gallivm: (trivial) Fix typo in comment introduced by 70dc8a

Fix typo in comment introduced by 70dc8a

Signed-off-by: Alexandre Demers <alexandre.f.demers@gmail.com>
Signed-off-by: Jose Fonseca <jfonseca@vmware.com>

mesa: improve ARB_copy_image internal format compat check

The memory layout of compatible internal formats may differ in bytes per
block, so TexFormat is not a reliable measure of compatibility. For example,
GL_RGB8 and GL_RGB8UI are compatible formats, but GL_RGB8 may be laid out in
memory as B8G8R8X8. If GL_RGB8UI has a 3 byte-per-block memory layout, the
existing compatibility check will fail.

Additionally, the current check allows any two compressed textures which share
block size to be used, whereas the spec gives an explicit table of compatible
formats.

v2: Use a switch instead of array iteration for block class and show the
    correct GL error when internal formats are mismatched.
v3: Include spec citations for new compatibility checks, rearrange check
    order to ensure that compressed, view-compatible formats return the
    correct result, and make style fixes. Original commit message amended
    for clarity.
v4: Reformatted spec citations.

Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

nir: Fix non-determinism in nir_lower_vars_to_ssa().

Previously, we stored derefs in a hash table, using the malloc'd pointer
as the key.  Then, we walked through the hash table and generated code,
based on the order of the hash table's elements.

Memory addresses returned by malloc are pretty much random, which meant
that the hash was random, and the hash table's elements would be walked
in some random order.  This led to successive compiles of the same
shader using different variable names and slightly different orderings
of phi-nodes.  Code could not be diff'd, and the final assembly would
sometimes change slightly too.

It turns out the only point of the hash table was to avoid inserting
the same node multiple times for different dereferences.  We never
actually searched the hash table!  This patch uses an intrusive
linked list instead.  Since exec_list uses head and tail sentinels,
checking prev or next against NULL will tell us whether the node is
already in the list.

Pair programming with Jason Ekstrand.

Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>

util: Fix foreach_list_typed_safe when exec_node is not at offset 0.

__next and __prev are pointers to the structure containing the exec_node
link, not the embedded exec_node.  NULL checks would fail unless the
embedded exec_node happened to be at offset 0 in the parent struct.

v2: Jason Ekstrand <jason.ekstrand@intel.com>:
   Use "(__node)->__field.next != NULL" to check for the end of the list
   instead of the "&__next->__field != NULL".  The former is far more
   obviously correct as it matches what the non-safe versions do.  The
   original code tried to avoid any use of __next as the client code may
   delete it during its execution.  However, since the looping condition is
   checked after the iteration clause but before the client code is
   executed, we know that __node is valid during the looping condition.

Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965: Use NIR for scalar VS when INTEL_USE_NIR is set.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

i965/fs: Add VS output support to nir_setup_outputs().

Adapted from fs_visitor::visit(ir_variable *).

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

i965/fs: Handle VS inputs in the NIR backend.

(Jason noted that this is not a good long term solution, and we should
instead improve nir_lower_io so that this extra set of MOVs is
unnecessary.  I tend to agree, but decided we could do that as a
follow-up improvement.)

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

i965/fs: Refactor fs_visitor::nir_setup_inputs().

No functional change.  In preparation for supporting vertex shaders,
this adds a switch statement on shader stage (since vertex attributes
and fragment shader varyings will need different handling).  It also
renames "varying" to "input", to be more general.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

i965: Implement NIR intrinsics for loading VS system values.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

nir: Add intrinsics for SYSTEM_VALUE_BASE_VERTEX and VERTEX_ID_ZERO_BASE

Ian and I added these around the time Connor was developing NIR.  Now
that both exist, we should make them work together!

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

i965/nir: Lower to registers a bit later.

We can't safely call nir_optimize() with register present, since several
passes called in the loop can't handle registers, and will fail asserts.

