st/mesa: don't call translate_*_program functions for NIR

move the initializaton to st_link_nir

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

st/mesa: finalize NIR after shader variant passes for TCS/TES/GS/CS

Same as VS and FS.

This might fix vertex color clamping.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

st/mesa: unify transform feedback info translation code

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

st/mesa: move vertex program preparation code into st_prepare_vertex_program

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

st/mesa: clean up more after the removal of st_compute_program

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

st/mesa: deduplicate st_common_program code in st_program_string_notify

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

st/mesa: sink TCS/TES/GS/CS translate code into st_translate_common_program

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

st/mesa: deduplicate cases in st_deserialise_ir_program

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

st/mesa: remove st_compute_program in favor of st_common_program

The conversion from pipe_shader_state to pipe_compute_state is done
at the end.

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

st/mesa: don't store stream output info to shader cache for tess ctrl shaders

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

st/mesa: simplify the signature of st_release_basic_variants

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

st/mesa: deduplicate code for ATI fs in st_program_string_notify

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

st/mesa: use *prog at the end of st_link_nir

Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>

appveyor: Cache meson's wrap downloads

This makes us less reliant on wrap-db (and reduces the amount of
downloading that goes on during the build).

Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/1936
Acked-by: Eric Engestrom <eric.engestrom@intel.com>

gitlab-ci: Set the meson wrapmode to disabled

This will prevent us from accidentally falling back to the wrap-db
instead of using locally installed versions.

Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

Revert "gitlab-ci: Disable meson-mingw32-x86_64 job again for now"

This reverts commit d60b8679a46212d992dcd94b9d3eeebfcff3f00a.

Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

gitlab-ci: Add a pkg-config for mingw

The one debian provides is broken in buster+, so I've just written my
own. This allows meson to find the installed zlib and prevents it from
falling back to wraps.

Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

meson: Don't use expat on windows

It's not really needed, and there's no debian package for it so we're
forced to fall back to wraps in mesa's CI. This can be problematic in
itself.

Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

st/mesa: fix crash for drivers supporting nir defaulting to tgsi

nvc0 and I assume radeonsi as well hit an assert inside glsl_to_tgsi as atan
instructions get inserted into the shader.

Fixes: cece947a8dfa ("glsl/builtin: Add alternate versions of atan using new ops")
Cc: Neil Roberts <nroberts@igalia.com>
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>

util/u_atomic: fix return type of p_atomic_{inc,dec}_return() and p_atomic_{cmp,}xchg()

We're trying to cast the return type to the type of the var, but instead
we were casting `sizeof(*v)`.

Fixes: 6df72e970c0e2794a565 ("util: Make u_atomic.h typeless.")
Fixes: 0a7f17cf5b591330a257 ("util/u_atomic: add p_atomic_xchg")
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>

mesa/math: delete duplicate extern symbol

It's already defined in `m_debug_util.h`, along with an explanation of
what it is and how to use it.

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>

mesa/math: delete leftover... from 18 years ago (!)

Left over from 0a79baf1bff93718e50a ("remove dead vertex assembly").

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>

lima: Fix crash when there are no vertex shader attributes

Signed-off-by: Andreas Baierl <ichgeh@imkreisrum.de>
Reviewed-by: Vasily Khoruzhick <anarsoul@gmail.com>

lima: Fix compiler warning in standalone compiler

'struct lima_context' has to be declared before usage in lima_program.h

Signed-off-by: Andreas Baierl <ichgeh@imkreisrum.de>
Reviewed-by: Vasily Khoruzhick <anarsoul@gmail.com>

