isl/state: Only set cube face enables if usage includes CUBE_BIT

It seems safe to set it all the time, but this reduces the diff between
the way i965 does it and what ISL does.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

isl/state: Use the layout for computing qpitch rather than dimensions

For depth/stencil 1-D textures on SKL, we want them layed out in the old
format that has been used since gen4.  In order for the surface state
fill-out code to handle, this it needs to distinguish based on layout
rather than just dimensionality.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

isl/state: Set the IntegerSurfaceFormat bit on Haswell

This fixes 688 Vulkan CTS tests on Haswell.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

isl/format: Mark R9G9B9E5 as containing 9-bit unsigned float channels

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

isl/state: Don't set RenderTargetViewExtent for texture surfaces

The docs specify that this only matters for render targets and surfaces
used with typed dataport messages.  On some platforms (gen4-6) the Depth
field has more bits than RenderTargetViewExtent so we can have textures
with more levels than we can render to.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

isl/state: Set SurfaceArray based on the surface dimension

According to the PRM, you can't set SurfaceArray for 3D or buffer textures.
There doesn't seem to be a good reason not to set it when we can.  On the
other hand, if we don't set it we can end up getting strange results for
1-layer array textures such as textureSize() returning the wrong results.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

isl/state: Don't force-disable L2 bypass for everything

We already set the bit in the few cases where it's required by the docs so
there's no need to set it all the time.  This has no noticable perf impact
for Dota 2 on Vulkan with the time demo I have.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

isl/state: Refactor the setup of clear colors

This commit switches clear colors to use #if's instead of a C if.  This
lets us properly handle SNB where the clear color field doesn't exist.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

isl/state: Refactor the per-gen isl_to_gen_h/valign tables

This moves the #if's around so that halign and valign have different sets
of #if conditions.  This also prepares us for SNB because isl_to_gen_halign
is not defined at all on gen6.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

isl/state: Return an extent3d from the halign/valign helper

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

isl/state: Put pitch calculations together

This is purely cosmetic, but it makes things look a bit more readable.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

isl/state: Put all dimension setup together and towards the top

This is purely cosmetic, but it makes things look a bit more readable.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

isl/state: Put surface format setup at the top

This is purely cosmetic, but it makes things look a bit more readable.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

isl/state: Remove some unused fields

They're already zero-initialized and we have no plans of doing anything
more interesting with them.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

isl/state: Don't use designated initializers for the surface state

While designated initializers are nice, they also force us to put some
things in the initializer and some things later.  Surface state setup is
complicated enough that this really hurts readability in the long run.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

genxml/gen8,9: Prefix the multisample format enum with MSFMT

This is what gen7 does and it's nice to have a prefix

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

i965/blorp: Only set src_z for gen8+ 3D textures

Otherwise, we end up with a bogus value in the third component.  On gen6-7
where we always use 2D textures, this can cause problems if the
SurfaceArray bit is set in the SURFACE_STATE.

Acked-by: Chad Versace <chad.versace@intel.com>

i965/gen7,8: Set SURFACE_IS_ARRAY for all non-3D texture types

There's no real reason why we shouldn't set this bit.  It does affect how
the sampler operates a bit but since you can have a 2D non-array view of a
2D_ARRAY texture that distinction is very weak.  Also, this is what ISL
will do and we would like this change to be isolated from using ISL.

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

i965/gen4: Subtract 1 from buffer sizes

The PRM states that the values put in Width, Height, and Depth should be
various bits from the value size - 1.  We seem to have done this wrong
more-or-less from the start.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>

i965: Remove fake W-tiled render target support

This hasn't been used since 1cfb4bc890b8 where we deleted the meta stencil
blit path.

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

i965/fs: Use a default Y coordinate of 0 for TXF on gen9+

Previously, we were incrementing length but not actually putting anything
in the Y coordinate.  This meant that 1-D TXF operations had a garbage
array index.  If the surface is emitted as 1-D non-array, the coordinate
gets discarded and it works fine.  If it happens to be bound as an array
surface, it may count as an out-of-bounds array access and you get zero.

Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>

i965/gen8: Use the qpitch from the aux_mt for AUX_QPITCH

Reviewed-by: Chad Versace <chad.versace@intel.com>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>

i965/blorp/gen8: Use the correct max level and layer in emit_surface_states

We were adding in the base which is wrong because the values given in the
miptree are relative to zero and not the base layer/level.