Notably, nir_lower_vec_alus() and nir_opt_algebraic() really don't want
registers.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

i965/nir: Optimize after nir_lower_var_copies().

Array variable copy splitting generates a bunch of stuff we want to
clean up before proceeding.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

i965/fs: Store a pointer to brw_sampler_prog_key_data in the visitor.

The NIR backend hardcodes brw_wm_prog_key at the moment, which won't
work when we support scalar VS.  We could use get_tex(), but it's a
static method.  I was going to promote it to fs_visitor, but then
realized that both parameters (stage and key) are already members.

It then occured to me that we could just set up a pointer in the
constructor, and skip having a function altogether.

This patch also converts all existing users to use key_tex.

v2: Make key_tex a "const brw_sampler_prog_key_data *" instead of
    non-const; word-wrap some lines.  (Review comments from Topi.)

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

tnl: HAVE_LE32_VERTS is never defined, remove associated code

Reviewed-by: Matt Turner <mattst88@gmail.com>

mesa: move LONGSTRING into generated enums.c

enums.c is the only place this directive is needed.

Reviewed-by: Matt Turner <mattst88@gmail.com>

mesa: remove _ASMAPI, ASMAPIP

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

mesa: remove _XFORMAPI

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

swrast: remove _BLENDAPI

_BLENDAPI boils down to __cdecl on Windows, but __cdecl is the default
calling convention so this serves no purpose.

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

mesa: use ARRAY_SIZE in _mesa_QueryMatrixxOES()

Reviewed-by: Matt Turner <mattst88@gmail.com>

mesa: remove register keyword, add const in _mesa_QueryMatrixxOES()

Reviewed-by: Matt Turner <mattst88@gmail.com>

mesa: reindent querymatrix.c

Use 3-space indents, not 4.  Move some comments after the case statements.

Acked-by: Matt Turner <mattst88@gmail.com>

mesa: move fpclassify work-arounds into c99_math.h

v2: Use #error in the #else clause, per Jose.

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>

gallivm: Prevent double delete on LLVM 3.6

std::unique_ptr takes ownership of MM, and a double delete could ensure
in case of an error,  as pointed out by Chris Vine in
https://bugs.freedesktop.org/show_bug.cgi?id=89387

Reviewed-by: Chris Vine <chris@cvine.freeserve.co.uk>

autogen.sh: pass --force to autoreconf, quote ORIGDIR

By passing --force autoreconf will update all the aux files, which would
otherwise be ignored if one updates autoconf/automake.

Quote the ORIGDIR variable to prevent fall-outs, when its name contains
space.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

glx: remove support for non-multithreaded platforms

Implicitly required for a while, although commit 9385c592c68 (mapi:
remove u_thread.h) was the one that put the final nail on the
coffin.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

glx: remove final reference to THREADS

Left over from commit 18db13f5865(mapi: THREADS was always defined,
remove it)

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

configure: require pthreads for POSIX builds

This has been an implicit rule for building mesa for a long time. Let's
make it official and just bail out at configure time. This way we can
cleaning up some of our glx code.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

egl/main: convert thread management to use c11 threads

Convert the code to use the C11 threads implementation, and nuke the
Windows non-pthreads code-path. The c11/threads_win32.h abstraction
should be better than the current code.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

egl/main: use c11/threads' mutex directly

Remove the inline wrappers/abstraction layer.

Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

nir/worklist: Don't change the start index when computing the tail index

Reviewed-by: Mark Janes <mark.a.janes@intel.com>

nir: Optimize a + neg(a)

Shader-db i965 instructions:
total instructions in shared programs: 1711180 -> 1711159 (-0.00%)
instructions in affected programs:     825 -> 804 (-2.55%)
helped:                                9
HURT:                                  0
GAINED:                                3
LOST:                                  3

Shader-db NIR instructions:
total instructions in shared programs: 606187 -> 606179 (-0.00%)
instructions in affected programs:     298 -> 290 (-2.68%)
helped:                                4
HURT:                                  0
GAINED:                                0
LOST:                                  0

Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Signed-off-by: Thomas Helland <thomashelland90@gmail.com>

nir: Optimize (a*b)+(a*c) -> a*(b+c)

Shader-db i965 instructions:
total instructions in shared programs: 1715894 -> 1710802 (-0.30%)
instructions in affected programs:     443080 -> 437988 (-1.15%)
helped:                                1502
HURT:                                  13
GAINED:                                4
LOST:                                  4

Shader-db NIR instructions:
total instructions in shared programs: 607710 -> 606187 (-0.25%)
instructions in affected programs:     208285 -> 206762 (-0.73%)
helped:                                769
HURT:                                  8
GAINED:                                0
LOST:                                  0

Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Signed-off-by: Thomas Helland <thomashelland90@gmail.com>

vbo: improve the code style by adjust the preprocessing c code directives

Brian Paul review suggestion: there's more macro use here than necessary.
Removed and redefine some #define preprocessing directives.
Removed the directive input parameter 'T' .
No functional changes.

Signed-off-by: Marius Predut <marius.predut@intel.com>
Reviewed-by: Brian Paul <brianp@vmware.com>

mesa: remove CPU_TO_LE32() for AIX

This is the only remnant of AIX-specific code in Mesa.  Probably long
unused.

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

mesa: remove #define __volatile

Not actually used anwhere in Mesa.

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

mesa: use strdup() instead of _mesa_strdup()

We were already using strdup() in various places in Mesa.  Get rid
of the _mesa_strdup() wrapper.  All the callers pass a non-NULL
argument so the NULL check isn't needed either.

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

st/glx: use strdup() instead of _mesa_strdup()

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

xlib: use strdup() instead of _mesa_strdup()

Reviewed-by: Jose Fonseca <jfonseca@vmware.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>

i915: add parens to silence operator precedence warning

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

i965: Fix out-of-bounds accesses into pull_constant_loc array

The piglit test glsl-fs-uniform-array-loop-unroll.shader_test was designed
to do an out of bounds access into an uniform array to make sure that we
handle that situation gracefully inside the driver, however, as Ken describes
in bug 79202, Valgrind reports that this is leading to an out-of-bounds access
in fs_visitor::demote_pull_constants().

Before accessing the pull_constant_loc array we should make sure that
the uniform we are trying to access is valid.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=79202
Reviewed-by: Matt Turner <mattst88@gmail.com>

i965/gen6 gs: Convert brw_imm_ud/brw_imm_d to src_reg

Same idea as this patch, only for gen6_gs_visitor:

commit 49a938a265f5959c9b558995cc658f80acb6eb18
Author: Jordan Justen <jordan.l.justen@intel.com>
Date:   Fri Feb 20 12:12:25 2015 -0800
    i965/fs: Use fs_reg for CS/VS atomics pixel mask immediate data

Suggested-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Chris Forbes <chrisf@ijw.co.nz>
Reviewed-by: Matt Turner <mattst88@gmail.com>

i965/fs: Use unsigned for CS/VS atomics pixel mask immediate data

brw_imm_ud(0xffff) should have been converted to fs_reg(0xffffu) to
make sure the uint32_t fs_reg constructor was matched.

commit 49a938a265f5959c9b558995cc658f80acb6eb18
Author: Jordan Justen <jordan.l.justen@intel.com>
Date:   Fri Feb 20 12:12:25 2015 -0800
    i965/fs: Use fs_reg for CS/VS atomics pixel mask immediate data

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

i965/gen8: Don't allocate hiz miptree structure

We now skip allocating a hiz miptree for gen8. Instead, we calculate
the required hiz buffer parameters and allocate a bo directly.