aco: emit_split_vector() s_memtime results

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>

aco: don't CSE s_memtime

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>

aco: fix scheduling with s_memtime/s_memrealtime

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>

intel/common: include unistd.h for ioctl() prototype on Solaris

Fixes build errors of:
In file included from ../src/intel/vulkan/anv_private.h:48,
                 from ../src/intel/vulkan/genX_blorp_exec.c:26:
../src/intel/common/gen_gem.h: In function ‘gen_ioctl’:
../src/intel/common/gen_gem.h:68:15: error: implicit declaration of function ‘ioctl’ [-Werror=implicit-function-declaration]
   68 |         ret = ioctl(fd, request, arg);
      |               ^~~~~
In file included from ../include/c11/threads_posix.h:35,
                 from ../include/c11/threads.h:66,
                 from ../src/mesa/main/mtypes.h:39,
                 from ../src/intel/compiler/brw_compiler.h:30,
                 from ../src/intel/vulkan/anv_private.h:51,
                 from ../src/intel/vulkan/genX_blorp_exec.c:26:
/usr/include/unistd.h: At top level:
/usr/include/unistd.h:471:12: error: conflicting types for ‘ioctl’
  471 | extern int ioctl(int, int, ...);
      |            ^~~~~
/usr/include/unistd.h:471:1: note: a parameter list with an ellipsis can’t match an empty parameter name list declaration
  471 | extern int ioctl(int, int, ...);
      | ^~~~~~
In file included from ../src/intel/vulkan/anv_private.h:48,
                 from ../src/intel/vulkan/genX_blorp_exec.c:26:
../src/intel/common/gen_gem.h:68:15: note: previous implicit declaration of ‘ioctl’ was here
   68 |         ret = ioctl(fd, request, arg);
      |               ^~~~~

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>

meson: recognize "sunos" as the system name for Solaris

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>

util: Solaris has linux-style pthread_setname_np

Fixes: dcf9d91a ("util: Handle differences in pthread_setname_np")
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>

util: Workaround lack of flock on Solaris

v2: Replace autoconf check for flock() with meson check

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>

util: Make Solaris implemention of p_atomic_add work with gcc

gcc is very particular about where you place the (void) cast
The previous placement made it error out with:

In file included from disk_cache.c:40:0:
../../src/util/u_atomic.h:203:29: error: void value not ignored as it ought to be
 #define p_atomic_add(v, i) ((void)         \
                              ^
disk_cache.c:658:4: note: in expansion of macro ‘p_atomic_add’
    p_atomic_add(cache->size, size);
    ^

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>

c99_compat.h: Don't try to use 'restrict' in C++ code

Fixes build failures on Solaris in C++ files using gcc:

../src/util/u_math.h:628:41: error: expected ‘,’ or ‘...’ before ‘dest’
  628 | util_memcpy_cpu_to_le32(void * restrict dest, const void * restrict src, size_t n)
      |                                         ^~~~
../src/util/u_math.h: In function ‘void* util_memcpy_cpu_to_le32(void*)’:
../src/util/u_math.h:641:18: error: ‘dest’ was not declared in this scope
  641 |    return memcpy(dest, src, n);
      |                  ^~~~
../src/util/u_math.h:641:24: error: ‘src’ was not declared in this scope
  641 |    return memcpy(dest, src, n);
      |                        ^~~
../src/util/u_math.h:641:29: error: ‘n’ was not declared in this scope; did you mean ‘yn’?
  641 |    return memcpy(dest, src, n);
      |                             ^
      |                             yn

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>

pan/midgard: Do not repeatedly spill same value

It doesn't make sense. You already spilled it once, and it didn't help.
Don't try again, or you'll end up in a loop.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

pan/midgard: Fix memory corruption in register spilling

Essentially an off-by-one error ... bit of an edge case, but seems to
occur in some glamor shaders.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

pan/midgard: Allow COMPUTE jobs in panfrost_bo_access_for_stage

Fixes: ada752afe40 ("panfrost: Extend the panfrost_batch_add_bo() API to pass access flags")
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

pan/midgard: Use 16-bit liveness masks

We'll want liveness per-byte, so we need to accomodate up to 16 bytes.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

panfrost: Disable frame throttling

The new frame throttling implemention interacts unfortunately with
pipelining, leading to fence fds leaking like crazy and ultimately apps
crashing quickly.

With this patch, apps still crash but not as quickly. We need to either
figure out the real cause or revert the core changes.

Nevertheless, we don't want frame throttling in the first place, so.

