Danylo Piliaiev [Thu, 5 May 2022 17:29:05 +0000 (20:29 +0300)]
tu: Do not flush ccu in clear/blits during renderpass
For clear/blits ccu flush not only worse for perf, but also messes up
flush_bits when executed in a conditional set of commands.
We already don't flush for 3d blits.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6419
Fixes:
487aa807bd1b70602fcb6fbdabd101d4cff7c07b
("tu: Rewrite flushing to use barriers")
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16352>
Samuel Pitoiset [Tue, 10 May 2022 12:02:09 +0000 (14:02 +0200)]
radv: remove useless check against nir_texop_samples_identical
nir_texop_samples_identical is lowered to
nir_texop_fragment_mask_fetch_amd earlier.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16426>
Marek Olšák [Sat, 7 May 2022 19:37:34 +0000 (15:37 -0400)]
radeonsi: report more missing/broken texture formats as unsupported on gfx6-9
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16386>
Marek Olšák [Sat, 7 May 2022 19:30:44 +0000 (15:30 -0400)]
radeonsi: report 10_10_10_2 scaled formats as unsupported on gfx6-9 as well
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16386>
Dave Airlie [Mon, 11 Apr 2022 03:42:58 +0000 (13:42 +1000)]
radv: optimised command buffer reset of vertex bindings.
This takes the buffer ptrs out of the struct, so they can be memset
separately and optimises the memset to be minimal for them.
Removing them from the struct avoids having to loop to clear them.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16393>
Dave Airlie [Mon, 9 May 2022 06:16:08 +0000 (16:16 +1000)]
radv: only dirty line stipple when necessary.
Line stipple only cares if things changes to/from line strip.
Reduces overheads seen running pixmark volplosion with zink.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16393>
Dave Airlie [Mon, 9 May 2022 06:06:58 +0000 (16:06 +1000)]
radv: cache the slab bo pointer in the pipeline.
Otherwise each emission has to pointer chase across cachelines.
Reduces overhead seen with zink on pixmark volplosion test.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16393>
Dave Airlie [Mon, 9 May 2022 05:51:45 +0000 (15:51 +1000)]
zink: don't finish barriers cmd buffer if not used.
finishing has some cpu overhead, which if the cmd buffer isn't
going to be used is pointless. just let it get reset next time.
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16393>
Dave Airlie [Mon, 9 May 2022 05:51:20 +0000 (15:51 +1000)]
zink: avoid extra submits for empty wait cmdbufs.
This just avoids cpu overheads from processing that won't get used.
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16393>
Vinson Lee [Sat, 7 May 2022 00:17:50 +0000 (17:17 -0700)]
radeonsi: Move NULL check before dereference.
Fix defect reported by Coverity Scan.
Dereference before null check (REVERSE_INULL)
check_after_deref: Null-checking desc suggests that it may be
null, but it has already been dereferenced on all paths leading to
the check.
Fixes:
2f83dce059 ("radeonsi: don't report R64_*INT as a sampler format because it doesn't work")
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16380>
Dave Airlie [Tue, 10 May 2022 01:47:33 +0000 (11:47 +1000)]
radeonsi: port tess ring calcs to the common helper.
This uses the common helper code to implement the tess ring sizing.
One question is if radeonsi should be using tess_offchip_ring_offset
in some places it's using tess_factor_ring_size?
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16415>
Dave Airlie [Tue, 10 May 2022 01:44:01 +0000 (11:44 +1000)]
ac: port radeonsi tess factor calculations to common code.
This was updated in
4e49a05e37232 ("radeonsi: increase the tesselation factor ring size")
and
9fecac091f315 ("radeonsi/gfx11: scattered register deltas")
This will apply this to radv.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16415>
Dave Airlie [Tue, 10 May 2022 01:41:54 +0000 (11:41 +1000)]
radv/ac: introduce a new common function for hs calcs.
