radv: simplify fast clear shaders

There is no need for anything but a noop shader here.

vulkan/wsi: fix out of tree build.

radv: start using defines for the user sgpr offsets

This adds some comments and adds defines for the user sgprs,
so that we can move them around easier later and not have
to change/revalidate every one of these.

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

radv: port to common wsi codebase

This drops all the radv WSI code in favour of using
the new shared code that was ported from anv

This regresses Talos for now, Jason has pointed out
the bug is in Talos and we should wait for them to fix it.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv: move to using shared wsi code

This moves the shared code to a common subdirectory
and makes anv linked to that code instead of the copy
it was using.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi: remove all anv references from WSI common code

the WSI code should be now be clean for sharing.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv: move common wsi code to x11/wayland common files.

Next task is to rename all the anv_ out of this,
and move to a common location

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi/wayland: add callback to get device format properties.

This avoids having to know the toplevel API name.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi/wl: stop using device in more places

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi: split out surface creation to avoid instance API

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi: move further away from passing anv displays around

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi: split image alloc/free out to separate fns.

This moves these outside the wsi platform code, so we can reuse
that code

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi: switch to using VkDevice in swapchain

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi/x11: more refactoring to use generic handles

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi/x11: start refactoring out the image allocation/free functionality

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi: drop device from get format

Just use the wsi_device instead.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi: remove device from get_support interface

replace with wsi_device and allocator.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi/x11: abstract WSI interface from internals.

This allows the API and the internals to be split, and the
internals shared.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi/x11: push anv_device out of the init/finish routines

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi: abstract wsi interfaces away from device a bit more.

This is a step towards separating out the wsi code for sharing

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi/x11: push device out of x11 connection fns.

just pass the allocator/wsi_interface instead.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi: drop device from get caps

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

anv/wsi: drop get present modes device arg

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

radv/anv/wsi: drop unneeded parameter

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

draw: improve vertex fetch (v2)

The per-element fetch has quite some calculations which are constant,
these can be moved outside both the per-element as well as the main
shader loop (llvm can figure out it's constant mostly on its own, however
this can have a significant compile time cost).
Similarly, it looks easier swapping the fetch loops (outer loop per attrib,
inner loop filling up the per vertex elements - this way the aos->soa
conversion also can be done per attrib and not just at the end though again
this doesn't really make much of a difference in the generated code). (This
would also make it possible to vectorize the calculations leading to the
fetches.)
There's also some minimal change simplifying the overflow math slightly.
All in all, the generated code seems to look slightly simpler (depending
on the actual vs), but more importantly I've seen a significant reduction
in compile times for some vs (albeit with old (3.3) llvm version, and the
time reduction is only really for the optimizations run on the IR).
v2: adapt to other draw change.

No changes with piglit.

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

draw: improved handling of undefined inputs

Previous attempts to zero initialize all inputs were not really optimal
(though no performance impact was measurable). In fact this is not really
necessary, since we know the max number of inputs used.
Instead, just generate fetch for up to max inputs used by the shader,
directly replacing inputs for which there was no vertex element by zero.
This also cleans up key generation, which previously would have stored
some garbage for these elements.
And also drop the assertion which indicates such bogus usage by a
debug_printf (the whole point of initializing the undefined inputs was to
make this case safe to handle).

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

gallivm: print out time for jitting functions with GALLIVM_DEBUG=perf

Compilation to actual machine code can easily take as much time as the
optimization passes on the IR if not more, so print this out too.

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

gallivm: Use native packs and unpacks for the lerps

For the texturing packs, things looked pretty terrible. For every
lerp, we were repacking the values, and while those look sort of cheap
with 128bit, with 256bit we end up with 2 of them instead of just 1 but
worse, plus 2 extracts too (the unpack, however, works fine with a
single instruction, albeit only with llvm 3.8 - the vpmovzxbw).