Fixes: a65e29ccb26 ("gallium: simplify throttle implementation")
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

mesa: fix invalid target error handling for teximage

This commit moves the target check before using _mesa_get_current_tex_object
to fix a "Mesa implementation error: bad target in _mesa_get_current_tex_object()"
error.

Fixes: 9dd1f7cec01 ("mesa: pass gl_texture_object as arg to not depend on state")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

radeonsi/nir: simplify si_lower_nir signature

just a cleanup

pan/midgard: Fix mir_mask_of_read_components with dot products

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

pan/midgard: Add perspective ops to mir_get_swizzle

I really need to just make this a table..

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

pan/midgard: Don't try to propagate swizzles to branches

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

pan/midgard: Allow non-contiguous masks in UBO lowering

We don't really need to impose this condition, but we do need to cope
with the slightly more general case.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

pan/midgard: Report read mask for branch arguments

Conditionals in particular read values.

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>

iris: finish aux import on get_param

A buffer and its aux are imported separately, if the aux import is
not completed yet when resource_get_param is called, merge the
separate aux a.k.a the 2nd image into the main image.

Fixes: 246eebba4a8 ("iris: Export and import surfaces with modifiers that have aux data")
Signed-off-by: James Xiong <james.xiong@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Jordan Justen <jordan.l.justen@intel.com>

mesa: Handle pbuffers in desktop GL framebuffer attachment queries

Once again, we were handling back-to-front in the GLES3 case, but not
the desktop GL case.

Fixes GTF-GL46.gtf30.GL3Tests.framebuffer_srgb.framebuffer_srgb_default_encoding when run with --deqp-surface-type=pbuffer --deqp-gl-context-type=egl.

Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

mesa: Make back_to_front_if_single_buffered non-static

So I can use it in fbobject.c in the next commit.

Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

mesa: Use ctx->ReadBuffer in glReadBuffer back-to-front tests

We were looking at ctx->DrawBuffer when asking about the read buffer,
which was good enough for CTS purposes, but definitely not right.

Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

etnaviv: remove variable from global namespace

Found out by accident this was clashing with another driver.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Cc: <mesa-stable@lists.freedesktop.org>

st/mesa: always allocate pack/unpack buffers as staging

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>

gallium/xlib: Fix xmesa drawable creation

The first time you call glXMakeCurrent, current != ctx. As a result we
would never look up whether the drawable already had an XMesaDrawable,
and would instead always create one. Then XMesaBufferList would have two
different buffers for the same XID, and you'd be reading and drawing to
different places, and that's not what you want at all.

Instead just always look up the drawable.

Fixes: db8be355 (gallium/xlib: Remove drawable caching from the MakeCurrent path)
Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/1196
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>

gitlab-ci: set a common job parent for test stage

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>

gitlab-ci: set a common job parent for build stage

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>

gitlab-ci: set a common job parent for container stage

While at it, rename to singular "container" for consistency.

Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>

Revert "radv: do not emit PKT3_CONTEXT_CONTROL with AMDGPU 3.6.0+"

This reverts commit 2ca8629fa9b303e24783b76a7b3b0c2513e32fbd.

This was initially ported from RadeonSI, but in the meantime it has
been reverted because it might hang. Be conservative and re-introduce
this packet emission.

Unfortunately this doesn't fix anything known.

Cc: 19.2 <mesa-stable@lists.freedesktop.org>
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

spirv: set correct dest_type for texture query ops

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>

turnip: more descriptor sets

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: push constants

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: depth/stencil

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: basic msaa working

Not perfect but gets through some tests.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: improve CmdCopyImage and implement CmdBlitImage

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: use nir_assign_io_var_locations instead of nir_assign_var_locations

Variables with same location should use the same driver_location.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: add missing nir passes

Avoids assert fails in ir3.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: add code to lower indirect samplers

Taken from nir_lower_samplers. Sampler arrays don't work though, this is
just to avoid an assert fail in ir3.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: fixup consts

Fix some mistakes in previous series.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: update some shader state bits from GL driver

Notably includes centroid varying bits that were missing.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: Emit clears of gmem using linear.