Reviewed-by: Chad Versace <chad.versace@intel.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>

i965: Drop the maximum 3D texture size to 512 on Sandy Bridge

The RenderTargetViewExtent field of RENDER_SURFACE_STATE is supposed to be
set to the depth of a 3-D texture when rendering.  Unfortunatley, that
field is only 9 bits on Sandy Bridge and prior so we can't actually bind
a 3-D texturing for rendering if it has depth > 512.  On Ivy Bridge, this
field was bumpped to 11 bits so we can go all the way up to 2048.  On Iron
Lake and prior, we don't support layered rendering and we use OffsetX/Y
hacks to render to particular layers so 2048 is ok there too.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>

i965/gen4-6: Handle gl_texture_object::BaseLevel and MinLayer correctly

This is basically a direct translation of what we do for gen7.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83036
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>

i965/gen4: Pull texture formats from the texture object not the miptree

This makes texture views sort-of work.  It doesn't add full texture view
support for gen4-5 but it is enough to fix the GL_ARB_copy_image formats
piglit test on Iron Lake.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=83036
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>

i965: Fix point size with tessellation/geometry shaders in GLES.

Our previous code worked for desktop GL, and ES without geometry or
tessellation shaders.  But those features require fancier point size
handling.  Fortunately, we can use one rule for all APIs.

Fixes a number of dEQP tests with EXT_tessellation_shader enabled:
dEQP-GLES31.functional.tessellation_geometry_interaction.point_size.*

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

.mailmap: fix my main address

i965: move vs outputs written into a helper

We will reuse this for fs key generation for the on disk shader
cache.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>

st/mesa: use a single memcpy in st_ReadPixels when possible

This avoids costly address recomputations, function overhead, and may trigger
large copy optimizations.

Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

glsl: only match gl_FragData and not gl_SecondaryFragDataEXT

There's special logic around finding gl_FragData. It latches onto any
array with FRAG_RESULT_DATA0. However gl_SecondaryFragDataEXT[], added
by GL_EXT_blend_func_extended, fits those parameters as well. The real
frag data array should have index 0 though, so we can use that to
distinguish them.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96617
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nv50,nvc0: fix start_instance in manual push path

The start instance is applied as an offset into the buffer directly,
ignoring the divisor, not as an instance id offset that respects the
divisor.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.2 12.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

translate: fix start_instance parameter in sse version

The generic version gets this right already, but this was using an
incorrect formula in SSE.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.2 12.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

anv/cmd: Dirty descriptor sets when a new pipeline is bound

Ever since c2581a9375ea, the binding table layout has depended on the
pipeline.  This means that whenever we change pipelines we also need to
re-emit binding tables for the new layout.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

anv/cmd: Move emit_descriptor_pointers to genX_cmd_buffer.c

It's tiny and fully generic so there's really no reason for it to be in a
gen7-specific file.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

anv/cmd: Move flush_descriptor_sets to anv_cmd_buffer.c

There's no good reason for recompiling it

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

spirv: Use the system value version of gl_FrontFace

SPIR-V treats it as an input but NIR wants the system value.  This
shouldn't have been too much of a surprise given that we have to do the
same conversion in the GLSL IR to NIR pass.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

i965: Reorganize prog_data->total_scratch code a bit.

Cc: "12.0" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

radeonsi: add a debug flag for unsafe math LLVM optimizations

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: use u_blitter for mipmap generation

This reduces time spend in glGenerateMipmap by a half.

v2: don't decompress the levels to be overwritten

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

gallium/u_blitter: implement mipmap generation

for pipe_context::generate_mipmap

first move some of the blit code from util_blitter_blit_generic
to a separate function, then use it from util_blitter_generate_mipmap

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

st/mesa: cache staging texture for glReadPixels

v2: add ST_DEBUG flag for disabling (suggested by Ilia)

Reviewed-by: Marek Olšák <marek.olsak@amd.com> (v1)

st/mesa: invalidate readpixels cache

Whenever a draw happens or some other function call might change the result
of future glReadPixels calls, we must invalidate the cache.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

st/mesa: add readpix_cache structure

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

st/mesa: move ReadPixels blit into a separate function

Reviewed-by: Marek Olšák <marek.olsak@amd.com>

st/mesa: flush bitmap cache before CopyImageSubData

Found by inspection.