This ports the radv code to the new ac code.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16415>
Mike Blumenkrantz [Fri, 29 Apr 2022 15:52:52 +0000 (11:52 -0400)]
zink: create images with modifiers any time there is an import handle
there's no way to pass in a plane offset otherwise, so this is necessary
in order to handle yuv imports
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16252>
Mike Blumenkrantz [Fri, 29 Apr 2022 15:02:55 +0000 (11:02 -0400)]
zink: rebind resources for export as needed
exporting all resources breaks suballocation, so instead just use the
existing heuristics and then forcibly rebind resources as needed
for this functionality
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16252>
Mike Blumenkrantz [Fri, 29 Apr 2022 15:01:49 +0000 (11:01 -0400)]
zink: add a LINEAR drm modifier if rebinding to add dmabuf export
this needs to have some kind of modifier, so be conservative here
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16252>
Mike Blumenkrantz [Fri, 29 Apr 2022 15:01:06 +0000 (11:01 -0400)]
zink: force memory exports if dmabuf bind is specified
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16252>
Mike Blumenkrantz [Fri, 29 Apr 2022 15:00:12 +0000 (11:00 -0400)]
zink: fix conditional for modifier usage
the shared check here was inaccurate; if the resource has dmabuf
export, then use modifiers
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16252>
Mike Blumenkrantz [Fri, 29 Apr 2022 14:59:31 +0000 (10:59 -0400)]
zink: add a bind flag to indicate a resource is exported as a dmabuf
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16252>
Mike Blumenkrantz [Fri, 29 Apr 2022 14:57:53 +0000 (10:57 -0400)]
zink: unset pipe_resource::next pointers when creating resource copies
this otherwise walks the chain and destroys all the resources
cc: mesa-stable
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16252>
Mike Blumenkrantz [Fri, 29 Apr 2022 14:52:27 +0000 (10:52 -0400)]
zink: always check for fb rebinds when starting renderpass
ensure the right image is being used
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16252>
Mike Blumenkrantz [Fri, 29 Apr 2022 11:42:05 +0000 (07:42 -0400)]
Revert "zink: export fd info for all 2d images"
This reverts commit
fda7371b4cfc5a15f2207f5165f3a8e14442a522.
this breaks suballocation
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16252>
Mike Blumenkrantz [Fri, 29 Apr 2022 11:41:52 +0000 (07:41 -0400)]
zink: create a copy context for the screen on init
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16252>
Mike Blumenkrantz [Fri, 29 Apr 2022 11:41:32 +0000 (07:41 -0400)]
zink: introduce copy contexts
this is for creating a context which is only used for copying images
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16252>
Mike Blumenkrantz [Fri, 29 Apr 2022 11:41:02 +0000 (07:41 -0400)]
zink: break out resource bind-adding into separate function for reuse
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16252>
Mike Blumenkrantz [Fri, 29 Apr 2022 11:40:44 +0000 (07:40 -0400)]
zink: add a flag to indicate whether a resource is exportable
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16252>
Mike Blumenkrantz [Fri, 29 Apr 2022 11:34:33 +0000 (07:34 -0400)]
zink: outdent some code
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16252>
Mike Blumenkrantz [Fri, 29 Apr 2022 11:26:09 +0000 (07:26 -0400)]
zink: remove unused headers/struct members/linkage
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16252>
Michael Skorokhodov [Tue, 10 May 2022 07:48:53 +0000 (10:48 +0300)]
glsl: Fix ir_quadop_vector validation
Some glcts tests have failed due to incorrect processing of `ir_quadop_vector` in
`ir_validation`. e.g:
`GLES31.functional.shaders.builtin_functions.integer.imulextended.int_highp_geometry`
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6461
Fixes:
23cde71b ("glsl: Stop lowering ir_quadop_vector.")