This is what we do in freedreno.

Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: Set up the correct tiling mode for small attachments.

Noticed while debugging a tiling-looking issue by comparing our gmem
blit setup to freedreno's.

Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: Tell spirv_to_nir that we want fragcoord as a sysval.

Fixes ir3 compiler failure failure in
dEQP-VK.renderpass.dedicated_allocation.formats.r8g8b8a8_unorm.clear.clear_draw
(now just a rendering failure where the subpass clear isn't happening)

Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: Fill in clear color packing for r10g11b11 and rgb9e5.

Fixes assertion failures in
dEQP-VK.api.image_clearing.core.clear_color_image.2d.* for these
formats, though the test set as a whole is stil failing.

Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: Drop unused tu_pack_clear_value() return.

Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: add anisotropy and compressed formats to device features

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: disable tiling as necessary

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: update setup_slices

Deal with tiled r8g8 having different alignment and other updates taken
from fd6_resource. Additionally track image samples/cpp.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: add VK_KHR_sampler_mirror_clamp_to_edge

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: add black border color

Avoids hangs and some texture tests are happy with just this.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: improve sampler descriptor

Fixes anisotropy and shadow texture

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: improve view descriptor

Changes to make compressed, tiled, 3d, etc textures work

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: add more 2d_ifmt translations

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: format table fixes

* Fix R16G16 SCALED and R16G16B16A16 SCALED having texture format
* Fix B5G6R5 swap value
* Use R8_UINT instead of R8_UNORM for S8_UINT rb format
* Disable 96-bit texture formats instead having a check for NPOT formats
* Don't fail assert on D24X8 format

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: add format_is_uint/format_is_sint

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: add astc format layout

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: fix assert failing for 0 color attachments

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: fix segmentation fault with compute pipeline

Not supported, so always set pointer to NULL

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: fix segmentation fault in events

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: fix 32 vertex attributes case

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

turnip: fix triangle strip

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

freedreno/regs: update a6xx 2d blit bits

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>

radv: rename VK_KHR_shader_float16_int8 structs/constants

Trivial change.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

v3d: drop unused shader_rec_count member from context

Looks like this was copied from the vc4 driver where it is actually
included in the submit CL ioctl.

Reviewed-by: Eric Anholt <eric@anholt.net>

freedreno/ir3: implement fquantize2f16

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Rob Clark <robclark@gmail.com>

freedreno/ir3: implement texop_texture_samples

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Rob Clark <robclark@gmail.com>

freedreno/ir3: fix GETLOD for negative LODs

Note: for output type U32, negative LOD is not sign extended from 16 bits

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Rob Clark <robclark@gmail.com>

freedreno/ir3: implement fdd{x,y}_coarse opcodes

Same as regular fddx/fddy.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Rob Clark <robclark@gmail.com>

freedreno/ir3: increase size of inputs/outputs arrays

Makes it possible to support 32 varyings.

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Rob Clark <robclark@gmail.com>

freedreno/ir3: remove input ncomp field

Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Rob Clark <robclark@gmail.com>

etnaviv: fix vertex buffer state emission for single stream GPUs

GPUs with a single supported vertex stream must use the single state
address to program the stream.

Fixes: 3d09bb390a39 (etnaviv: GC7000: State changes for HALTI3..5)
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Jonathan Marek <jonathan@marek.ca>

gallivm/draw/swr: make the gs_iface not depend on tgsi.

This gs_iface doesn't seem to require a dependence on the tgsi
context, except for the swr end prim code.

This refactors the API to include all the info that the swr
code needs in the interface rather than having to dig it out of
the struct inheritance.

This is a precursor to adding NIR support to llvmpipe.

Reviewed-by: Jan Zielinski <jan.zielinski@intel.com>

iris: Implement the Gen < 9 tessellation quads workaround

Fixes several CTS tests:
- KHR-GL46.tessellation_shader.vertex.vertex_spacing
- KHR-GL46.tessellation_shader.tessellation_shader_point_mode.points_verification

Fixes: 823609b1a39 ("iris/WIP: add broadwell support")