Cc: 11.2 12.0 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

st/mesa: flush bitmap cache before texture functions

As far as I can tell, a sequence of glBitmap followed by texture functions
that refer to a texture bound as the framebuffer is well within what should
be allowed.

Found by inspection.

Cc: 11.2 12.0 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

st/mesa: flush bitmap cache before compute dispatch

In the unlikely case that a program uses glBitmap to render to a framebuffer
whose texture is bound in a compute shader.

Found by inspection.

Cc: 11.2 12.0 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

i965: get PrimitiveMode from the program rather than the shader struct

This is more consistent with what we do elsewhere and will allow
us to only cache one of the values in the shader cache.

Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>

clover: Fix build against clang SVN >= r273191

setLangDefaults() now requires PreprocessorOptions as an argument.

Reviewed-and-Tested-by: Michel Dänzer <michel.daenzer@amd.com>

i965: Fix multiplication of immediates on Cherryview/Broxton.

Cherryview and Broxton don't support DW x DW multiplication.  We have
piles of code to handle this, but apparently weren't retyping in the
immediate case.

For example,
tests/spec/arb_tessellation_shader/execution/dvec3-vs-tcs-tes
makes the simulator angry about instructions such as:

   mul(8) r18<1>:D r10.0<8;8,1>:D 0x00000003:D

Just retype to W or UW.  It should be safe on all platforms.

Cc: "12.0" <mesa-stable@lists.freedesktop.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95462
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>

anv: Add proper support for depth clamping

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

anv/cmd_buffer: Split emit_viewport in two

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

anv/cmd_buffer: Set depth/stencil extent based on the image

It used to be based on the framebuffer which isn't quite right.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

anv/cmd_buffer: Don't crash if push constants are provided for missing stages

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

anv/pipeline: Do invariance propagation on SPIR-V shaders

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

nir/alu_to_scalar: Respect the exact ALU operation qualifier

Just setting builder->exact isn't sufficient because that only applies to
instructions that are built with the builder but instructions created
manually and only inserted using the builder are left alone.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

nir: Add a pass for propagating invariant decorations

This pass is similar to propagate_invariance in the GLSL compiler.  The
real "output" of this pass is that any algebraic operations which are
eventually consumed by an invariant variable get marked as "exact".

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>

nir/algebraic: Remove imprecise flog2 optimizations

While mathematically correct, these two optimizations result in an
expression with substantially lower precision than the original.  For any
positive finite floating-point value, log2(x) is well-defined and finite.
More precisely, it is in the range [-150, 150] so any sum of logarithms
log2(a) + log2(b) is also well-defined and finite as long as a and b are
both positive and finite.  However, if a and b are either very small or
very large, their product may get flushed to infinity or zero causing
log2(a * b) to be nowhere close to log2(a) + log2(b).

This imprecision was causing incorrect rendering in Talos Principal because
part of its HDR rendering process involves doing 8 texture operations,
clamping the result to [0, 65000], taking a dot-product with a constant,
and then taking the log2.  This is done 6 or 8 times and summed to produce
the final result which is written to a red texture.  In cases where you
have a region of the screen that is very dark, it can end up getting a
result value of -inf which is not what is intended.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96425
Cc: "11.1 11.2 12.0" <mesa-stable@lists.freedesktop.org>

i965: Delete redundant extension enables

A nearly identical block already exists in the gen >= 6 block above.

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

mesa: Fix incorrect "see also" comments

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

mesa: Silence unused parameter warning

main/pipelineobj.c: In function ‘delete_pipelineobj_cb’:
main/pipelineobj.c:110:30: warning: unused parameter ‘id’ [-Wunused-parameter]
 delete_pipelineobj_cb(GLuint id, void *data, void *userData)
                              ^

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

gallium: make image_view const

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

gallium: make constant_buffer const

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

gallium: make shader_buffers const

Be consistent with the rest of the "set_xyz" state interfaces.

Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: use trapezoid distribution for tess on Fiji and Polaris

This yields a small performance improvement in Unigine Heaven.

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

radeonsi/sid: add Fiji+ tesselation distribution mode

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

radeonsi: emit PA_SC_RASTER_CONFIG_1 only once

It is the same for all SEs.

Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

radeonsi: fix calculation of valid RB mask per SE

The old calculation treated too many RBs as disabled.

