review.tizen.org Git - platform/upstream/mesa.git/log

zink: fix implicit feedback loop detection

the code here was all expecting the VkPipelineStageFlags bitfield expansions,
but u_foreach_bit() gives the actual bit, so implicit feedback loops were never
actually being detected

instead, convert back to the bitfield at the top of the loop so the value works
as expected

Fixes: 9ba06579035 ("zink: make implicit feedback loop application stricter")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20863>

radv: print depth image size with RADV_DEBUG=img

This turned out to be useful when investigating a GPU hang.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20872>

intel/common: add a INTEL_DECODE variable to parameter decoder at runtime

Sometimes you want to diff 2 runs with INTEL_DEBUG=bat, but a tiny
allocation change can mess quite badly with offsets printed in the
decoding, making it hard to look at the diff with meld.

Fortunately our decoder can avoid printing offsets. We just need a
variable to specify that.

We still use the defaults specified by the driver but you can turn
things on/off with :

INTEL_DECODE=+color,-offsets,-floats INTEL_DEBUG=bat ./my_app

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20874>

intel/decoder: print out compute push constants

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20874>

intel/compiler/mesh: handle const data in task & mesh programs

Started showing up when nir_opt_large_constants call was moved in 88756cee8da.
Fixes dEQP-VK.mesh_shader.ext.smoke.monolithic.fullscreen_gradient*

Suggested-by: Kenneth Graunke <kenneth@whitecape.org>
Fixes: 88756cee8da ("intel/compiler: Run nir_opt_large_constants before scalarizing consts")
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20876>

intel/compiler: fix generation of vec8/vec16 alu instruction

I stumbled on this when I inserted some suboptimal lowering code after all
optimizations. Adding certain subset of optimizations after my lowering code
actually avoided this bug, so I think it's not possible to hit this on upstream.

Let's fix this for the next person generating suboptimal code...

Reviewed-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20857>

loader,glx,egl: remove is_different_gpu variable from loader

v2: fd number is different (Pierre-Eric)
v1: remove is_different_gpu (Pierre-Eric)

Signed-off-by: Yogesh Mohan Marimuthu <yogesh.mohanmarimuthu@amd.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13422>

glx: remove is_different_gpu variable from struct dri_screen

v2: fd number is different (Pierre-Eric)
v1: remove is_different_gpu (Pierre-Eric)

Signed-off-by: Yogesh Mohan Marimuthu <yogesh.mohanmarimuthu@amd.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13422>

egl: remove is_different_gpu variable from struct dri2_egl_display

v2: fd number is different (Pierre-Eric)
v1: remove is_different_gpu (Pierre-Eric)

Signed-off-by: Yogesh Mohan Marimuthu <yogesh.mohanmarimuthu@amd.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13422>

loader,glx,egl,vl,d3d: loader_get_user_preferred_fd() function to return original_fd

v1: return original_fd in loader_get_user_preferred_fd() (Pierre-Eric)
v2: fix *original_id crash for android, haiku... (Pierre-Eric)
remove extra comment (Pierre-Eric)
v2: also return render fd in case of original_fd passed is NULL (Pierre-Eric)

Signed-off-by: Yogesh Mohan Marimuthu <yogesh.mohanmarimuthu@amd.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13422>

loader,glx: add render_gpu tag psc->driScreen and psc->fd

v1: add render_gpu_tag (Pierre-Eric)

Signed-off-by: Yogesh Mohan Marimuthu <yogesh.mohanmarimuthu@amd.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13422>

egl: add render_gpu tag to dri2_dpy->fd and dri2_dpy->dri_screen variable

v1: add render_gpu_tag (Pierre-Eric)

Signed-off-by: Yogesh Mohan Marimuthu <yogesh.mohanmarimuthu@amd.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13422>

egl/wayland: for prime, allocate linear_copy from display GPU VRAM

Allocates VRAM in display GPU in case of prime. Then the dma_buf is imported
into prime GPU.

v4: add image tag to __DRIimage (Marek Olšák)
v3: move display fd opening to separate commit (Pierre-Eric)
image_format_to_fourcc() non-static to seperate commit (Pierre-Eric)
v2: close query fds after linear_copy buffer import (Marek Olšák)
use image_format_to_fourcc() from loader_dri3_helper.c (Marek Olšák)

Signed-off-by: Yogesh Mohanmarimuthu <yogesh.mohanmarimuthu@amd.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13422>

loader: make image_format_to_fourcc() non-static

the image_format_to_fourcc() function will be used from
egl/wayland hence make it non-static. Also move the function
into loader_dri_helper.c from loader_dri3_helper.c since
loader_dri3_helper library depends on xcb which will make
egl wayland depend on xcb indirectly.

v2: add loader tag to extern image_format_to_fourcc() (Marek Olšák)

V3: move image_format_to_fourcc to loader_dri_helper.c
Signed-off-by: Yogesh Mohanmarimuthu <yogesh.mohanmarimuthu@amd.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13422>

egl/wayland: keep display fd open for prime

Keep the display fd open for creating DRI screen on display gpu in
case of prime.

