Qiang Yu [Tue, 11 Oct 2022 08:29:44 +0000 (16:29 +0800)]
ac/nir/ngg,ac/llvm,aco: save nogs ngg culling one lds dword
TES rel patch id is <256, so we can use an existing unused LDS
byte instead of extra dword.
To ease the programing, change the index of repacked_arg_vars
for these variables.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18832>
Qiang Yu [Sun, 31 Jul 2022 09:53:19 +0000 (17:53 +0800)]
ac/nir/ngg: save and restore no_varying/no_sysval_output
These are used by radeonsi for param export count, should
be saved and restore.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18832>
Qiang Yu [Sat, 23 Jul 2022 09:30:52 +0000 (17:30 +0800)]
ac/nir/ngg: allow passthrough with vs primitive id output
vertex primtive id and passthrough are not exclusive, just need
to get correct vertex index when passthrough.
radeonsi won't disable passthrough when vs primitive id output,
this is also for fixing the crash of the assertion.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18832>
Qiang Yu [Fri, 22 Jul 2022 09:42:41 +0000 (17:42 +0800)]
ac/nir/ngg,radv: move LDS layout calculation out of nir ngg lowering
Use lds base load intrinsics in nir ngg lowering to get layout, left
its calulation to driver.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18832>
Qiang Yu [Fri, 22 Jul 2022 07:25:45 +0000 (15:25 +0800)]
nir: add two amd ngg lds base load intrinsics
These two values are not known when compile for radeonsi.
They are relocated when link/upload time.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18832>
Qiang Yu [Thu, 14 Jul 2022 11:56:59 +0000 (19:56 +0800)]
ac/nir/ngg: pass primitive_id_location as param for nogs lower
radeonsi need to use packed driver location for all outputs,
while radv need to use VARYING_SLOT_*. To meet both drivers'
needs.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18832>
Qiang Yu [Fri, 10 Jun 2022 12:54:43 +0000 (20:54 +0800)]
ac/nir/ngg: support user edge flags for ngg lower
Pack user edge flag into arg code is ported from radeonsi
gfx10_ngg_build_export_prim.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18832>
Qiang Yu [Tue, 7 Jun 2022 07:46:36 +0000 (15:46 +0800)]
ac/llvm: get back intrinsics used by NGG
Will be used by radeonsi.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Qiang Yu <yuq825@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18832>
Brian Paul [Wed, 26 Oct 2022 21:39:39 +0000 (15:39 -0600)]
glx: clean-ups in drisw_glx.c
Replace tabs with spaces. Fix up function pointer calls (don't use
the old style (*foo)(arg) syntax).
Signed-off-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19329>
Brian Paul [Wed, 26 Oct 2022 21:37:31 +0000 (15:37 -0600)]
glx: clean-ups in create_context.c
Replace tabs w/ spaces, remove trailing whitespace.
Signed-off-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19329>
Brian Paul [Wed, 26 Oct 2022 21:36:15 +0000 (15:36 -0600)]
frontends/dri: clean-ups in dri_util.c
Replace tabs with spaces. Rename __ATTRIB macro to SIMPLE_CASE to
be a bit more readable.
NFC.
Signed-off-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19329>
Brian Paul [Wed, 26 Oct 2022 21:34:38 +0000 (15:34 -0600)]
frontend/dri: assorted clean-ups in dri-screen.c
Replace tabs with spaces, fix indentation.
Move 'format' var decl and type (it's an integer array index, not
actually a mesa format).
NFC.
Signed-off-by: Brian Paul <brianp@vmware.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19329>
Yusuf Khan [Sun, 9 Oct 2022 00:39:01 +0000 (19:39 -0500)]
nv50/ir: nir_op_b2i8 and nir_op_b2i16
Signed-off-by: Yusuf Khan <yusisamerican@gmail.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19256>
Yiwei Zhang [Mon, 24 Oct 2022 23:45:50 +0000 (23:45 +0000)]
docs: update to latest venus driver support
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19285>
Yiwei Zhang [Tue, 25 Oct 2022 05:37:42 +0000 (05:37 +0000)]
venus: add VK_EXT_depth_clip_control support
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19285>
Yiwei Zhang [Mon, 24 Oct 2022 23:38:38 +0000 (23:38 +0000)]
venus: add VK_EXT_primitives_generated_query support
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19285>
Yiwei Zhang [Mon, 24 Oct 2022 22:58:52 +0000 (22:58 +0000)]
venus: sync to latest venus protocol headers
This brings in:
- VK_KHR_push_descriptor
- VK_EXT_depth_clip_control
- VK_EXT_primitives_generated_query
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19285>
Yiwei Zhang [Tue, 25 Oct 2022 01:18:51 +0000 (01:18 +0000)]
venus: handle VkAndroidHardwareBufferFormatProperties2ANDROID
Fixes:
4d80ccbf2d0 ("venus: Enable VK_KHR_format_feature_flags2")
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19287>
Yiwei Zhang [Mon, 24 Oct 2022 23:01:22 +0000 (23:01 +0000)]
venus: remove redundant codes
This is some left over from prior 1.3 effort.