Reviewed-by: Emma Anholt <emma@anholt.net>
Signed-off-by: Mykhailo Skorokhodov <mykhailo.skorokhodov@globallogic.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16420>
Benjamin Cheng [Mon, 4 Apr 2022 04:35:19 +0000 (00:35 -0400)]
radv: fix memory leak of descriptor set layout
We need to be able to track the descriptor sets explicity to unref the
descriptor sets, otherwise these descriptor sets will not unref the
descriptor set layout it holds.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6222
Fixes:
66f7289d568
("radv: add reference counting for descriptor set layouts")
Tested-by: Jakob Bornecrantz <jakob@collabora.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15741>
Mike Blumenkrantz [Tue, 10 May 2022 19:32:20 +0000 (15:32 -0400)]
kopper: remove unused zink include
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16441>
Daniel Schürmann [Tue, 4 Jan 2022 09:59:07 +0000 (09:59 +0000)]
radv: call nir_lower_vars_to_ssa later
Totals from 599 (0.44% of 134913) affected shaders: (GFX10.3)
CodeSize: 1243732 -> 1240512 (-0.26%); split: -0.26%, +0.00%
Instrs: 223370 -> 222702 (-0.30%); split: -0.30%, +0.00%
Latency: 901762 -> 885446 (-1.81%); split: -1.81%, +0.00%
InvThroughput: 305546 -> 305536 (-0.00%)
SClause: 12690 -> 12554 (-1.07%); split: -1.13%, +0.06%
Copies: 23779 -> 23778 (-0.00%)
PreSGPRs: 17781 -> 17829 (+0.27%)
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5668>
Daniel Schürmann [Wed, 24 Jun 2020 15:22:54 +0000 (16:22 +0100)]
radv: move two NIR passes out of tight optimization loop
nir_remove_dead_variables
nir_opt_shrink_vectors
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5668>
Iván Briano [Tue, 10 May 2022 01:30:20 +0000 (18:30 -0700)]
anv: re-alloc push constants after secondary command buffers
If the secondary command buffer executed used push constants on a
different set of stages than the primary is using, we may end up not
reallocating them for the primary, getting misrender artifacts at best,
or a nice GPU hang at worst.
Fixes the tests from a CTS from the future:
dEQP-VK.dynamic_rendering.random.*
Cc: mesa-stable
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16439>
Adam Jackson [Fri, 6 May 2022 19:31:26 +0000 (15:31 -0400)]
dri: Remove what's left of utils.[ch]
Which is just moving the struct definition for __DRIconfigRec into
dri_util.h.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16377>
Adam Jackson [Fri, 6 May 2022 19:21:53 +0000 (15:21 -0400)]
dri: Move dri{Get,Index}ConfigAttrib near their vtable
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16377>
Adam Jackson [Fri, 6 May 2022 19:15:58 +0000 (15:15 -0400)]
dri: Move driC{reate,oncat}Configs near their only caller
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16377>
Adam Jackson [Fri, 6 May 2022 18:11:04 +0000 (14:11 -0400)]
dri: move driQueryRendererIntegerCommon near its one caller
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16377>
Adam Jackson [Fri, 6 May 2022 14:49:04 +0000 (10:49 -0400)]
dri: Move driCopySubBufferExtension into drisw
drisw is the only consumer so rename things to make it a bit more
obvious.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16377>
Adam Jackson [Thu, 5 May 2022 20:52:07 +0000 (16:52 -0400)]
dri: Inline dri2InvalidateDrawable into its one caller
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16377>
Charmaine Lee [Mon, 9 May 2022 06:42:41 +0000 (23:42 -0700)]
svga/nir: enable lower double option
There is no translation support for dfloor, mod, and pow
in svga_tgsi_vgpu10, just enable the lowering passes in the
nir compile option.
Reviewed-by: Neha Bhende <bhenden@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16412>
Charmaine Lee [Fri, 6 May 2022 21:45:55 +0000 (14:45 -0700)]
svga/tgsi: add translation for DTRUNC
DTRUNC is not lowered in glsl-nir-tgsi path. Add the translation
in svga_tgsi_vgpu10().
Reviewed-by: Neha Bhende <bhenden@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16412>
Charmaine Lee [Tue, 10 May 2022 16:34:57 +0000 (09:34 -0700)]
svga: enable PIPE_CAP_TGSI_TEXCOORD for vgpu10 and up
Revise commit
0d456315f95 to enable PIPE_CAP_TGSI_TEXCOORD for
vgpu10 and up only.
Reviewed-by: Neha Bhende <bhenden@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16434>
Mike Blumenkrantz [Tue, 10 May 2022 17:17:59 +0000 (13:17 -0400)]
zink: ci updates
I don't know how !16167 made this pass, but no driver seems to be able to
pass this test, so ignore
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Mike Blumenkrantz [Tue, 10 May 2022 03:37:34 +0000 (23:37 -0400)]
docs: EXT_robustness2 for lavapipe
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Mike Blumenkrantz [Tue, 26 Apr 2022 13:48:57 +0000 (09:48 -0400)]
lavapipe: ci updates
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Mike Blumenkrantz [Sat, 23 Apr 2022 15:42:34 +0000 (11:42 -0400)]
lavapipe: set robustBufferAccess2 and robustImageAccess2
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Mike Blumenkrantz [Sat, 23 Apr 2022 15:40:48 +0000 (11:40 -0400)]
gallivm: fix oob txf swizzling
this wasn't taking into account the format swizzle, returning broken
alpha values in most cases
Fixes:
0b6554ba6f2 ("gallivm,llvmpipe: handle TXF (texelFetch) instruction, including offsets")
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Erik Faye-Lund [Wed, 21 Apr 2021 09:12:35 +0000 (11:12 +0200)]
lavapipe: add support for null-descriptors from EXT_robustness2
Mike says this should be enough for full support, so let's see ;)
Nope, no robust image or buffer access implemented yet.