v2:
 * Update hz_height calculation as suggested by Topi
v3:
 * Bail if we failed to create the bo (Ben)
v4:
 * CEILING => DIV_ROUND_UP
 * Make sure mt->logical_depth0 being 0 would not cause trouble
 * Fail if Y tiling is not returned

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67564
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>

i965/gen7: Don't allocate hiz miptree structure

We now skip allocating a hiz miptree for gen7. Instead, we calculate
the required hiz buffer parameters and allocate a bo directly.

v2:
 * Update hz_height calculation as suggested by Topi
v3:
 * Bail if we failed to create the bo (Ben)
v4:
 * CEILING => DIV_ROUND_UP
 * Make sure mt->logical_depth0 being 0 would not cause trouble
 * Fail if Y tiling is not returned

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67564
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>

i965/gen8: Don't rely directly on the hiz miptree structure

We are still allocating a miptree for hiz, but we only use fields from
intel_miptree_aux_buffer. This will allow us to switch over to not
allocating a miptree.

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>

i965/gen7: Don't rely directly on the hiz miptree structure

We are still allocating a miptree for hiz, but we only use fields from
intel_miptree_aux_buffer. This will allow us to switch over to not
allocating a miptree.

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>

i965/hiz: Start to separate miptree out from hiz buffers

Today we allocate a miptree's for the hiz buffer. We needed this in
the past because we would point the hardware at offsets of the hiz
buffer. Since the hiz format is not documented, this is not a good
idea.

Since moving to support layered rendering on Gen7+, we no longer point
at an offset into the buffer on Gen7+.

Therefore, to support hiz on Gen7+, we don't need a full miptree
structure allocated.

This patch starts to create a new auxiliary buffer structure
(intel_miptree_aux_buffer) that can be a more simplistic miptree
side-band buffer associated with a miptree. (For example, to serve the
needs of the hiz buffer.)

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>

mesa/scissor: fix typos in debug names

Just noticed this when working on virgl.

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

nvc0: fix wrong max value for driver queries

The maximum value of a Gallium HUD's panel is automatically adjusted
when the current value is greater than the max. If we set the
pipe_query_driver_info::max_value to UINT64_MAX, the maximum value is
never adjusted and this results in a flat line instead of a pretty curve
which is correctly scaled.

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

i965: Silence GCC maybe-uninitialized warning.

brw_shader.cpp: In function ‘bool brw_saturate_immediate(brw_reg_type, brw_reg*)’:
brw_shader.cpp:618:31: warning: ‘sat_imm.brw_saturate_immediate(brw_reg_type, brw_reg*)::<anonymous union>::ud’ may be used uninitialized in this function [-Wmaybe-uninitialized]
       reg->dw1.ud = sat_imm.ud;
                               ^

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

i915: Fix GCC unused-but-set-variable warning in release build.

i915_fragprog.c: In function ‘i915ValidateFragmentProgram’:
i915_fragprog.c:1453:11: warning: variable ‘k’ set but not used [-Wunused-but-set-variable]
       int k;
           ^

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

Add macro for unused function attribute.

Suggested-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>

meta: Plug memory leak

It looks like this has existed since
commit f5a477ab76b6e0b268387699cd2253a43db0dfae
Author: Ian Romanick <ian.d.romanick@intel.com>
Date:   Mon Dec 16 11:54:08 2013 -0800

    meta: Refactor shader generation code out of mipmap generation path

Valgrind was complaining on fbo-generatemipmap-formats

v2: Instead, do the allocation after the early return block (v2)

Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

i965/fs: Don't issue FB writes for bound but unwritten color targets.

We used to loop over all color attachments, and emit FB writes for each
one, even if the shader didn't write to a corresponding output variable.
Those color attachments would be filled with garbage (undefined values).

Football Manager binds a framebuffer with 4 color attachments, but draws
to it using a shader that only writes to gl_FragData[0..2].  This meant
that color attachment 3 would be filled with garbage, resulting in
rendering artifacts.  Now we skip writing to it, fixing rendering.