Signed-off-by: Yiwei Zhang <zzyiwei@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19287>
Dave Airlie [Tue, 25 Oct 2022 03:43:56 +0000 (13:43 +1000)]
nir/lower_bool_to_int32: add support for lowering functions.
Change the function parameters to 32-bit.
Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19291>
Lionel Landwerlin [Wed, 26 Oct 2022 10:54:57 +0000 (13:54 +0300)]
nir/divergence_analysis: add missing desc_set_address_intel
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: mesa-stable
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19320>
Lionel Landwerlin [Wed, 26 Oct 2022 10:54:26 +0000 (13:54 +0300)]
nir/divergence_analysis: add some missing RT intrinsics
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: mesa-stable
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19320>
Lionel Landwerlin [Fri, 21 Oct 2022 08:37:32 +0000 (11:37 +0300)]
vulkan/wsi/wl: correctly find whether the compositor uses the same GPU
Using the wl_drm protocol we can check whether the compositor uses the
same GPU as the application.
This allows to run vulkan applications using a DG2 GPU with the
compositor using another card.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: mesa-stable
Reviewed-by: Simon Ser <contact@emersion.fr>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19224>
Lionel Landwerlin [Fri, 21 Oct 2022 09:25:18 +0000 (12:25 +0300)]
anv: init major/minor before WSI
So that we can provide that information to WSI if it asks for it
immediately.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: mesa-stable
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19224>
Lionel Landwerlin [Wed, 26 Oct 2022 14:40:34 +0000 (17:40 +0300)]
anv: disable mesh in memcpy
We can't have streamout and mesh enabled at the same time.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes:
ef04caea9b8b ("anv: Implement Mesh Shading pipeline")
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19323>
Christophe [Wed, 26 Oct 2022 14:19:10 +0000 (16:19 +0200)]
Zink: add Zink profiles file
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19192>
Christophe [Mon, 24 Oct 2022 16:28:09 +0000 (18:28 +0200)]
zink: add profile documentation
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19192>
Mike Blumenkrantz [Wed, 30 Jun 2021 18:35:02 +0000 (14:35 -0400)]
zink: handle broken resource mapping deadlocks
some apps (most notably Wolfenstein: The New Order) have broken multi-context
buffer usage in which one context will attempt to write to a buffer while
another context holds unflushed usage, and the unflushed context will never
flush until the buffer write completes
it's impossible to handle this scenario correctly without deadlocking,
so add some handling to try waiting and then yolo the buffer write if
a deadlock would occur
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19141>
Jason Ekstrand [Tue, 25 Oct 2022 16:07:24 +0000 (11:07 -0500)]
nir: Unconditionally call nir_trim_vector in nir_lower_readonly_images_to_tex
It will already short-circuit if the number of components matches.
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19301>
Jason Ekstrand [Tue, 25 Oct 2022 16:06:07 +0000 (11:06 -0500)]
nir: Misc. style fixes to nir_lower_readonly_images_to_tex
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19301>
Jason Ekstrand [Tue, 25 Oct 2022 16:04:33 +0000 (11:04 -0500)]
nir: Use nir_shader_instructions_pass in nir_lower_readonly_images_to_tex
nir_shader_lower_instructions is overkill and this makes the pass
generally easier to understand.
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19301>
Jason Ekstrand [Tue, 25 Oct 2022 15:55:43 +0000 (10:55 -0500)]
nir: Reformat a comment
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19301>
Lucas Stach [Mon, 24 Oct 2022 16:40:50 +0000 (18:40 +0200)]
etnaviv: properly reference flush_resources
The flush_resources recorded in the context need to stay alive until
the context is flushed, at which point additional resolve operations
are done to those resources. While the backing BO is alive due to being
referenced in the cmdstream, the resource might already be destroyed
at this point.
Keep a reference to the resource to make sure it is still available at
context flush time.
Fixes:
7b9d8d1936d7 ("etnaviv: flush used render buffers on context flush when neccessary")
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19280>
Michel Dänzer [Fri, 21 Oct 2022 10:57:53 +0000 (12:57 +0200)]
winsys/amdgpu: Set RADEON_FLAG_32BIT again
Avoids hang running
rendercheck -t cacomposite -f a8r8g8b8
via glamor on Navi 14.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7167
Fixes:
7833c5139a54 ("winsys/amdgpu: use cached GTT for command buffers and don't set the 32BIT flag")
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19276>
SoroushIMG [Wed, 26 Oct 2022 15:36:19 +0000 (16:36 +0100)]
zink: limit gl_Layer clamping to drivers that need it
So far, only IMG drivers cannot handle out of bounds layer values.
Ideally, a vulkan extension will be drafted to detail this behavior.