Cc: 11.0 11.1 11.2 12.0 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

radeonsi: raise SI_PM4_MAX_DW

The old limit, introduced in commit afa752d3f03ac6697581ff5d324e8ac0512ef513,
was exceeded by 4 SE configurations which hit si_write_harvested_raster_configs.

Cc: 11.1 11.2 12.0 <mesa-stable@lists.freedesktop.org>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

gallivm: don't use integer min/max sse intrinsics with llvm >= 3.9

Apparently, these are deprecated. There's some AutoUpgrade feature which
is supposed to promote these to cmp/select, which apparently doesn't work
with jit code. It is possible it's not actually even meant to work (see
the bug filed against llvm which couldn't provide an answer neither)
but in any case this is meant to be only temporary unless the intrinsics
are really illegal. So, just use the fallback code (which should be cmp/select,
we're actually doing cmp/sext/trunc/select, but in any case llvm 3.9 manages
to optimize this back to pmin/pmax in the end).

This addresses https://llvm.org/bugs/show_bug.cgi?id=28176

CC: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Tested-by: Vinson Lee <vlee@freedesktop.org>
Tested-by: Aaron Watry <awatry@gmail.com>

nvc0: don't make use of push hint if there are no non-const user vbos

This makes the check match up what we do on nv50 as well - there's no
point in switching over the push path if everything's in managed
buffers. This can happen when a shader uses a vertex without an enabled
array - we end up passing it a constant attribute.

This also has the effect of "fixing" some flickering in Talos. I have no
idea why. I've stared at the push logic forwards, backwards, and
sideways. By always forcing the push path (which is slow), the
flickering also goes away, but other rendering is still wrong
(specifically draw 383068 as identified in the bug). However by not
switching over to the push path, draw 383068 is correct.

Note that other flickering remains in Talos, like the red/green
walls/floors. This takes care of the shadow flickering though.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90513
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

gk104/ir: fix tex use generation to be more careful about eliding uses

If we have a loop, instructions before the tex might be added as tex
uses, and those may in fact dominate all other uses of the tex results.
This however doesn't mean that we don't need a texbar after the tex.
Only check if uses dominate each other they are dominated by the tex.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96565
Fixes: 7752bbc44 (gk104/ir: simplify and fool-proof texbar algorithm)
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: "11.2 12.0" <mesa-stable@lists.freedesktop.org>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

nv50: add support for GL_EXT_window_rectangles

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

nvc0: add support for GL_EXT_window_rectangles

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

st/mesa: add support for GL_EXT_window_rectangles

Make sure to pass the requisite information in draws, blits, and clears
that work on the context's draw buffer.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

gallium: add PIPE_CAP_MAX_WINDOW_RECTANGLES to all drivers

This says how many window rectangles are supported by the
implementation, although it may not exceed PIPE_MAX_WINDOW_RECTANGLES.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Brian Paul <brianp@vmware.com>

gallium: add API for setting window rectangles

Window rectangles apply to all framebuffer operations, either in
inclusive or exclusive mode. They may also be specified as part of a
blit operation.

In exclusive mode, any fragment inside any of the specified rectangles
will be discarded.

In inclusive mode, any fragment outside every rectangle will be
discarded.

The no-op state is to have 0 rectangles in exclusive mode.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

mesa: add GL_EXT_window_rectangles state storage/retrieval functionality

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Brian Paul <brianp@vmware.com>

glapi: add GL_EXT_window_rectangles entrypoints

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Brian Paul <brianp@vmware.com>

nv50/ir: add missing strings for some recent sysvals

This is pretty useful for debugging purposes and those should
not be omitted.

Fixes: 517a93b3 ("nvc0: add ARB_shader_draw_parameters support")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>

swr: Update screen->context pointer with multiple contexts.

A pipe pointer in the screen allows for access to current device context
 in flush_frontbuffer and resource_destroy.  This wasn't tracking current
context in multi-context situations.

v2: More caffeine.  Corrected compare, removed unnecessary set of
screen-pipe in create_context, and added a few comments.

scons: put the generated git_sha1.h file in top-level src/ directory

To match what's done in the automake build.

v2: Use git rev-parse to get a 10-character hash ID
    Fix Python imports

Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: José Fonseca <jfonseca@vmware.com>

swr: switch from overriding -march to selecting features

Acked-by: Chuck Atkins <chuck.atkins@kitware.com>
Tested-by: Chuck Atkins <chuck.atkins@kitware.com>

mesa: remove remaining tabs in api_validate.c

Reviewed-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>

i965/fs: indirect addressing with doubles is not supported in CHV/BSW/BXT

From the Cherryview's PRM, Volume 7, 3D Media GPGPU Engine, Register Region
Restrictions, page 844:

  "When source or destination datatype is 64b or operation is integer DWord
   multiply, indirect addressing must not be used."

v2:
- Fix it for Broxton too.

v3:
- Simplify code by using subscript() and not creating a new num_components
variable (Kenneth).