Signed-off-by: Yogesh Mohanmarimuthu <yogesh.mohanmarimuthu@amd.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13422>

loader,glx,egl/x11: init dri_screen_display_gpu in struct loader_dri3_drawable

Initialize dri_screendisplay_gpu variable in struct laoder_dri3_drawable.
Also make dri_screen_display_gpu variable as input parameter to function
loader_dri3_drawable_init() since dri_screen variable is initialized this way.
This also helps to avoid duplicate initializing dri_screen_display_gpu
in glx and egl code.

Signed-off-by: Yogesh Mohanmarimuthu <yogesh.mohanmarimuthu@amd.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13422>

egl: create DRI screen for display GPU in case of prime

The created DRI screen can be used to allocate VRAM memory from
display GPU in case of prime.

v2: remove extra whitespace (Marek Olšák)

Signed-off-by: Yogesh Mohanmarimuthu <yogesh.mohanmarimuthu@amd.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13422>

egl,egl/x11: keep display fd open for prime

Keep the display fd open for creating DRI screen on display gpu in
case of prime. Also close the fd opened in dri2_display_destroy()

Signed-off-by: Yogesh Mohanmarimuthu <yogesh.mohanmarimuthu@amd.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13422>

egl: add fd_display_gpu to struct dri2_egl_display

fd opened on display gpu is saved in fd_display_gpu. It is later used
to create dri screen on display gpu.

Signed-off-by: Yogesh Mohanmarimuthu <yogesh.mohanmarimuthu@amd.com>
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13422>

intel/genxml: set unused 3DSTATE_PS_EXTRA field as mbz

Wa_14015360517 mentions situations where HW produces invalid
occlusion query results when "Pixel Shader Does not write to RT"
bit is set.

"When Pixel Shader Kills Pixel is set, SW must perform a dummy render
target write from the shader and not set this bit, so that Occlusion
Query is correct."

Another situation is when writing to UAV or to NULL render target.
Patch sets field as 'must be zero' to discourage possible use of it.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20849>

mesa: move component bits queries as GL ES only

These enums have been removed/deprecated long time ago from desktop
GL. Here we remove them from modern GL while still allow for compat
contexts (~old apps). This change matches proprietary drivers and makes
our behaviour same within CTS with some tests.

Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20719>

v3dv: ensure we allocate at least the requested space for a CL

While we are already ensuring we allocate at least 8192 bytes should
this not be the first allocation and our allocations are typically just
a few bytes, multilayered framebuffers with large numbers of layers may
require more space than that in a single allocation.

Fixes: 3325950648 ('v3dv: increase BO allocation size when growing CLs')
Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20871>

mesa: (more) correctly handle incomplete depth textures

according to GL spec, incomplete shadow samplers should return 0

this is technically possible for drivers to do using a RGBA texture in
the sense that somehow it's been working, but it's broken at the gallium-level
for what drivers should be expecting to see in such circumstances given
that such scenarios have been binding a RGBA texture to use with shadow samplers

instead, we can give drivers a fallback Z32 texture to avoid format/sampler
mismatches and complying with expected behavior

see also KHR-GL46.incomplete_texture_access.sampler for driver-specific testing

Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20817>

mesa: populate gl_program::ShadowSamplers mask from shader data

this data is already processed, it just isn't being propagated for whatever reason

Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20817>

mesa: remove dead parameter doc for _mesa_new_texture_object()

Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20817>

zink: pass depth swizzle data block to shader compile

with everything now hooked up, this should fix all related test failures

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20598>

zink: remove old depth swizzle workaround

this is already handled in match_tex_dests(), so it does nothing here

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20598>

zink: create another samplerview for shadow textures

when doing legacy depth texture mode sampling, it's necessary to keep
another view that has the right (R in component 0) swizzle so that depth
values can actually be returned in cases where it would otherwise be
a constant value due to swizzling

this also allows zink_sampler_view::shadow_needs_shader_swizzle to be removed

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20598>

zink: plug in the program/module parts of shadow texture mode emulation

this is clunky because of how big the swizzle data block is,
but the gist of it is the data block is stored onto the shader module key
after all the other data, and then it gets manually hashed/compared in
relevant cases

it's gross, but so is this functionality

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20598>

zink: block pipeline fast-pathing for any programs using depth texture modes

the data for this is too big to compress into a shader key, so these pipelines
will always consume more cpu