Ideally we'd use more clever pack for llvmpipe backend conversion too
since we actually use the "wrong" shuffle (which is more work) when doing
the fs twiddle just so we end up with the wrong order for being able to
do native pack when converting from 2x8f -> 1x16b. But this requires some
refactoring, since the untwiddle is separate from conversion.

This is only used for avx2 256bit pack/unpack for now.

Improves openarena scores by 8% or so, though overall it's still pretty
disappointing how much faster 256bit vectors are even with avx2 (or
rather, aren't...). And, of course, eliminating the needless
packs/unpacks in the first place would eliminate most of that advantage
(not quite all) from this patch.

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

anv: drop pointless struct decl.

Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>

radv: drop pointless struct decl.

Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>

radv: move to using shared vk_alloc inlines.

This moves to the shared vk_alloc inlines for vulkan
memory allocations.

Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>

anv: move to using vk_alloc helpers.

This moves all the alloc/free in anv to the generic helpers.

Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>

vulkan: add vk_alloc.h shared allocation inlines.

vulkan allocation allows for overriding the allocator used,
add some macros for anv/radv to share for this.

Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>

anv: drop local MIN/MAX macros.

Use the ones from mesa, most places already did.

Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>

radv: drop local MIN/MAX macros.

Use the ones in macros.h instead.

Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>

util: move min/max/clamp macros to util macros.h

Although the vulkan drivers include mesa macros.h, for
radv I'd like to move away from that.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>

radv: make use of shared vector helper.

This removes the vector code from radv in favour of sharing
code with anv.

Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>

anv: port to using new u_vector shared helper.

This just removes the anv vector code and uses the new helper.

Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>

util: add vector util code.

This is ported from anv, both anv and radv can share this.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>

svga: minor code improvements in svga_validate_pipe_sampler_view()

Use the 'texture' local var in more places.
Rename 'pFormat' to 'viewFormat'.

Reviewed-by: Charmaine Lee <charmainel@vmware.com>

intel: genxml: add SAMPLER_BORDER_COLOR_STATE structures

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>

st/va: force to flush the last p frame in idr period

During dual instance encoding submission, if the second encode task and first
encode task have no reference dependency, e.g. p following with idr-frame,
there is a chance the second task will use for its reconstructed picture
buffer the same buffer used by first task for its reference/reconstructed
picture. In this case, buffer corruption may occur depending on encoding
speed. Fix is to force flush these two tasks separately to avoid race condition

Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=98005
Signed-off-by: Boyuan Zhang <boyuan.zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>

egl/surfaceless: Fix segfault in eglSwapBuffers

Since commit 63c5d5c6c46c8472ee7a8241a0f80f13d79cb8cd, the surfaceless
platform has allowed creation of pbuffer surfaces. But the vtable entry
for eglSwapBuffers has remained NULL.

Discovered by running a little pbuffer test.

Cc: Gurchetan Singh <gurchetansingh@chromium.org>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

radeonsi: rename prefixes from radeon to si

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Acked-by: Edward O'Callaghan <funfunctor@folklore1984.net>

radeonsi: merge radeon_llvm_context and si_shader_context

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Acked-by: Edward O'Callaghan <funfunctor@folklore1984.net>

radeonsi: import all TGSI->LLVM code from gallium/radeon

Acked-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Acked-by: Edward O'Callaghan <funfunctor@folklore1984.net>

gallium/radeon: simplify initialization of 64-bit gallivm builders

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Acked-by: Edward O'Callaghan <funfunctor@folklore1984.net>

gallium/radeon: remove unused radeon_llvm_reg_index_soa

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Acked-by: Edward O'Callaghan <funfunctor@folklore1984.net>

radeonsi: move LLVM ALU codegen into radeonsi

Acked-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Acked-by: Edward O'Callaghan <funfunctor@folklore1984.net>

genxml: add generated headers to EXTRA_DIST

Building the Mesa 12.0.3 distfile failed on a system without python
as generated files were not included in the distfile.

Cc: "12.0" <mesa-stable@lists.freedesktop.org>
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

mesa: automake: include mesa_glinterop.h in distfile

Add mesa_glinterop.h to the list of headers that will get included
in the distfile as it is required to build Mesa itself.

