mesa/teximage: show internal format when printing verbose api log

This would have saved me some time to determine that the problem was
coming from lack of RGBA32F / 8x MSAA support on nv50.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8405>

radv: enable DCC for layered color images on GFX10+

There is still some CTS failures on GFX9.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8368>

radv: do not use predication when the range doesn't cover the whole image

The predication is based on the mip level, so if the image has layers
and DCC is enabled, it should only be used if the range of layers
covers the whole image.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8368>

radv: clean up radv_decompress_dcc_compute()

Remove one old comment because it supports decompressing layers.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8368>

radv: add support for fast-clearing DCC layers on GFX9+

Layers are contiguous in memory.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8368>

ac/surface: initialize dcc_slice_size on GFX9+

Will be used by RADV to implement DCC layers.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8368>

wsi/x11: Make sure wsi_x11_connection::is_xwayland is always initialized

We wouldn't initialize it if the X server didn't support the RANDR
extension (though that's unlikely these days).

Fixes: b5268d532a01 "wsi/x11: Detect Xwayland"
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8393>

wsi/x11: Always free randr_reply in wsi_x11_connection_create

Fixes memory leak.

Fixes: b5268d532a01 "wsi/x11: Detect Xwayland"
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8393>

wsi/x11: Use wsi_x11_get_connection in x11_present_to_x11_dri3

Instead of creating a new wsi_x11_connection every time.

Fixes memory leak and bad performance.

Fixes: 4292fb213928 "wsi/x11: Use PresentOptionAsync for MAILBOX present mode with Xwayland"
Reported-by: Chris Rankin <rankincj@yahoo.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8393>

nir/lower_vec_to_movs: don't vectorize unsupports ops

If the instruction being coalesced would be vectorized but the target
doesn't support vectorizing that op, skip coalescing.
Reuse the callbacks from alu_to_scalar to describe which ops should not
be vectorized.

Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Eric Anholt <eric@anholt.net>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6506>

aco: Spill more optimally before loops.

This further reduces the dead code emitted by the spiller.
Some minimal amount of dead IR is still emitted sometimes,
but that doesn't generate any compiled code at the end.

Totals from 1953 (1.40% of 139391) affected shaders:
VGPRs: 206980 -> 206588 (-0.19%)
SpillSGPRs: 24719 -> 16423 (-33.56%); split: -33.58%, +0.02%
CodeSize: 28448516 -> 28343836 (-0.37%); split: -0.38%, +0.01%
MaxWaves: 8960 -> 8992 (+0.36%)
Instrs: 5422049 -> 5408334 (-0.25%); split: -0.26%, +0.01%
Cycles: 511240864 -> 512460764 (+0.24%); split: -0.02%, +0.26%
VMEM: 346681 -> 346468 (-0.06%); split: +0.27%, -0.33%
SMEM: 124160 -> 122802 (-1.09%); split: +0.33%, -1.42%
VClause: 81102 -> 81163 (+0.08%); split: -0.01%, +0.09%
SClause: 174404 -> 174237 (-0.10%); split: -0.23%, +0.13%
Copies: 530216 -> 532961 (+0.52%); split: -0.90%, +1.42%
Branches: 189114 -> 189221 (+0.06%); split: -0.13%, +0.18%
PreSGPRs: 206017 -> 206526 (+0.25%); split: -0.08%, +0.33%
PreVGPRs: 183103 -> 182964 (-0.08%)

Co-authored-by: Daniel Schürmann <daniel@schuermann.dev>
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8026>

aco: Keep live-though variables and constants spilled.

This noticably reduces the amount of dead code emitted by our
spiller, when eg. previously a constant was spilled then
rematerialized before a loop, but then spilled again inside the loop.

Fossil DB changes on Navi 10:
Totals from 263 (0.19% of 139391) affected shaders:
VGPRs: 30044 -> 30028 (-0.05%)
SpillSGPRs: 8800 -> 4948 (-43.77%)
CodeSize: 4496040 -> 4335448 (-3.57%); split: -3.57%, +0.00%
Instrs: 843942 -> 819219 (-2.93%); split: -2.93%, +0.00%
Cycles: 76485744 -> 73549080 (-3.84%); split: -4.04%, +0.20%
VMEM: 38204 -> 38147 (-0.15%); split: +0.08%, -0.23%
SMEM: 17872 -> 17959 (+0.49%)
SClause: 24298 -> 24012 (-1.18%)
Copies: 98023 -> 82960 (-15.37%); split: -15.38%, +0.01%
Branches: 29074 -> 27632 (-4.96%)
PreVGPRs: 25291 -> 25241 (-0.20%)

Co-authored-by: Daniel Schürmann <daniel@schuermann.dev>
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8026>

radv: Enable sparse buffer and image support.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7953>

radv: Add sparse image queries.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7953>

radv: Add image sparse memory update implementation.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7953>

radv: Create sparse images.