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20598>

zink: rework depth sampler splatting in shaders

this enables passing a zink_fs_shadow_key to the compiler to manually
apply a swizzle other than R/R/R/R to depth texture results

currently no data is passed, so the previous splatting behavior is preserved

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20598>

zink: add a fs shader key member to indicate depth texturing mode

this does nothing now besides track the data

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20598>

zink: track depth swizzle on samplerviews

this will provide info for shader rewrites

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20598>

zink: add an extra_data param to zink_shader_compile

this is extra shader key data that can be used in various ways per stage
and is too large to fit into the shader key

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20598>

zink: break out tex dest rewriting into separate function

no functional changes

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20598>

zink: flag old-style shadow tex mask for fragment shaders

this will be useful for handling depth texturing modes

only 32 are tracked now for performance reasons

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20598>

radv/winsys: fix incorrect PCIID for GFX11 in the null winsys

Fixes: bbad550f3d4 ("radv/winsys: fill real info for CHIP_GFX1100")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20850>

intel/fs: avoid cmod optimization on instruction with different write_mask

I've been running into failures with tests like :

dEQP-VK.robustness.robustness2.bind.notemplate.rgba32i.unroll.nonvolatile.uniform_buffer_dynamic.no_fmt_qual.len_4.samples_1.1d.frag

With the load_global_const_block_intel NIR intrinsic, you can load a
vec8/vec16 with a predicate. The predicate is correctly uniformized to
feed into the SEND instruction's flag register.

The problem is that a series of optimization first remove the
find_live_channel and then changes the broadcast into a simple MOV
instruction, on the assumption that the first channel is always active
if there is not control flow. This is correct.

But after that the cmod optimzation will remove this instruction :

mov.nz.f0.0(16) null:D, vgrf16+0.0<0>:D NoMask

because it seems to be equivalent to :

cmp.g.f0.0(16) vgrf16:D, vgrf12:D, 63d

In this case vgrf16 is the predicate to the load block SEND
instruction. Since the execution mask is different between both, some
of the channels of the SEND instruction end up not being loaded or
loaded with the wrong predication and we end up with incorrect UBO
data.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: mesa-stable
Reviewed-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20852>

ci/piglit: Exclude swapbuffers front-readback tests with PIGLIT_PLATFORM=gbm.

These are expected to fail by the design of gbm. Don't make each driver
track them.

Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Karol Herbst <kherbst@redhat.com>
Acked-by: Martin Roukala <martin.roukala@mupuf.org>
Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20798>

ci/piglit: Add some common piglit skips for Mesa CI's testing of glx.

Since our X servers don't have a compositor, and we run tests in parallel,
various swap and frontbuffer tests won't ever be stable. Rather than
having every driver have to track those flakes, make a general X11 skips
list as a known issue of our CI rather than pointing fingers at drivers.

Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Karol Herbst <kherbst@redhat.com>
Acked-by: Martin Roukala <martin.roukala@mupuf.org>
Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20798>

ci: Move PIGLIT_PLATFORM settings out of the .tomls.

I'm going to add some automatic platform-based skips lists shortly (like
all-skips but more targeted), and this avoids needing to add them to each
.toml.

Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Karol Herbst <kherbst@redhat.com>
Acked-by: Martin Roukala <martin.roukala@mupuf.org>
Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Acked-by: Erico Nunes <nunes.erico@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20798>

ci/freedreno: Switch the piglit job to using a deqp-runner suite.

This is one of the few remaining piglit-runner.sh users.

I think the notable change here is that we no longer set
EGL_PLATFORM=surfaceless like the piglit-runner.sh script did.

Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Karol Herbst <kherbst@redhat.com>
Acked-by: Martin Roukala <martin.roukala@mupuf.org>
Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20798>

ci/zink: Clear issue #7781 flakes.

These tests all had valgrind UAF complaints that got fixed with the MR
closing that bug.

Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Karol Herbst <kherbst@redhat.com>
Acked-by: Martin Roukala <martin.roukala@mupuf.org>
Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20798>

ci/zink: Drop glx-swap-copy xfails.