Signed-off-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95462
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/fs: Fix single-precision to double-precision conversions for CHV/BSW/BXT

From the Cherryview PRM, Volume 7, 3D Media GPGPU Engine,
Register Region Restrictions:

   "When source or destination is 64b (...), regioning in Align1
    must follow these rules:

    1. Source and destination horizontal stride must be aligned to
       the same qword.
    (...)"

v2:
- Fix it for Broxton too.

v3:
- Remove inst->regs_written change as it is not necessary (Ken)

Cc: "12.0" <mesa-stable@lists.freedesktop.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=95462
Tested-by: Mark Janes <mark.a.janes@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

docs: Mention GL_ARB_ES3_1_compatibility in release notes.

Ilia reminded me that I forgot this.

i965: Fix comment about CS scratch space encodings on Broadwell+.

I typo'd this.

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

docs: Update ARB_ES3_1_compatibility status for i965.

i965: Drop perf_debug about rasterizer discard in SOL vs. clipper.

I recently experimented with performing rasterizer discard in the SOL
unit instead of the clipper, and as far as I can tell, it's basically
the same performance.  The clipper comes directly after SOL anyway,
and setting the clipper to REJECT_ALL should be pretty darn cheap.

Keep the perf_debug on Sandybridge, where the GS actually does work.

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

i965: Enable GL_ARB_ES3_1_compatibility on Gen8+ if CS are available.

There are almost no tests in any test suite, but what little I've found
seems to work.  Ilia believes everything is in place.

v2: Predicate the enable on ES 3.1 being available (Gen8+) and also
    ARB_compute_shader being available (requested by Ilia).

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

mesa: If validation fails in a debug context just emit a debug message

There are quite a few pipelines that desktop applications (including a
bunch of piglit test) can expect to have run but don't meet the GLES
requirements.  Instead of failing validation, just emit a debug message.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96358
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
Cc: Gregory Hainaut <gregory.hainaut@gmail.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>

glsl: Always strip arrayness in precision_qualifier_allowed

Previously some callers of precision_qualifier_allowed would strip the
arrayness from the type and some would not.  As a result, some places
would not notice that float[6], for example, needed a precision
qualifier.

Fixes the new piglit test no-default-float-array-precision.frag.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=96358
Cc: "12.0" <mesa-stable@lists.freedesktop.org>
Cc: Gregory Hainaut <gregory.hainaut@gmail.com>
Cc: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>

mesa/main: Update _mesa_new_shader.

Left over from 31dee99e052902bc08ddbb1009748dc982ac3211.  It should fix
Clang Windows build.

Trivial.

st/vdpau: we support lumakeying now

Signed-off-by: Christian König <christian.koenig@amd.com>

vl: support luma keying for interlaced surfaces as well

We had the CSC code twice in there, factor it out into a separate function.

Signed-off-by: Christian König <christian.koenig@amd.com>

i965: remove remaining tabs in brw_link.cpp

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

vbo: Use a bitmask to track the active arrays in vbo_save*.

The use of a bitmask makes functions iterating only active
attributes less visible in profiles.

v2: Use _mesa_bit_scan{,64} instead of open coding.
v3: Use u_bit_scan{,64} instead of _mesa_bit_scan{,64}.

Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Mathias Fröhlich <Mathias.Froehlich@web.de>

vbo: Use a bitmask to track the active arrays in vbo_exec*.

The use of a bitmask makes functions iterating only active
attributes less visible in profiles.

v2: Use _mesa_bit_scan{,64} instead of open coding.
v3: Use u_bit_scan{,64} instead of _mesa_bit_scan{,64}.

Reviewed-by: Brian Paul <brianp@vmware.com>
Signed-off-by: Mathias Fröhlich <Mathias.Froehlich@web.de>