Disable all metadata for now.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7953>

radv/winsys: Fix offset in range merging.

If we change the virtual address we also have to change the offset in the buffer
to be mapped.

Fixes: 715df30a4e2 "radv/amdgpu: Add winsys implementation of virtual buffers."
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7953>

radv/winsys: Fix inequality for sparse buffer remapping.

Found a case where we mapped a range too many.

Per the comment the constraint is:

/* [first, last] is exactly the range of ranges that either overlap the
 * new parent, or are adjacent to it. This corresponds to the bind ranges
 * that may change.
 */

So that means that after the ++last we the ranges[last] should still
be adjacent. So we need to test the post-increment value to see whether
it is adjacent.

Failure case:
  ranges:
    0: 0 - ffff
    1: 10000 - 1ffff
    2: 20000 - 2ffff
    3: 30000 - 3ffff
  new range: 10000 - 1ffff

wrong first, last: 0,3
  However range 3 clearly isn't adjacent at all.

Fixes: 715df30a4e2 "radv/amdgpu: Add winsys implementation of virtual buffers."
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7953>

ac/surf: Use correct tilemodes on GFX8 for PRT.

Otherwise addrlib will assign the non-PRT tiling indices anyway ...

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7953>

ac/surf: Add sparse texture info to radeon_surf.

For GFX9 I didn't reuse the existing mipmap offset/pitch because
last time we did that there was a revert request from Marek.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7953>

ac/surf: Implement PRT layout.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7953>

ac/surf: Prepare for 64-bit flags.

Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7953>

radv: only re-initialize HTILE after ds compute resolves if compressed

If the current layout isn't compressed we don't have to re-initialize
the HTILE metadata.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8389>

radv: re-initialize HTILE properly after depth/stencil compute resolves

This was added to workaround some CTS failures which no longer happen.
Note that radv_clear_htile() will only clear the depth or stencil
bytes of the HTILE buffer based on the aspect.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8389>

radv: disable TC-compat HTILE in GENERAL for Detroit: Become Human

The game has invalid usage of render loops and enabling TC-compat
HTILE in GENERAL introduces rendering issues.

Gitlab: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3063
Cc: mesa-stable
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8359>

radv: fix crashes when fast-clearing in a secondary command buffer

iview can be NULL inside a secondary command buffer.

Fixes: 00064713a30 ("radv: determine at creation if an image view can be fast cleared")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8408>

zink: use hardware int64 when supported

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8403>

zink: add missing 64-bit integer ops

This seems to be all that's missing to support hardware int64 instead of
lowering it.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8403>

zink: add missing opcodes

This fixes a bunch of conversion-tests in the arb_gpu_shader_int64
piglit test-suite for me.

Fixes: 55e2d836f37 ("zink: enable 64bit pipe caps")
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8403>

dri: enable glthread + radeonsi workaround for CS:GO

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4021
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8383>

ac/rgp: add missing include

The build would fail without this include if -std=gnu17 is used.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4057
Fixes: ffdfe136e67 ("ac/sqtt: move rgp/sqtt def to ac")
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8376>

ci: re-mark some depth/stencil resolve CTS as expected failures

They were removed during the last update but they still fail.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8385>

radv: do not predicate FMASK decompression when DCC+MSAA is used

Even if the FCE predicate is FALSE, we might still need to decompress
FMASK if compressed rendering was used. FMASK decompressions should
never been predicated.

This fixes a ton of CTS failures and a rendering issue with Control
when DCC+MSAA is force-enabled.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8331>

radv: determine at creation if an image view can be fast cleared

This can be determined earlier than every time a clear is performed
by the driver, it probably saves a bunch of CPU cycles.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8370>

ci: Add a few more drivers to the cross builds

This is mostly to get additional -Werror coverage to avoid introducing
unforced ILP32 or big-endian errors. i386 adds lavapipe, r600, nouveau,
zink, and all the classic drivers. ppc64le adds lavapipe and zink, and
also adds -Werror for symmetry with the other cross builds. s390x also
adds lavapipe and zink.

Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8394>

zink: Fix VK_FORMAT_A8B8G8R8_SRGB_PACK32 mapping on big-endian

The PACK32 Vulkan formats need to map to the corresponding XYZWxxxx pipe
formats, since the latter take endianness into account.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8394>

nouveau: pacify gcc on ILP32

The gcc we're using (and quite possibly newer ones) throws a really
stupid error:

   ../src/gallium/drivers/nouveau/nouveau_buffer.c:765:22: error: cast from pointer to integer of different size [-Werror=pointer-to-int-cast]
      buffer->address = (uint64_t)user_ptr;

Which... address is a uint64_t, and user_ptr is a void *, so this is
completely unambiguously safe to do. Apparently casting to uintptr_t
squelches this, so do that instead.

Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8394>

etnaviv: add set_stream_output_targets(..) stub

When running gles3 deqp's with ETNA_MESA_DEBUG=deqp we fake streamout support.
CSO thinks that streamout is supported and calls ctx->pipe->set_stream_output_targets(..)
in cso_destroy_context(..) which results in a null-pointer access.

Add a stub to make development easier.

Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8382>

zink: setup compiler options during init

This avoids duplicating the options just to change a single one.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8402>

zink: rename zink vs pipe variables

The general convention we use is "pscreen" for the pipe_screen, and
"screen" for the zink_screen, so let's stick with that here also.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8402>

util/u_thread: Disable pthread_barrier_t on Haiku

* Causes a lockup on SMP systems

Reviewed-by: Alexander von Gluck IV <kallisti5@unixzen.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8323>

hgl: Major refactor and cleanup

* Drop old-timey GLDisplatcher
* Refactor haiku-softpipe fixing some hacks
* Bubble BBitmap up to winsys

Reviewed-by: Alexander von Gluck IV <kallisti5@unixzen.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8323>

targets/haiku-softpipe: Restore GalliumContext

Reviewed-by: Alexander von Gluck IV <kallisti5@unixzen.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8323>

aux/driver_ddebug: Normalize pid type from Haiku

Reviewed-by: Alexander von Gluck IV <kallisti5@unixzen.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8323>

frontends/hgl: set framebuffer id

Reviewed-by: Alexander von Gluck IV <kallisti5@unixzen.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8323>

frontends/hgl: set state_manager

Reviewed-by: Alexander von Gluck IV <kallisti5@unixzen.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8323>

hgl: use local headers instead of system header

* We should choose the header we're building with
  and not the one installed by a previous mesa install

Reviewed-by: Alexander von Gluck IV <kallisti5@unixzen.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8323>

include: fix export in Haiku OpenGL kit headers

Reviewed-by: Alexander von Gluck IV <kallisti5@unixzen.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8323>

meson: fix Haiku EGL build; no dri requirement

Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8323>

hgl/meson: add version to libGL.so

Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8323>

targets/haiku-softpipe/meson: add libswpipe.so to install directory

Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8323>

util/meson: Add libnetwork dependency for Haiku

Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8323>

util: implement GET_PROGRAM_NAME for Haiku

Reviewed-by: Alexander von Gluck IV <kallisti5@unixzen.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8323>

nir/opt_vectorize: fix srcs_equal() with two different non-const

To match hash_alu_src(), this should return false if both are different
non-const ssa defs.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8391>

nir/opt_vectorize: fix typo in instr_can_rewrite()

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8391>

android: ac/radv: fix typo in ac_rgp.h listed in Makefile.sources

Fixes the following building error:

error: external/mesa/src/amd/Android.mk: libmesa_amd_common: Unused source files: common/ac_rgph.

Fixes: 4ec5cf5318f4 ("ac/radv: move radv_rgp.c to ac")
Signed-off-by: Mauro Rossi <issor.oruam@gmail.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8371>

gallium/u_threaded: skip draws if user index buffer size has size == 0

This happens when all draws have count == 0.

