().

This commit introduces a proper type_size_dvec4() function, that we use
to measure vertex inputs.

Measuring double-precision vertex inputs as dvec4 is required because
ARB_vertex_attrib_64bit states that these uses the same number of
locations than the single-precision version. That is, two consecutives
dvec4 would be located in location "x" and location "x+1", not "x+2".

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

docs: true up nvc0 status - images, etc

Images aren't supported on maxwell, but neither is tessellation. Don't
overly confuse matters by trying to expose those subtleties in the
GL3.txt file/relnotes.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Acked-by: Dave Airlie <airlied@redhat.com>

st/mesa: enable ARB_ES3_1_compatibility when ES 3.1 would be exposed

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Dave Airlie <airlied@redhat.com>

mesa: remove separate enable for KHR_robust_buffer_access_behavior

This extension appears to be a strict subset of the ARB version. Also
remove it from GL3.txt since it doesn't seem relevant.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>

glsl: add support for explicit components to frag outputs

V2: fix error checking for arrays and components. V1 was
only taking into account all the array elements and all the
components of one of the varyings during the comparision
and treating the other as a single slot/component.

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

mesa: add view classes for 3d astc formats

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

glsl: add EXT_clip_cull_distance support based on ARB_cull_distance

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Dave Airlie <airlied@redhat.com>

nvc0: expose robust buffer access

We apparently pass all the relevant CTS tests. There are probably some
shortcomings, but they can be addressed down the line.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>

i965: Use ISL for surface format introspection

With this, we can delete the surface format table in brw_surface_formats.c
because all of the information we need is now in ISL.

anv/formats: Use isl_format_supports* for format introspection

isl: Add per-gen format introspection

This is just a copy-and-paste from brw_surface_formats.c.  For the
supports_vertex_fetch function, we do a bit more work so that it properly
handles Bay Trail.

isl: Add the ISL_FORMAT_R32G32_FLOAT_LD format

isl: Add support for quering the string name of a format

i965: Enable ARB/KHR_robust_buffer_access_behavior on BYT and HSW+

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

main: Add extension enable bits for KHR_robust_buffer_access_behavior

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nir/lower_samplers: Protect against sampler index overflow

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

glsl: Add an option to clamp block indices when lowering UBO/SSBOs

This prevents array overflow when the block is actually an array of UBOs or
SSBOs.  On some hardware such as i965, such overflows can cause GPU hangs.

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

glsl/linker: Add a helper variable for compiler options

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

i965/draw: Use the real size for index buffers

Previously, we were using the size of the whole BO which may be
substantially larger than the actual index buffer size.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/draw: Use the real size for vertex buffers

Previously, we were using the size of the BO which may be substantially
larger than the actual vertex buffer size.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/draw: Use 3-channel formats for vertex fetch when possible.

For a long time, several of the 3-channel vertex formats didn't exist so we
faked them with 4-channel versions.  Starting with Sandy Bridge, we can use
R16G16B16_FLOAT and 8 and 16-bit integer formats become available on
Haswell and Bay Trail.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/surface_formats: Update the VB column for new formats added on BYT

Bay Trail and Haswell added a bunch of new vertex formats.  There was also
the addition of 64-bit passthrough formats for BDW+.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/draw: Properly handle rounding when dividing by InstanceDivisor

The old code always divided rounded down and then subtracted 1.  What we
wanted was to divide rounded up and then subtract 1 which is equivalent to
subtracting 1 and then dividing rounded down.

Cc: "11.1 11.2" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/draw: Account for BaseInstance in VBO bounds

Cc: "11.1 11.2" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/draw: Use worst-case VBO bounds if brw->num_instances == 0

Previously, we only handled the "I don't know what's going on" case for
things with InstanceDivisor == 0.  However, in the DrawIndirect case we can
get num_instances == 0 and we don't know what's going on with the instanced
ones either.  This commit makes the worst-case bound the default and then
conservatively tightens the bound.

Cc: "11.1 11.2" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/draw: Delay when we get the bo for vertex buffers

The previous code got the BO the first time we encountered it.  However,
this can potentially lead to problems if the BO is used for multiple arrays
with the same buffer object because the range we declare as busy may not be
quite right.  By delaying the call to intel_bufferobj_buffer, we can ensure
that we have the full range for the given buffer.

Cc: "11.1 11.2" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/draw: Stop relying on min_index == -1 for invalid index bounds

The vbo layer passes an index_bounds_valid flag that we should be using
instead.  This also fixes a bug when min_index == -1 and basevertex != 0
where we were actually comparing min_index + basevertex == -1 which was
false and we were getting the wrong buffer-sizing path.

Cc: "11.1 11.2" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

vbo: Declare the index range invalid for DrawTransformFeedback

Right now, we're setting the range to [0, 0] which is obviously bogus.
Instead, we should set it to be invalid like we do for DrawIndirect.

Cc: "11.1 11.2" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

vbo: Declare the index range invalid for DrawIndirect

Right now, we're just setting the range to [0, MAX_UINT32] which, while
correct isn't helpful.  With DrawIndirect, you can't really know what the
actual range is so we may as well flag it as being an invalid range.  This
is what we do for draws with index buffer which is similar (the indices
aren't statically known) if a bit simpler.

Cc: "11.1 11.2" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

mesa/teximage: fix GL_FLOAT in comment

Noticed by Brian. Trivial.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>

glsl: fix explicit location validation for doubles

Previously we would fail to find a match for the second half of a
dvec4 as 'i' would get incremented to 1 before we added the var to
the array at component 0.

Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

docs: update ARB_cull_distance status.

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

st/mesa: reenable culling

Now the lowering pass is fixed, reenable ARB_cull_distance.

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