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Mike Blumenkrantz [Sat, 23 Apr 2022 13:39:06 +0000 (09:39 -0400)]
llvmpipe: always set ssbo data pointers for draw
skipping these would lead to reading the previously-set data if
a null buffer was set after a valid buffer
cc: mesa-stable
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Mike Blumenkrantz [Fri, 22 Apr 2022 18:47:38 +0000 (14:47 -0400)]
gallivm/draw: fix oob ubo reads
the first case in gallivm did no checking, so this was always going to be
broken, so instead just copy the now-correct ssbo logic
cc: mesa-stable
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Mike Blumenkrantz [Tue, 26 Apr 2022 14:23:36 +0000 (10:23 -0400)]
st/draw_feedback: set constant buffer stride
these shaders all do float-based reads, so using vec4 for robustness
checks breaks gallivm
cc: mesa-stable
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Mike Blumenkrantz [Fri, 22 Apr 2022 18:16:42 +0000 (14:16 -0400)]
lavapipe: fix CmdPushDescriptorSetWithTemplateKHR with refcounting
this is a cmdbuf function, which means it gets enqueued, which means
the template can't be destroyed until the cmdbuf has finished using it
cc: mesa-stable
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Mike Blumenkrantz [Fri, 22 Apr 2022 17:52:22 +0000 (13:52 -0400)]
gallivm: fix ssbo oob reads
this was checking (offset < size) when really it needs to be
(offset < size && extent < size && offset >= 0)
Fixes:
591899eedd2 ("gallivm/nir: Add a short circuit uniform-offset mode for load_ssbo/load_shared.")
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Mike Blumenkrantz [Sat, 23 Apr 2022 15:42:57 +0000 (11:42 -0400)]
lavapipe: handle null samplerviews
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Erik Faye-Lund [Wed, 14 Jul 2021 08:17:41 +0000 (10:17 +0200)]
lavapipe: tolerate NULL sampler-buffer view descriptors
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Erik Faye-Lund [Mon, 26 Apr 2021 15:11:01 +0000 (17:11 +0200)]
lavapipe: tolerate NULL image-buffer view descriptors
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Erik Faye-Lund [Mon, 26 Apr 2021 15:10:36 +0000 (17:10 +0200)]
lavapipe: tolerate NULL image view descriptors
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Erik Faye-Lund [Mon, 26 Apr 2021 14:44:27 +0000 (16:44 +0200)]
lavapipe: tolerate NULL UBO and SSBO descriptors
Reviewed-by: Dave Airlie <airlied@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16137>
Georg Lehmann [Mon, 9 May 2022 09:52:26 +0000 (11:52 +0200)]
nir/opt_algebraic: Simple xor/ishr optimizations.
The first pattern here removes the xor-swap pattern.
Foz-DB GFX10_3:
Totals from 305 (0.23% of 134913) affected shaders:
CodeSize: 1589040 -> 1585164 (-0.24%)
Instrs: 284344 -> 283375 (-0.34%)
Latency: 4205148 -> 4198472 (-0.16%); split: -0.16%, +0.00%
InvThroughput: 708745 -> 708739 (-0.00%)
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16411>
Georg Lehmann [Sat, 16 Apr 2022 11:36:17 +0000 (13:36 +0200)]
nir/opt_algebraic: Fix mask in shift by constant combining.
The comment above is correct, but the code to calculate the mask was broken.
No Foz-db changes outside of noise.