Writes to gl_FragColor initialize outputs[0..nr_color_regions-1] to
GRFs, while writes to gl_FragData[i] initialize outputs[i].

Thanks to Jason Ekstrand for tracking this down.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=86747
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Cc: mesa-stable@lists.freedesktop.org

i965/fs: Make emit_shader_time_end() insert before EOT.

Previously, we emitted the shader-time epilogue from emit_fb_writes(),
during the middle of looping through color regions (or emit_urb_writes
for the VS).  This is duplicated several times and rather awkward.

I need to fix a bug in our FB write handling, and it will be a lot
easier if we move emit_shader_time_end() out of there.

Now, we simply emit FB writes/URB writes, and subsequently have
emit_shader_time_end() insert instructions before the final SEND with
EOT.  Not only is this simpler, it's actually a slight improvement:
we now include the MOVs to set up the final FB write payload in our
shader-time measurements.

Note that INTEL_DEBUG=shader_time only exists on Gen7+, and uses
send-from-GRF.  (In the past, we might have hit trouble where both
attempt to use MRFs for messages; that's not a problem now.)

v2: Rebase on v3 of the previous patch and other shader_time fixes.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com> [v1]
Acked-by: Matt Turner <mattst88@gmail.com>
Cc: mesa-stable@lists.freedesktop.org

i965/fs: Make get_timestamp() pass back the MOV rather than emitting it.

This makes another part of the INTEL_DEBUG=shader_time code emittable
at arbitrary locations, rather than just at the end of the instruction
stream.

v2: Don't lose smear!  Caught by Topi Pohjolainen.
v3: Don't set smear on the destination of the MOV.  Thanks Topi!

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Cc: mesa-stable@lists.freedesktop.org

i965/fs: Make emit_shader_time_write return rather than emit.

Instead of emit_shader_time_write, we now do emit(SHADER_TIME_ADD(...)).
The advantage is that we can also insert a shader time write at an
arbitrary location in the instruction stream, rather than being
restricted to emitting at the end.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Cc: mesa-stable@lists.freedesktop.org

i965/fs: Set smear on shader_time diff register.

The ADD(diff, diff, fs_reg(-2u)) instruction reads diff, which is a
width 1 register.  We need to read it as <0,1,0> with a subreg of 0,
which is what smear accomplishes.

Fixes assertion:
brw_eu_emit.c:285: validate_reg: Assertion `hstride == 0' failed.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=86974
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Cc: mesa-stable@lists.freedesktop.org

i965/fs: Set force_writemask_all on shader_time instructions.

These computations don't have anything to do with the currently
executing channels, so they should use force_writemask_all.

This fixes assert failures.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=86974
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Cc: mesa-stable@lists.freedesktop.org

r600g: Use R600_MAX_VIEWPORTS instead of 16

Lets define R600_MAX_VIEWPORTS instead of using 16 here and there
in the code when looping through viewports and scissors. It is
easier to understand what this number represents.

v2: Missed a case where R600_MAX_VIEWPORTS should have been used.

Signed-off-by: Alexandre Demers <alexandre.f.demers@gmail.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>

i915: Remove unused IS_GEN2 macro

Inspired by Damien's recent libdrm changes.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

i915: Remove (mostly) unused IS_915 macro

Inspired by Damien's recent libdrm changes.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

i915: Remove (mostly) unused IS_PNV, IS_PNVG, and IS_PNVGM macros

Inspired by Damien's recent libdrm changes.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

i915: Remove IS_9XX macro

Since the i915 / i965 split, IS_9XX just means IS_GEN3.  Inspired by
Damien's recent libdrm changes.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

i915: Remove unused IS_MOBILE macro

Inspired by Damien's recent libdrm changes.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

i965: Don't write past the end of the application supplied buffer

Both the AMD and Intel APIs provide a dataSize parameter, and this
function would merrily ignore it.  Neither API specifies what to do when
the buffer isn't big enough.  I take the easy route of writing all the
complete bits of data that will fit.  With more complete specs, we could
probably do something different.