Corrects a regression introduced in a89faa2022fd995af2019c886b152b49a01f9392.

Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

egl: remove docs directory from EXTRA_DIST

The egl docs directory no longer exists as of
88b5c36fe1a1546bf633ee161a6715efc593acbd.

Remove it from EXTRA_DIST to unbreak 'make dist'

Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Tested-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

genxml: avoid using a GNU make pattern rule

% pattern rules are a GNU extension.  Convert the use of one to a
inference rule to allow this to build on OpenBSD.

This is a related change to the one made in
e3d43dc5eae5271e2c87bab702aa7409d3dd0b23

Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

configure.ac: use a single require_libdrm helper

Rather than having 4-5 places which do the explicit check/message just
polish the gallium helper and use it everywhere.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

configure.ac: remove no longer needed *_pci_id logic

Previously it was used to differentiate between the different codepaths
in the loader. Although strictly speaking the (core) of the loader is
only used when a hardware device is available. The latter of which in
itself requires libdrm (one of the codepaths available).

That said, all the configure toggles which relate to enabling/using hw
device should attribute and require libdrm, so there's no need to keep
this code around.

With this gallium_require_drm_loader becomes an empty stub, so nuke that
one as well.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

loader: cleanup copyright section

With previous patches nearly all the original code (as seen in the
various loaders) is gone.

Update the copyright/license section to reflect that.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

loader: remove loader_get_driver_for_fd() driver_type

Reminiscent from the pre-loader days, were we had multiple instances of
the loader logic in separate places and one could build a "GALLIUM_ONLY"
version.

Since that is no longer the case and the loaders (glx/egl/gbm) do not
(and should not) require to know any classic/gallium specific we can
drop the argument and the related code.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

loader: remove final sysfs codepath in loader_get_device_name_for_fd()

Effectively everyone with actual hardware and/or requesting the
"device_name" requires a working libdrm. Thus they could/should already
be using the (now only) codepath.

Apart from the code simplification, we can slim down our configure.ac
even further. But that will be done in separate patch(es).

Cc: Gary Wong <gtw@gnu.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

travis: remove no longer needed libudev-dev dependency

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

scons: remove all libudev references

Analogous to previous automake/autoconf commit.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

scons: loader: use libdrm when available

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

gbm: remove superfluous/incorrect udev comment

The gbm_device_get_backend_name() provides an (somewhat) internal name
of the implementation/backend used. Is has nothing to do with the udev,
one cannot and should not attempt to derive the name from it.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

automake: remove all the libudev references

As of last commit nothing in mesa depends on libudev.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

loader: remove libudev_get_device_name_for_fd and related code

With this all the libudev related code is now gone.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

loader: reimplement loader_get_user_preferred_fd via libdrm

Currently not everyone has libudev and with follow-up patches we'll
completely remove the divergent codepaths.

Use the libdrm drm device API to construct the required ID_PATH_TAG-like
string, to preserve the current functionality for libudev users and
allow others to benefit from it as well.

v2: Drop ranty comments, pick the correct device
v3: \n -> \0 in PCI_ID_PATH_TAG_LENGTH comment (Axel).
v4: Use snprintf (Nicolai)

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>

loader: annotate __driConfigOptionsLoader as static

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

loader: separate USE_DRICONF code into separate function

Improves readability and allows us to do further cleanups a lot easier.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

loader: slim down loader_get_pci_id_for_fd implementation(s)

Currently mesa has three code paths in the loader - libudev, manual
sysfs and drm ioctl one.

Considering the issues we had with libudev - strip those down in favour
of the libdrm drm device API. The latter can be implemented in any way
depending on the platform and can be reused by others.

v2: Use correct message on drmGetDevice failure. (Nicolai)

Cc: Jonathan Gray <jsg@jsg.id.au>
Cc: Jean-Sébastien Pédron <dumbbell@FreeBSD.org>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>

configure.ac: mark libdrm as have_pci_id provider

With follow on work, we'll untangle and simplify all the different
codepaths in loader. Then again, we forget to set have_pci_id when
libdrm is present (one of the codepaths available).