Fixes:
0e6581b87dc ("nir/algebraic: Reassociate shift-by-constant of shift-by-constant")
Signed-off-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15990>
Timur Kristóf [Fri, 1 Apr 2022 15:55:28 +0000 (17:55 +0200)]
radv/amdgpu: Use scheduled dependency for submitting to multiple queues.
This code path will be used on kernels that don't have the
"gang submit" feature in amdgpu yet.
It does each submission separately, and uses scheduled dependencies
to connect them.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-By: Tatsuyuki Ishi <ishitatsuyuki@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16361>
Timur Kristóf [Tue, 10 May 2022 10:47:46 +0000 (12:47 +0200)]
radv/amdgpu: Pass correct struct type instead of repeating the cast.
Signed-off-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/16361>
Timur Kristóf [Thu, 5 May 2022 19:51:44 +0000 (21:51 +0200)]
radv/amdgpu: Pass new queue submit info structure to internal function.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-By: Tatsuyuki Ishi <ishitatsuyuki@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16361>
Timur Kristóf [Thu, 5 May 2022 19:50:44 +0000 (21:50 +0200)]
radv/amdgpu: Initialize semaphore info with a designated initializer.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-By: Tatsuyuki Ishi <ishitatsuyuki@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16361>
Timur Kristóf [Thu, 5 May 2022 19:43:09 +0000 (21:43 +0200)]
radv/amdgpu: Group queue submit info into a structure.
This prepares RADV for submitting to multiple queues at the same time.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-By: Tatsuyuki Ishi <ishitatsuyuki@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16361>
Timur Kristóf [Wed, 23 Mar 2022 18:25:38 +0000 (19:25 +0100)]
ac/llvm: Remove now-superfluous intrinsics and ABI callbacks.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13155>
Timur Kristóf [Wed, 23 Mar 2022 18:20:47 +0000 (19:20 +0100)]
aco: Remove now-superfluous intrinsics.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13155>
Timur Kristóf [Wed, 23 Mar 2022 18:13:20 +0000 (19:13 +0100)]
ac/nir: Remove now-superfluous ac_nir_lower_tess_to_const.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13155>
Timur Kristóf [Wed, 16 Mar 2022 11:39:47 +0000 (12:39 +0100)]
radv: Lower ABI in NIR for tess/ESGS/NGG shader arguments.
Totals from 58720 (45.64% of 128653) affected shaders:
VGPRs: 2056576 -> 2056568 (-0.00%)
CodeSize:
153906488 ->
154609536 (+0.46%); split: -0.00%, +0.46%
MaxWaves: 1665056 -> 1665058 (+0.00%)
Instrs:
29242117 ->
29417265 (+0.60%); split: -0.00%, +0.60%
Latency:
134149608 ->
134238738 (+0.07%); split: -0.00%, +0.07%
InvThroughput:
20348833 ->
20349641 (+0.00%); split: -0.00%, +0.00%
VClause: 489376 -> 489385 (+0.00%)
SClause: 729383 -> 729382 (-0.00%)
Copies: 2672567 -> 2673132 (+0.02%); split: -0.00%, +0.02%
Branches: 1192117 -> 1192121 (+0.00%); split: -0.00%, +0.00%
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13155>
Timur Kristóf [Wed, 23 Mar 2022 17:44:57 +0000 (18:44 +0100)]
nir: Add upper bound for AMD shader arg intrinsics.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13155>
Timur Kristóf [Fri, 1 Oct 2021 11:19:27 +0000 (13:19 +0200)]
radv: Move radv_nggc_settings enum out of radv_cmd_buffer.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13155>
Timur Kristóf [Wed, 29 Sep 2021 16:25:03 +0000 (18:25 +0200)]
ac/nir: Add ac_nir_load_arg helper for shader arguments.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13155>
Jason Ekstrand [Fri, 15 Apr 2022 22:14:26 +0000 (17:14 -0500)]
gallium: Bump PIPE_MAX_SHADER_IMAGES to 64
This is required by OpenCL
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15988>
Jason Ekstrand [Mon, 18 Apr 2022 17:26:40 +0000 (12:26 -0500)]
mesa/st: Clamp MaxImageUniforms to MAX_IMAGE_UNIFORMS
We're about to bump PIPE_MAX_SHADER_IMAGES to 64 and don't want to
overflow any state tracker arrays for GL.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15988>
Jason Ekstrand [Mon, 18 Apr 2022 17:03:35 +0000 (12:03 -0500)]
gallium/u_threaded_context: Use PIPE_MAX_SHADER_SAMPLER_VIEWS for sampler_buffers
Sampler views and samplers may not be the same limit; in fact one is 32
while the other is 128. The sampler_buffers field is tracking sampler
views (yes, naming is confusing) so we should use the right limit.