These now skip, since there are no preserved configs any more.

Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Karol Herbst <kherbst@redhat.com>
Acked-by: Martin Roukala <martin.roukala@mupuf.org>
Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20798>

ci/zink: Drop xfail for copy-sub-buffer.

This is a skip these days, since the ext isn't supported in the kopper
world.

Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Karol Herbst <kherbst@redhat.com>
Acked-by: Martin Roukala <martin.roukala@mupuf.org>
Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20798>

ci: Drop windowoverlap xfails, since it's always skipped.

e31d08d307d0aa2b2c7441e85f12c7e5deee4bc9 made it a skip, so no need to
track these any more.

Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Acked-by: Karol Herbst <kherbst@redhat.com>
Acked-by: Martin Roukala <martin.roukala@mupuf.org>
Acked-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20798>

ci/virgl: Disable iris traces for now while it's unstable.

4 spurious fails today it looks like with the same
"[drm:virtio_gpu_get_capsets] *ERROR* timed out waiting for cap set 0"
signature. It had started by at least Jan 20.

https://gitlab.freedesktop.org/mesa/mesa/-/jobs/35240569#L2929 (today)

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20868>

radv: stop using radv_pipeline_has_stage() in BindPipeline

this incurs a small amount of unnecessary cpu overhead as compared to just
checking bitflags

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20615>

hasvk: check the return value of anv_execbuf_add_bo_bitset()

This is the Hasvk version of Anv's:
3d37950fd9dd ("anv: check the return value of anv_execbuf_add_bo_bitset()")

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20800>

hasvk: don't leave undefined values in exec->syncobj_values

This is the Hasvk version of Anv's:
ad6a036a6815 ("anv: don't leave undefined values in exec->syncobj_values")

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20800>

ci: Uprev kernel to 6.1.7

Once prepared the archive in gfx-ci/linux for kernel 6.1 pointing to 6.1.7 and
with the patches for mesa, update the link. Also, enable some kconfigs needed
for the Adreno and USB.

One job in AMD required an expectation file to be updated with one test. Also,
an mt8192 device tree has been included in the arm64 trees for lava build.

Co-developed-by: Daniel Stone <daniels@collabora.com>
Co-developed-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Co-developed-by: Helen Koike <helen.koike@collabora.com>
Co-developed-by: David Heidelberg <david.heidelberg@collabora.com>
Signed-off-by: Sergi Blanch Torne <sergi.blanch.torne@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20855>

util/fossilize_db: fix macOS inotify build error

require <sys/inotify.h> for fossilize_db

Fixes: 3b69b67545b ("util/fossilize_db: add runtime RO foz db loading via FOZ_DBS_DYNAMIC_LIST")

Signed-off-by: Juston Li <justonli@google.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20755>

radeonsi/vcn: check fence before destroying decoder

Add a fence_wait before destroying decoder to make sure destory message has
been processed by firmware.

Signed-off-by: Boyuan Zhang <boyuan.zhang@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Sajeesh Sidharthan <sajeesh.sidharthan@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20805>

radeonsi/vcn: check fence before destroying dpb

Add a fence_wait before destroying dpb buffer in dpg_unref_list to make sure
previous decode job has been done. (for DPB_DYNAMIC_TIER_2)

Signed-off-by: Boyuan Zhang <boyuan.zhang@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Sajeesh Sidharthan <sajeesh.sidharthan@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20805>

Revert "aco: Combine v_cvt_u32_f32 with insert to v_cvt_pk_u8_f32."

This reverts commit 6d020540474b43c7e9a3175638bb04be68d80dbb.

v_cvt_pk_u8_f32 returns 0xff instead of v_cvt_u32_f32 & 0xff if the input is
larger than 255.

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8128

Cc: mesa-stable
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20829>

zink: update gl43 profile to allow imageRobustAccess

Now that zink can use imageRobustAccess and handle txf with invalid LODs,
update the profile to require either VK_EXT_image_robustness or VK_EXT_robustness2.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20808>

zink: lower LOD-invalid txf when imageRobustAccess2 is missing

GL robust buffer access applies to texelFetch with out of bounds LODs.
imageRobustAccess2 guarantees this, but imageRobustAccess does not.
Therefore, the txf robustness lowering pass from earlier is used
to provide this guarantee and support ARB/KHR robust_buffer_access_behavior.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20808>

zink: add robust_access field to shader key

Since shaders can be shared even between robust and non-robust
contexts, the robustness state needs to be tracked.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20808>

zink: add VK_EXT_image_robustness

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20808>

zink: add zink_cs_key

Also, add shader key handling to compute programs.
This will be used later on for the rb_image workaround.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20808>

zink: add pass checking for lod overflow in txf

Will be used later to workaround missing robustImageAccess2 along
with robustImageAccess.

Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20808>

zink: use actual swapchain object for surface comparison

the outer swapchain object is persistent, which means checking it
will never yield an update after the first check

fixes #8122

Fixes: b2739c9f005 ("zink: set surface->dt when updating swapchain"
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20814>

aco/tests: update assembler tests for latest LLVM 16

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20747>

aco/tests: fix assembler.gfx11.vop12c_v128 with LLVM 15

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8089
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20747>

radeonsi: respect smoothing_enabled

When this was last changed, the smoothing_enabled flag seems to have
been forgotten about, breaking line-smoothing (and probably also polygon
smoothing).

Fixes: 4147add280e ("radeonsi: update db_eqaa even if msaa is disabled")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20810>

v3dv: increase BO allocation size when growing CLs

This can significantly reduce the number of allocations required
to record large command buffers with thousands of draw calls and
it improves significantly vkoverhead's results.

fps_avg helped:   serious_sam_trace01_1280x720.gfxr:                17.96 -> 18.50 (3.01%)
fps_avg helped:   vehicle-game-2.gfxr:                              9.09 -> 9.33 (2.67%)
fps_avg helped:   ue4_shooter_game_high_quality.gfxr:               17.40 -> 17.66 (1.48%)
fps_avg helped:   ue4_shooter_game_low_quality.gfxr:                25.25 -> 25.61 (1.44%)
fps_avg helped:   ue4_sun_temple.gfxr:                              24.93 -> 25.15 (0.86%)
fps_avg helped:   vkQuake_capture_frames_1_through_1200.gfxr:       52.54 -> 52.95 (0.78%)
fps_avg helped:   rbdoom-3-bfg_capture.gfxr:                        13.87 -> 13.96 (0.69%)
fps_avg helped:   sponza_demo02_800x600.gfxr:                       19.59 -> 19.63 (0.21%)

fps_avg HURT:   ue4_shooter_game_shooting_high_quality.gfxr:      19.14 -> 18.95 (-0.98%)
fps_avg HURT:   sponza_demo01_800x600.gfxr:                       20.03 -> 20 (-0.09%)
fps_avg HURT:   quake3e_capture_frames_1_through_1800.gfxr:       57.34 -> 57.34 (-0.01%)
fps_avg HURT:   serious_sam_trace02_1280x720.gfxr:                38.27 -> 38.27 (-0.01%)

total fps_min in shared programs: 311.74 -> 313.91 (0.69%)
fps_min in affected programs: 311.74 -> 313.91 (0.69%)
helped: 10
HURT: 2

total fps_max in shared programs: 319.54 -> 320.24 (0.22%)
fps_max in affected programs: 319.54 -> 320.24 (0.22%)
helped: 6
HURT: 6

total fps_avg in shared programs: 315.41 -> 317.36 (0.62%)
fps_avg in affected programs: 315.41 -> 317.36 (0.62%)
helped: 8
HURT: 4

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20787>

v3dv: drop unused field from v3dv_cmd_buffer

Not needed since we moved to the common sync framework.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20787>

egl/dri2: avoid undefined unlocks

unlocks were incorrectly added to paths using dri2_egl_display() as
well as those using dri2_egl_display_lock()

pthread_mutex_unlock() when unlocked is documented by posix as
being undefined behaviour. On OpenBSD pthread_mutex_unlock() will call
abort(3) if this happens.

Fixes: f1efe037dfd ("egl/dri2: Add display lock")
Reviewed-by: Rob Clark <robclark@freedesktop.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20712>

ci: bump Mold to the 1.10.0

mold 1.9.0 was up to 10% slower than 1.8.0 on some multicore machines. We fixed the performance regression and made it even faster than 1.8.0.

Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20819>

ci: Sir trace has small invisible change in rendering

Unrelated to MR itself.

Signed-off-by: David Heidelberg <david.heidelberg@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20819>

radv: add an early out in radv_cmd_buffer_flush_dynamic_state()

no point checking all the states if they're known to be unset

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20653>

v3dv: use common code for descriptor update template

Use the common code and structs for this. Depends on !14780

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18625>

glx: fix a macro being invoked with the wrong parameter name

Strangely, this was somehow compiling with GCC but my futile efforts to
build mesa with msan caused me to find clang refusing to compile because
of this. Unknown how many bugs this could fix or how GCC did manage to
find "config" in scope but it's fairly obvious that this is the correct
parameter that should be used.

Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20830>

radv: Enable extended SAH for shallow BVHs

Reviewed-by: Friedrich Vock <friedrich.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20656>

radv: Wrap internal build type inside a build_config struct

This will be useful for finer control over build configurations.

Reviewed-by: Friedrich Vock <friedrich.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20656>

radv: Add a shader variant for PLOC with extended SAH

Reviewed-by: Friedrich Vock <friedrich.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20656>

radv/bvh: Add a define for extended SAH

This will be used to only chose depth aware SAH when we know that it's
more optimal and doesn't increase build overhead too much.

Reviewed-by: Friedrich Vock <friedrich.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20656>

radv/bvh/meson: Add the option to set defines

This is useful for compiling different variants of the same shader.

Reviewed-by: Friedrich Vock <friedrich.vock@gmx.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20656>

ci: be explicit about the `meson setup` subcommand

The old way of doing things is deprecated.

Signed-off-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20827>

r600/sfn: Fix splitting of multislot alu ops

The old source must not incorporate the start slot as lookup
index.

Fixes: commit 573448f93cbbeeae22295e7f943c7120598780d4
r600/sfn: prepare alu split for use of variable length dot

Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8120

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20826>

r600/sfn: Fix opcode and result dest slot mask for variable size dot

Fixes: commit 2df023a1f1990aad6c20eca85af19c7d21a43203
r600/sfn: pre-evaluate allowed dest mask in Alu instructions

Related: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8120

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20826>

glthread: disallow glthread if buffer uploads are unsupported

to remove fail paths that defeat the purpose of glthread.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824>

glthread: do vertex uploads if an index buffer is present for MultiDrawElements

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824>

glthread: remove the vbo_upload_ratio_too_large fallback for glMultiDrawElements

This would be possible to implement, but we don't know of any app where
it would help.

Now glthread fully handles all non-VBO uploads except glDrawIndirect.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824>

glthread: make marshal functions for glBegin/End attribs non-static

for a future commit

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824>

glthread: pack and name the type of glthread_vao::Attrib

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824>

mesa: move gl_vertex_format_user definition into glthread.h

glthread.h needs it and it can't include mtypes.h because mtypes.h
includes it.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824>

glthread: do vertex uploads if an index buffer is present for glDrawElements

glthread didn't implement uploading non-VBO vertices if indices were
in a buffer. This implements that.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824>

glthread: change multi_draw_elements_async() to never fail due to large size

Some callers (not visible here) silently ignore the return value. Remove
the return value and handle the failure in multi_draw_elements_async.

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824>

glthread: execute glMultiDrawArrays(draw_count < 0) asynchronously

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824>

glthread: set GL_OUT_OF_MEMORY if we fail to upload vertices

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824>

glthread: set GL_OUT_OF_MEMORY if we fail to upload indices

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824>

glthread: handle GL_*_ARRAY in glEnable/Disable

Surprisingly, the GL compatibility profile allows these in both
glEnableClientState and glEnable.

Fixes: 0b1dd185913 - glthread: track which vertex array attribs are enabled

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824>

glapi: autogenerate function parameters with no space between * and variable

print "int *v" instead of "int * v".

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824>

mesa: allow GL_UNSIGNED_INT64_ARB as vertex format for ARB_bindless_texture

This wasn't implemented, but the spec requires it.

Fixes: 1fe7b1f9724 - mesa: implement ARB_bindless_texture

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824>

util: fix util_is_vbo_upload_ratio_too_large

It was wrong. For example, if the draw vertex count was 10 and the upload
vertex count was 150, u_vbuf wouldn't unroll the draw and would instead
memcpy 150 vertices. This fixes that case.

Fixes: 068a3bf0d7c - util: move and adjust the vertex upload heuristic equation from u_vbuf

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20824>

glthread: fix an upload buffer leak

Fixes: befbd54864d29 - glthread: don't use atomics for refcounting to decrease overhead on AMD Zen

Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20804>

anv: fix preemption enable emission in gpu_memcpy

This has to be before the MI_BATCH_BUFFER_END otherwise it has no
effect.

This also was messing around with you batch length alignment.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: b9aa66d5d0 ("anv: disable preemption for 3DPRIMITIVE during streamout")
Reviewed-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20802>