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Axel Davy <axel.davy@ens.fr>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

gm107/ir: fix bit offset of tex lod setting for indirect texturing

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

gm107/ir: fix texturing with indirect samplers

The indirect handle has to come right after the coordinates, so if there
was a sample/bias/depth compare/offset, everything would end up being
shifted by one argument position.

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

gallium/tgsi: add missing #include

Reviewed-by: Edward O'Callaghan <funfunctor@folklore1984.net>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

st/va: set default rt formats when calling vaCreateConfig

As specified in va.h, default value should be set on attributes
not present in the input list.

Signed-off-by: Julien Isorce <j.isorce@samsung.com>
Reviewed-by: Mark Thompson <sw@jkqxz.net>

i965: Fix gl_InvocationID in dual object GS where invocations == 1.

dEQP-GLES31.functional.geometry_shading.instanced.geometry_1_invocations
draws using a geometry shader that specifies

   layout(points, invocations = 1) in;

and then uses gl_InvocationID.  According to the Haswell PRM, the
"GS Instance ID 0" (and 1) thread payload fields are undefined in
dual object mode:

   "If 'dispatch mode' is DUAL_OBJECT this field is not valid."

But there's no point in using them - if there's only one invocation,
the ID will be 0.  So just load a constant.

Cc: mesa-stable@lists.freedesktop.org
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

anv: Get rid of anv_cmd_buffer_emit_state_base_address

All code that would have once called this can now call the gen-specific
version.  The switching version is no longer needed.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

anv/cmd_buffer: Move descriptor flushing into genX_cmd_buffer.c

It really should have gone here all along.  We were trying a bit too hard
to make it gen-agnostic just because it didn't have any #if's.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

anv/cmd_buffer: Expose ensure_push_constant_*

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

anv/cmd_buffer: Unify flush_compute_state across gens

With one small genxml change, the two versions were basically identical.
The only differences were one #define for HSW+ and a field that is missing
on Haswell but exists everywhere else.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

anv/cmd_buffer: Move Begin/End/Execute to genX_cmd_buffer.c

vkBeginCommandBuffer and vkCmdExecuteCommands both call into the
gen-specific emit_state_base_address function and vkEndCommandBuffer
belongs with begin.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

anv/cmd_buffer: Move state base address re-emit into ExecuteCommands

This has two primary advantages.  First, it means that the batch_chain code
knows less about the actual command buffer contents which is good because
improves separation.  Second, it means that it only gets re-emitted once
after all of the secondaries instead of once after each secondary which is
just wasteful.  It also has the advantage of cleaning the code up a bit.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com>

doc/features.txt: factor out radeonsi as GL45 complete

V2. add i965/hsw+ to list
V3. rebased on master.
V4. 'DONE' -> 'DONE ()'.
V5. remove i965/hsw+ from list :/

Signed-off-by: Edward O'Callaghan <funfunctor@folklore1984.net>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

i965: Silence unused parameter warnings

brw_link.cpp:76:44: warning: unused parameter ‘shader_type’ [-Wunused-parameter]
                            gl_shader_stage shader_type,
                                            ^
brw_nir.c: In function ‘brw_nir_lower_vs_inputs’:
brw_nir.c:194:55: warning: unused parameter ‘devinfo’ [-Wunused-parameter]
                         const struct gen_device_info *devinfo,
                                                       ^
brw_vec4_visitor.cpp:914:37: warning: unused parameter ‘sampler’ [-Wunused-parameter]
                            uint32_t sampler,
                                     ^
brw_vec4_visitor.cpp:1146:34: warning: unused parameter ‘stream_id’ [-Wunused-parameter]
 vec4_visitor::gs_emit_vertex(int stream_id)
                                  ^

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

glsl: Remove unused function import_prototypes

Once upon a time, this was used to extract prototypes from the shader
containing GLSL built-in functions.  This was removed by f5692f45 in
November 2010 for Mesa 7.10.