Fixes:
e9c41b321488 ("gallium/u_threaded: add buffer lists - tracking of buffers referenced by tc")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15988>
Jason Ekstrand [Fri, 15 Apr 2022 21:35:41 +0000 (16:35 -0500)]
shader_info: Bump the number of images and textures supported
OpenCL requires up to 128 read-only images and up to 64 write images.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15988>
Jason Ekstrand [Fri, 15 Apr 2022 21:44:47 +0000 (16:44 -0500)]
util/bitset: Support larger ranges in BITSET_TEST/CLEAR_RANGE
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15988>
Jason Ekstrand [Fri, 15 Apr 2022 21:16:03 +0000 (16:16 -0500)]
nir: Gather samplers_used separately from textures
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15988>
Jason Ekstrand [Fri, 15 Apr 2022 20:32:29 +0000 (15:32 -0500)]
shader_info: Make images_used a bitset
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15988>
Jason Ekstrand [Fri, 15 Apr 2022 20:52:54 +0000 (15:52 -0500)]
nir: Stop assuming shader_info::textures_used is 32-bit
This isn't a hot path. We don't need to be manually using the
INSIDE_WORD version which will assert if we ever get a bigger texture
index.
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15988>
Jason Ekstrand [Fri, 15 Apr 2022 20:40:13 +0000 (15:40 -0500)]
nir: Set image_buffers and msaa_images in lower_samplers_as_deref
This is where we set images_used so it's less likely that things will
accidentally get out-of-sync.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15988>
Icecream95 [Fri, 10 Dec 2021 10:51:06 +0000 (23:51 +1300)]
pan/mdg: Fix multiple spilt writes in the same bundle
If two instructions in a single bundle both write to a spilt
destination, then we need to reuse the fill and spill instructions,
otherwise the value will be overwritten.
This and the rest of this set of Midgard bug fixes were found from a
vertex shader in Firefox WebRender that is used when a video is
clipped, for example by setting the border-radius CSS property.
CC: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16382>
Icecream95 [Fri, 10 Dec 2021 10:50:00 +0000 (23:50 +1300)]
pan/mdg: Return the instruction from mir_insert_instruction_*_scheduled
We can't return a pointer to the bundle itself because it might move
about in memory.
CC: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16382>
Icecream95 [Fri, 10 Dec 2021 10:46:37 +0000 (23:46 +1300)]
pan/mdg: Fix disassembly of store instructions
The mask does apply for store instructions, so pass it onto
print_vec_swizzle after converting it to the right format.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16382>
Icecream95 [Fri, 10 Dec 2021 10:40:01 +0000 (23:40 +1300)]
pan/mdg: Fix mask usage when filling before a spill
Check the bytemask against 0xFFFF rather than 0xF so that the fill is
skipped for a .xyzw write rather than a .x write.
Set the mask on the store to 0xF when doing a read so that all
components are written back.
Fixes:
31d26ebf1b9 ("pan/mdg: Fill from TLS before spilling non-SSA nodes")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16382>
Icecream95 [Fri, 10 Dec 2021 10:44:25 +0000 (23:44 +1300)]
pan/mdg: Use MAX2 to set min_alignment
If a value is written in a vector CSEL but then written again by other
instructions, it still needs full alignment, so set min_alignment
using MAX2 to avoid ever reducing it.
Fixes:
1798f6bfc33 ("pan/midgard: Fix masks/alignment for 64-bit loads")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16382>
Icecream95 [Mon, 9 May 2022 22:40:49 +0000 (10:40 +1200)]
pan/mdg: Keep min_bound at 16 when alignment requires it
Otherwise LCRA will try to divide by zero when calculating m_max.