I noticed this while looking into an unused parameter warning.  The
warning was actually useful!

brw_performance_monitor.c: In function 'brw_get_perf_monitor_result':
brw_performance_monitor.c:1261:37: warning: unused parameter 'data_size' [-Wunused-parameter]
                             GLsizei data_size,
                                     ^

v2: Fix checks to include offset in the calculation.  Noticed by Jan.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>

i965: Silence unused parameter warning

All dd functions take a gl_context as the first parameter.  Instead of
removing it, just silence the warning.

brw_performance_monitor.c: In function 'brw_new_perf_monitor':
brw_performance_monitor.c:1354:41: warning: unused parameter 'ctx' [-Wunused-parameter]
 brw_new_perf_monitor(struct gl_context *ctx)
                                         ^

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Carl Worth <cworth@cworth.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

i965: Silence many 'static' is not at beginning of declaration warnings

What a useful warning. #ThanksGCC

brw_performance_monitor.c:153:1: warning: 'static' is not at beginning of declaration [-Wold-style-declaration]
 const static struct gl_perf_monitor_counter gen5_raw_chaps_counters[] = {
 ^
brw_performance_monitor.c:185:1: warning: 'static' is not at beginning of declaration [-Wold-style-declaration]
 const static int gen5_oa_snapshot_layout[] =
 ^
brw_performance_monitor.c:221:1: warning: 'static' is not at beginning of declaration [-Wold-style-declaration]
 const static struct gl_perf_monitor_group gen5_groups[] = {
 ^
brw_performance_monitor.c:240:1: warning: 'static' is not at beginning of declaration [-Wold-style-declaration]
 const static struct gl_perf_monitor_counter gen6_raw_oa_counters[] = {
 ^
brw_performance_monitor.c:281:1: warning: 'static' is not at beginning of declaration [-Wold-style-declaration]
 const static int gen6_oa_snapshot_layout[] =
 ^
brw_performance_monitor.c:317:1: warning: 'static' is not at beginning of declaration [-Wold-style-declaration]
 const static struct gl_perf_monitor_counter gen6_statistics_counters[] = {
 ^
brw_performance_monitor.c:332:1: warning: 'static' is not at beginning of declaration [-Wold-style-declaration]
 const static int gen6_statistics_register_addresses[] = {
 ^
brw_performance_monitor.c:346:1: warning: 'static' is not at beginning of declaration [-Wold-style-declaration]
 const static struct gl_perf_monitor_group gen6_groups[] = {
 ^
brw_performance_monitor.c:356:1: warning: 'static' is not at beginning of declaration [-Wold-style-declaration]
 const static struct gl_perf_monitor_counter gen7_raw_oa_counters[] = {
 ^
brw_performance_monitor.c:402:1: warning: 'static' is not at beginning of declaration [-Wold-style-declaration]
 const static int gen7_oa_snapshot_layout[] =
 ^
brw_performance_monitor.c:470:1: warning: 'static' is not at beginning of declaration [-Wold-style-declaration]
 const static struct gl_perf_monitor_counter gen7_statistics_counters[] = {
 ^
brw_performance_monitor.c:493:1: warning: 'static' is not at beginning of declaration [-Wold-style-declaration]
 const static int gen7_statistics_register_addresses[] = {
 ^
brw_performance_monitor.c:515:1: warning: 'static' is not at beginning of declaration [-Wold-style-declaration]
 const static struct gl_perf_monitor_group gen7_groups[] = {
 ^

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Carl Worth <cworth@cworth.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

i965/fs: Silence unused parameter warning

I don't this opt_cmod_propagation_local ever used the fs_visitor.