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

glsl: Remove prototypes for nonexistent functions

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

glsl: Replace assert with unreachable

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

anv: replace , with ; in anv_batch_emit()

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel: aubinator: use different colors to signal batch start/end

This makes the stream of commands a bit easier to read.

v2 (Ken): Use bold text on green headers for easier readability;
          swap the green and blue headers so the majority stay blue.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

st/glsl_to_tgsi: fix [ui]vec[34] conversion to double

The corresponding opcodes for integers need to be treated the same as F2D.

Fixes GL45-CTS.gpu_shader_fp64.conversions.

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

st/glsl_to_tgsi: fix atomic counter addressing

When more than one atomic counter buffer is in use, UniformStorage[n].opaque
is set up to contain indices that are contiguous across all used buffers.

This appears to be used by i965 via NIR, but for TGSI we do not treat atomic
counter buffers as opaque, so using the data in the opaque array is incorrect.

Fixes GL45-CTS.compute_shader.resource-atomic-counter.

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

st/glsl_to_tgsi: fix a corner case of std140 layout in uniform buffers

See the comment in the code for an explanation. This fixes
GL45-CTS.buffer_storage.map_persistent_draw.

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

st/mesa: fix fragment shader output mapping

Properly handle the case where there is a gap in the assigned output locations,
e.g. a fragment shader writes to color buffer 2 but not to color buffers 0 & 1.

Fixes GL45-CTS.gtf33.GL3Tests.explicit_attrib_location.explicit_attrib_location_pipeline.

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

glsl: print non-zero bindings of variables

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

radeonsi: unify the constant load paths

Remove the split between direct and indirect.

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

radeonsi: fix indirect loads of 64 bit constants

This fixes GL45-CTS.compute_shader.fp64-case3.

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

gbm: add a couple missing includes

Needed for memset() and drmIoctl().

Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Edward O'Callaghan <funfunctor@folklore1984.net>

glsl: fail compilation of compute shaders when unsupported

Generally, we only check for the presence of compute shaders during
parsing when we find any language (like layout qualifiers) that are
specific to compute shaders, however, it is possible to define an
empty compute shader does not use any language specific to compute
shaders at all and we should fail the compilation anyway. dEQP checks
this.

This patch adds a check for compute shader availability after we have
parsed the source code. At this point we know the effective GLSL version
and also extensions enabled in the shader.

Fixes a subcase of the following dEQP tests:
dEQP-GLES31.functional.debug.negative_coverage.callbacks.shader.compile_compute_shader
dEQP-GLES31.functional.debug.negative_coverage.get_error.shader.compile_compute_shader
dEQP-GLES31.functional.debug.negative_coverage.log.shader.compile_compute_shader

The tests still fail because there is one more subcase that fails that needs
another fix.

Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>

egl/android: fix error in droid_add_configs_for_visuals()

This was some kind of leftover in commit acd35c8 and format_count
array variable (declared in outer scope) should be used instead.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Fixes: acd35c8758dc73240903 ("egl/android: tweak droid_add_configs_for_visuals()")
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

radeonsi: shorten "shader->selector" to "sel" in si_shader_create

Reviewed-by: Edward O'Callaghan <funfunctor@folklore1984.net>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

radeonsi: clear DB_RENDER_OVERRIDE

Vulkan doesn't set these fields even though it doesn't use HiS.
HiS is disabled by programming DB_SRESULTS_COMPARE_STATEn to 0.

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

glsl: Disable textureOffset(sampler2DArrayShadow, ...) in GLSL ES.

This has apparently never existed in GLSL ES.

Fixes dEQP-GLES3.functional.shaders.texture_functions.invalid
.textureoffset_sampler2darrayshadow_vec4_ivec2_vertex and
.textureoffset_sampler2darrayshadow_vec4_ivec2_fragment

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98244
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>