Fixes: 85b6ba136bdc2db5 "st/mesa: implement Driver.DrawGallium callbacks

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8345>

vbo: fix a index buffer map failure with size = 0 in get_minmax_indices_gallium

Fixes: 85b6ba136bdc2db5 "st/mesa: implement Driver.DrawGallium callbacks

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8345>

mesa: skip draws w/ count == 0 and instance_count == 0 in draw_gallium_fallback

Fixes: 85b6ba136bdc2db5 "st/mesa: implement Driver.DrawGallium callbacks

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8345>

gallium: skip draws with count == 0 or instance_count == 0 in drivers

Fixes: 85b6ba136bdc2db5 "st/mesa: implement Driver.DrawGallium callbacks"

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8345>

util: Avoid pthread_setaffinity_np on NetBSD

NetBSD's variant has a different prototype from the Linux version
the code expects. It might make sense to add support for NetBSD's
version, however, since NetBSD defaults to not allowing non-root
users to set processor affinity, there would be little gain here.

This is a build fix for NetBSD.

Signed-off-by: Nia Alarie <nia@NetBSD.org>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
CC: 20.3 <mesa-stable@lists.freedesktop.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7947>

clover: Add constructor for clover::module.

Fix defect reported by Coverity Scan after commit 95527fe2292
("clover/module: add a printf support to module (v5)").

Uninitialized scalar variable (UNINIT)
uninit_use_in_call: Using uninitialized value
m.printf_strings_in_buffer when calling module.

Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8322>

iris: Drop fast_clear_color's blorp_flags param

Now that conditional fast clears are disabled, the blorp_flags parameter
is unused.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7762>

iris: Delete iris_resolve_conditional_render

This function has no more users.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7762>

iris: Disable conditional fast clears

For color buffers, conditional fast clears can cause aux-state tracking
to lose information necessary for resolves later on.

For depth buffers, they never actually worked because they occurred
unconditionally. Even if they were conditional, they would suffer from
the same issues as color buffers.

Enables iris to pass the nv_conditional_render-clear-bug piglit test.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3565
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7762>

iris: Make can_fast_clear_depth return constants

Make can_fast_clear_depth more consistent with can_fast_clear_color.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7762>

iris: Explain how conditional aux accesses work

Apart from an issue with fast clears that will be addressed soon,
aux-state tracking with conditional rendering works because the
aux-state info needed for performing required resolves is never lost.

Add comments explaining how this works. Assertions are omitted to avoid
having to pass render_condition variables into
iris_resource_prepare_access and iris_resource_finish_write.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7762>

anv: Bump maxGeometryInputComponents to 128 on Gen8+

See 8e627af59d5bd416 and 1e3e72e3054de for why this may not be a good
idea on Gen7.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8397>

nir/lower_locals_to_regs: Use the imul_imm helper instead of forcing it.

Cleaned up a bit of addressing math in the shader I just had to debug.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8373>

gallium/ntt: Work around virglrenderer UIF handling bug.

Until just recently ("vrend: Fix TGSI UIF/IF behavior"), virgl does "if
(any(bvec4(src0)))" instead of "if (src0.x != 0)", despite the tgsi.rst
documentation and tgsi_exec agreeing on the second form.  It's harmless to
work around it, since apparently NTT was the only one to not have scalar
swizzled the if condition.

Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8373>

gallium/ntt: Drop comment about needing array_id for svga tess.

The svga tess stuff is about internally making a temp array, not arrayid
of the outputs.

Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8373>

gallium/ntt: Drop comment about needing loop label setup.

BRK/CONT don't take a label, as shown by tgsi_opcode_tmp.h and the lack of
any users of a label on those instructions in tree.  I can't find any user
of ENDLOOP's label.  Additionally, GLSL-to-TGSI apparently never set up
the BGNLOOP label, so even nvfx's usage probably wants us to not set it.

Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8373>

gallium/ntt: Add support for emitting TXF_LZ.

Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8373>

gallium/ntt: Emit sample index when necessary for image load/store.

Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8373>

gallium/ntt: Emit SSBO buffer declarations.

This doesn't set the atomic flag the way the old TGSI path used to,
leaving us with a possible delta in caching modes on nv50 tgsi.

Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8373>

gallium/ntt: Drop XXX comment about supporting carry opcodes.

The GLSL-to-NIR path didn't do carry, so we don't need to.

Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8373>

ci: Disable the freedreno farm, which went down last night.

We're debugging it now, but let merges flow until then.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8395>

nv50: fake enough resume support pre-nva0 to pass gles3 requirements

GLES3 supports pause/resume of xfb. However since there's no geometry
shader support in ES3, it's a lot easier to figure out the offsets to
use. This makes it work for the dEQP tests at least.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8346>

nv50: only support 4 components in separate xfb mode

This is consistent with what the blob reports.

Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8346>

virgl: fix modifier truncation

Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8372>

radv/winsys: set has_packed_math_16bit in null winsys

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8392>

microsoft/compiler: do not lower away 64-bit ffma

In DXIL, the FMA instruction only supports 64-bit operations. However,
back when we implemented support for this, there were only a single
switch for lowering all ffma instructions, so we couldn't easily use it.

But now that there's separate flags to lower ffma on 16, 32 and 64 bit,
we can lower 16 and 32 bit ffmas, and leave 64 bit ffmas alone.

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8349>

microsoft/compiler: correct dxil fma opcode

When I originally added the FFMA opcode here, I added the FMAD opcode
instead of the FMA opcode. The reason for this is that it works on
32-bit values as well, so that seemed like a better fit.

But that's not correct, as the FMA opcode isn't a fused operation, so
let's correct the opcode.

This isn't currently in use, because we currently lower away all ffma
opcodes on the NIR level, but that's about to change.

While we're at it, let's also update the opcode name to match the DXIL
documentation.

Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8349>

zink: disable render_condition_enable during blit

We don't support stencil-exports yet, and even when we will, we might
not support it on all hardware. So we really need an alternative plan
here, even when render_condition_enable is true.

Fixing this properly is much more involved, and depends on reworking
render-condition along the lines that we do in !7746 to support pausing
and resuming properly first. So let's do the minimal thing, which is to
allow this to work in cases where no render-condition is active.

Fixes: 767f70dfe13 ("gallium/util: fix util_can_blit_via_copy_region for conditional rendering")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4056
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8379>

radv: only set BO metadata for the first plane

To properly support multi-planar images, we don't want to set metadata
on anything other than the first plane. To achieve this radv currently
checks for the image TILING and assumes LINEAR means it's not the first
plane.

However this doesn't account for images with a single LINEAR plane. We
still want to set metadata on those, e.g. to properly set the scanout
bit in the tiling flags.

Instead of checking for LINEAR, check if the offset is zero. Only the
first plane has a zero offset on AMD.

This mirrors the radeonsi logic [1].

While at it, move the metadata declaration into the if block.

[1]: https://gitlab.freedesktop.org/mesa/mesa/-/blob/6fecdc6dda6da15d616a31900508214c81cd256e/src/gallium/drivers/radeonsi/si_texture.c#L710

Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8086>

zink: clamp shader input/output max values

some vulkan drivers (e.g., amdvlk) advertise absolutely huge values here,
resulting in bitmask overflows everywhere since gallium assumes a max of
32 for vertex inputs and nir uses 64bit types for others

Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8318>

radv: support SpvCapabilitySparseResidency

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7775>

radv: implement is_sparse_texels_resident and sparse_residency_code_and

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7775>

ac/nir: implement sparse image/texture loads

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7775>

ac/nir: implement nir_op_vec5

Since sparse fetch/load uses vec5 destinations, it may be possible that we
encounter nir_op_vec5.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7775>

aco: form sparse load clauses

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7775>

aco: implement sparse image loads

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7775>

aco: implement sparse texture fetches

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7775>

aco: implement nir_op_vec5

Since sparse fetch/load uses vec5 destinations, it may be possible that we
encounter nir_op_vec5.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7775>

aco: move MIMG VDATA to its own operand

We will want both a VDATA operand and a sampler for some TFE/LWE MIMG
instructions.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7775>

aco: fix MIMG_instruction::lwe comment

The ISA docs were inconsistent about what this flag does, but that seems
fixed in the RDNA doc.

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7775>

aco: fix unreachable() for uniform 8/16-bit nir_op_mov from VGPR

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Fixes: d20a752c0de ("aco: use Builder::copy more")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8380>

ci: Add *ignore_scheduled_pipelines to mesa/gallium rules templates

These are currently not used directly, but maybe this can prevent copy
and paste accidents.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8360>

ci: Rule out scheduled pipelines in .windows-build-rules

The lack of this broke scheduled pipelines, because they attempted
to create a meson-windows-vs2019 job, which couldn't work (because the
windows_build_vs2019 job doesn't exist in scheduled pipelines).

Fixes: 84c8a35aa2ca "CI: Add Windows source dependency map"
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8360>

util: Disable memstream for Apple builds

Not all SDK versions support open_memstream. Maybe some other day.

Fixes: af8d488ea5e ("util,ac,aco,radv: Cross-platform memstream API")
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8269>