brw_fs_cmod_propagation.cpp:52:40: warning: unused parameter 'v' [-Wunused-parameter]
 opt_cmod_propagation_local(fs_visitor *v, bblock_t *block)
                                        ^

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

i965/fs: Silence unused parameter warning

Unused since b18fd23.

brw_fs.cpp:2878:44: warning: unused parameter 'dispatch_width' [-Wunused-parameter]
 clear_deps_for_inst_src(fs_inst *inst, int dispatch_width, bool *deps,
                                            ^

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

i965/fs: Silence unused parameter warning

brw_fs_visitor.cpp:2162:56: warning: unused parameter 'offset_components' [-Wunused-parameter]
                          fs_reg offset_value, unsigned offset_components,
                                                        ^

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

main: Add entry point for TextureBufferRange.

v2: Review by Martin Peres
   - Get rid of difficult-to-follow code copied and pasted from
     the original TexBufferRange

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

main: Add check_texture_buffer_target.

Creates a shared function to ensure that texture buffer target is
GL_TEXTURE_BUFFER. Helps to clean up the Tex[ture]Buffer[Range] functions.

v2: Review from Anuj Phogat
   - Split rebase of Tex[ture]Buffer[Range]

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

main: Add check_texture_buffer_range.

Creates a shared function that TexBufferRange and TextureBufferRange can use
to check the buffer range. This cleans up TexBufferRange considerably.

v2: Review from Anuj Phogat
   - Split rebase of Tex[ture]Buffer[Range]

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

main: Cosmetic changes for Texture Buffers.

Adds a useful comment and some whitespace. Fixes an error message.

v2: Review from Anuj Phogat
   - Split rebase of Tex[ture]Buffer[Range]

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

main: Refactor _mesa_texture_buffer_range.

Changes how the caller is identified in error messages, moves a check for
ARB_texture_buffer_object from the entry points to the shared code in
_mesa_texture_buffer_range, and removes an unused argument (GLenum target).

v2: Review from Anuj Phogat
   - Split rebase of Tex[ture]Buffer[Range]

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

main: Use _mesa_lookup_bufferobj_err to simplify Tex[ture]Buffer[Range].

v2: Review from Anuj Phogat
   - Split rebase of Tex[ture]Buffer[Range]
   - Closing curly brace on the same line as else

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

main: Add utility function _mesa_lookup_bufferobj_err.

This function is exposed to mesa driver internals so that texture buffer
objects and array objects can use it.

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

main: Checking for cube completeness in GetCompressedTextureImage.

v2: Review from Anuj Phogat
   - Remove redundant copies of the cube map block comment
   - Replace redundant "if (!texImage) return;" statements with
     assert(texImage)

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

main: Add TEXTURE_CUBE_MAP support for glCompressedTextureSubImage3D.

v2: Review from Anuj Phogat
   - Remove redundant copies of the cube map block comment
   - Replace redundant "if (!texImage) return;" statements with
     assert(texImage)

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

main: assert(texImage) in ARB_DSA texture cube map functions.

ARB_direct_state_access functions that deal with texture cube
maps need to make sure that texture images are not NULL before operating on
them. In the following cases, the error check functions already throw an
error if texImage == NULL, so an assert can be raised instead.

v2: Review from Anuj Phogat
   - Replace redundant "if (!texImage) return;" statements with
     assert(texImage)

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

main: Remove redundant copy of cube map block comment in GetTextureImage.

The comment describing why ARB_direct_state_access texture cube map functions
use _mesa_cube_level_complete is very long.  To save room in the files,
readers are now referred to one central comment on texturesubimage in
teximage.c.

v2: Review from Anuj Phogat
   - Remove redundant copies of the cube map block comment

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

main: Remove redundant NumLayers checks.

ARB_direct_state_access texture functions that operate on cube maps no longer
need to verify that cube map texture objects contain six texture images
because _mesa_cube_level_complete now does that for them.