Fixes:
553c2cf16b7 ("pan/mdg: Set RA bounds for fp16")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16382>
Pierre-Eric Pelloux-Prayer [Tue, 3 May 2022 13:34:38 +0000 (15:34 +0200)]
u_threaded: clear non-async debug callback correctly
The following sequence:
glEnable(GL_DEBUG_OUTPUT_KHR);
glEnable(GL_DEBUG_OUTPUT_SYNCHRONOUS_KHR);
glDebugMessageCallbackKHR(my_callback, NULL);
Will cause the 2nd call to be ignored - but since the callback
function used by _mesa_update_debug_callback is always the
same (_debug_message), this means we'll keep using it, causing
"my_callback" to be called from driver-internal threads.
So instead of skipping the 2nd call, make sure we pass the
information to the driver.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5206
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16300>
Martin Roukala (né Peres) [Mon, 2 May 2022 13:33:35 +0000 (16:33 +0300)]
ci: add ACO_DEBUG to the list of variables to pass down for testing
This parameter is used by radv-ci to perform extra validation while
running VKCTS.
This change catches more issues, which are will be addressed in !16248:
- dEQP-VK.subgroups.ballot_broadcast.compute.subgroupbroadcast_i8vec2,Crash
- dEQP-VK.subgroups.ballot_broadcast.compute.subgroupbroadcast_u8vec2,Crash
Signed-off-by: Martin Roukala (né Peres) <martin.roukala@mupuf.org>
Suggested-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Acked-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16277>
Pierre-Eric Pelloux-Prayer [Thu, 5 May 2022 15:02:37 +0000 (17:02 +0200)]
radeonsi/tests: reformat with black
Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16347>
Pierre-Eric Pelloux-Prayer [Thu, 5 May 2022 13:48:29 +0000 (15:48 +0200)]
radeonsi/tests: add a heuristic to pick the baseline
When the baseline for the exact GPU tested isn't available,
try to make a reasonnable guess: look for another baseline
from a GPU in the same class.
Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16347>
Pierre-Eric Pelloux-Prayer [Thu, 21 Apr 2022 14:50:39 +0000 (16:50 +0200)]
radeonsi/tests: allow empty line and comments in csv files
This will help describe why tests are failing.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16347>
Jordan Justen [Thu, 21 Apr 2022 09:48:58 +0000 (02:48 -0700)]
nir/divergence: handle more *_intel intrinsics
v2: fix topo/btd (Lionel)
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Acked-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16421>
Timur Kristóf [Thu, 5 May 2022 18:50:11 +0000 (20:50 +0200)]
radv: Only emit what is appropriate to the queue family for preambles.
This makes the function easier to read and follow.
Also prepares for some future changes when we'll want to
submit to multiple queues at once.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-By: Tatsuyuki Ishi <ishitatsuyuki@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16358>
Timur Kristóf [Thu, 5 May 2022 18:40:13 +0000 (20:40 +0200)]
radv: Don't use pointers to pointers when updating the preambles.
Instead, just use the pointers from the queue structure.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-By: Tatsuyuki Ishi <ishitatsuyuki@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16358>
Timur Kristóf [Thu, 5 May 2022 18:16:30 +0000 (20:16 +0200)]
radv: Rename fill_geom_tess_rings to radv_fill_shader_rings.
This function already handles more than just the geometry and tess
rings, and it will include more things in the future, such as
the task shader rings.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-By: Tatsuyuki Ishi <ishitatsuyuki@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16358>
Timur Kristóf [Thu, 5 May 2022 18:09:27 +0000 (20:09 +0200)]
radv: Don't create continue preamble when it's not needed.
Previously we would always create it, but would return NULL when
it wasn't needed. Now, don't create it when not needed.
Additionally, don't create the continue preamble when we can use
IB BOs, because then we never use a continue preamble.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-By: Tatsuyuki Ishi <ishitatsuyuki@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16358>
Timur Kristóf [Thu, 5 May 2022 18:04:46 +0000 (20:04 +0200)]
radv: Refactor cache flush code for the initial preambles.
The "initial" and "initial full flush" preambles both need a
cache flush with a slight difference. Improve the readability of
this code.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-By: Tatsuyuki Ishi <ishitatsuyuki@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16358>
Timur Kristóf [Thu, 5 May 2022 17:17:42 +0000 (19:17 +0200)]
radv: Initialize BO pointers when creating preambles.
This aims to improve the readability of this function.
Initialize the BO pointers to the queue's current pointers
at the beginning instead of in the else branches later.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Reviewed-By: Tatsuyuki Ishi <ishitatsuyuki@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16358>