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

main: _mesa_cube_level_complete checks NumLayers.

_mesa_cube_level_complete now verifies that a cube map texture object actually
has six texture images before proceeding.

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

r300g: fix sRGB->sRGB blits

Cc: 10.5 10.4 <mesa-stable@lists.freedesktop.org>

r300g: fix a crash when resolving into an sRGB texture

Cc: 10.5 10.4 <mesa-stable@lists.freedesktop.org>

r300g: use memset for clearing the shader key

r300g: remove the broken SNORM->UNORM shader lowering pass

Not used anymore.

r300g: fix RGTC1 and LATC1 SNORM formats

Cc: 10.5 10.4 <mesa-stable@lists.freedesktop.org>

r300g: Fix the ATI1N swizzle (RGTC1 and LATC1)

This fixes the GL_COMPRESSED_RED_RGTC1 part of piglit's rgtc-teximage-01
test as well as the precision part of Wine's 3dc format test (fd.o bug
89156).

The Z component seems to contain a lower precision version of the
result, probably a temporary value from the decompression computation.
The Y and W component contain different data that depends on the input
values as well, but I could not make sense of them (Not that I tried
very hard).

GL_COMPRESSED_SIGNED_RED_RGTC1 still seems to have precision problems in
piglit, and both formats are affected by a compiler bug if they're
sampled by the shader with a swizzle other than .xyzw. Wine uses .xxxx,
which returns random garbage.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=89156
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Cc: 10.5 10.4 <mesa-stable@lists.freedesktop.org>

radeonsi: Add additional information to shader dumps

This adds SGPR count, VGPR count, shader size, LDS size, and scratch
usage to shader dumps.

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

radeonsi/compute: Use value from compiler for COMPUTE_PGM_RSRC1.FLOAT_MODE

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

clover: Return the minimum required value for CL_DEVICE_SINGLE_FP_CONFIG v2

This means dropping CL_FP_DENORM from the current return value.

v2:
  - Add comments about minimum values for OpenCL 1.2.

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>

freedreno/ir3: get the # of miplevels from getinfo

This fixes ARB_texture_query_levels to actually return the desired
value.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Rob Clark <robclark@freedesktop.org>
Cc: "10.4 10.5" <mesa-stable@lists.freedesktop.org>

freedreno/ir3: fix array count returned by TXQ

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Rob Clark <robclark@freedesktop.org>
Cc: "10.4 10.5" <mesa-stable@lists.freedesktop.org>

freedreno: move fb state copy after checking for size change

Fixes: 1f3ca56b ("freedreno: use util_copy_framebuffer_state()")
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Rob Clark <robclark@freedesktop.org>
Cc: "10.4 10.5" <mesa-stable@lists.freedesktop.org>

nir: Make the printer include nir_variable::location too.

Being able to see both location and driver_location can be useful when
debugging IO mistakes.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>

i965/fs: Implement SIMD16 dual source blending.

From the SNB PRM, volume 4, part 1, page 193:

"The dual source render target messages only have SIMD8 forms due to
 maximum message length limitations. SIMD16 pixel shaders must send two of
 these messages to cover all of the pixels. Each message contains two colors
 (4 channels each) for each pixel in the message payload."

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=82831
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

nir: Only do gl_FrontFacing workaround in glsl_to_nir for the FS.

Vertex shaders can have shader inputs where location happens to be
VARYING_SLOT_FACE.  Without predicating this on the shader stage,
we suddenly end up with load_front_face intrinsics in vertex shaders,
which is nonsensical.

Fixes spec/arb_vertex_buffer_object/pos-array when using NIR for VS.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>

nir: Plumb the shader stage into glsl_to_nir().

The next commit needs to know the shader stage in glsl_to_nir().
To facilitate that, we pass the gl_shader rather than the raw exec_list
of instructions.  This has both the exec_list and the stage.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jason Ekstrand <jason.ekstrand@intel.com>