But for now if KHR-GL46.texture_cube_map_array.color_depth_attachments
fails, then needs_sanitised_layer is probably needed.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19163>
SoroushIMG [Sun, 16 Oct 2022 18:41:16 +0000 (19:41 +0100)]
zink: clamp gl_Layer output to 0, if framebuffer is not layered
GL spec forces driver to ignore gl_Layer, if layered rendering
is not enabled.
Since vulkan doesn't have the same bavior, emulate this by forcing
gl_Layer to 0, based on driver internal state.
This was seen as failure in
KHR-GL46.texture_cube_map_array.color_depth_attachments
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19163>
SoroushIMG [Thu, 13 Oct 2022 21:52:26 +0000 (22:52 +0100)]
zink: add new framebuffer_is_layered state
This state is needed to make sure gl_Layer values are set to 0,
when the framebuffer is not layered accorfing to GL spec.
Specifically Section 9.8 Layered Framebuffers of GL46 spec:
A layer number written by a geometry shader has no effect if
the framebuffer is not layered.
Vulkan has no carve out for this, so zink must handle this by
sanitising gl_Layer (next commit in the series).
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19163>
SoroushIMG [Thu, 13 Oct 2022 21:48:44 +0000 (22:48 +0100)]
zink: add pushconst only pipeline layout
Now that all gfx pipelines share the same push constant layout,
create a screen wide push const only layout that is compatible
with all future programs.
This layout will be used to update push constant values, so that
the update can happen at any point before draw call.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19163>
SoroushIMG [Mon, 24 Oct 2022 14:13:28 +0000 (15:13 +0100)]
zink: cleanup zink_pipeline_layout_create
move the hashing to the caller, since it's not related to this.
Additionally, remove dependance on zink_program argument.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19163>
SoroushIMG [Mon, 24 Oct 2022 13:45:12 +0000 (14:45 +0100)]
zink: use unified pushconst layour for passthorugh tcs
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19163>
SoroushIMG [Thu, 13 Oct 2022 16:49:50 +0000 (17:49 +0100)]
zink: cleanup pushconst interface between driver/compiler
Extend vs_pushconst structure to all gfx stages and make sure,
the push constant memory layout is defined in one place and
is therefore always correct.
No functional change, but should make adding new members to
zink_*_push_constant easier.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19163>
SoroushIMG [Wed, 19 Oct 2022 12:19:20 +0000 (13:19 +0100)]
lavapipe: stop allocating 0 size const buffer
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19163>
Lionel Landwerlin [Wed, 26 Oct 2022 10:55:25 +0000 (13:55 +0300)]
anv: enable localized loads for lower_shader_calls
On Q2RTX shaders :
Instructions in all programs: 31039 -> 26150 (-15.8%)
SENDs in all programs: 1587 -> 1148 (-27.7%)
Loops in all programs: 4 -> 4 (+0.0%)
Cycles in all programs: 420218 -> 392179 (-6.7%)
Spills in all programs: 157 -> 132 (-15.9%)
Fills in all programs: 337 -> 262 (-22.3%)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Acked-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16556>
Lionel Landwerlin [Wed, 19 Oct 2022 14:10:48 +0000 (17:10 +0300)]
radv: tweak lower_shader_calls parameters
On Q2RTX shaders :
MaxWaves: 62 -> 69 (+11.29%)
Instrs: 41626 -> 41575 (-0.12%); split: -0.27%, +0.15%
CodeSize: 224960 -> 223740 (-0.54%); split: -0.62%, +0.08%
VGPRs: 800 -> 704 (-12.00%)
Scratch: 75776 -> 70656 (-6.76%)
Latency: 922219 -> 977997 (+6.05%)
InvThroughput: 212154 -> 201746 (-4.91%); split: -5.54%, +0.64%
VClause: 1120 -> 1155 (+3.12%); split: -1.88%, +5.00%
SClause: 1148 -> 1144 (-0.35%); split: -0.70%, +0.35%
Copies: 5840 -> 5788 (-0.89%); split: -0.94%, +0.05%
PreVGPRs: 753 -> 651 (-13.55%)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16556>
Lionel Landwerlin [Thu, 20 Oct 2022 08:53:15 +0000 (11:53 +0300)]
nir/lower_shader_calls: run opt_cse after lower stack intrinsics
In particular when using scratch_base_ptr
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16556>
Lionel Landwerlin [Wed, 18 May 2022 15:31:27 +0000 (18:31 +0300)]
nir/lower_shader_calls: move scratch loads closer to where they're needed
The intel backend compiler is not dealing with the scratch loads
emitted by this pass very well. There are 2 reasons for this :
- all loads are at the top of the shader
- the loads are global load intrinsics (cannot be differentiated
from ssbo loads for example)
This leads the backend to generate ridiculous amount of spills.
To help a bit (actually quite a lot), we can move the scratch loads in
the blocks where they're needed, using the dominance information.
Quite often that also ends up moving loads in a block that might not
be reached by all the lanes, so we're potentially avoiding some loads.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16556>
Lionel Landwerlin [Mon, 22 Aug 2022 07:33:40 +0000 (10:33 +0300)]
nir/lower_shader_calls: add a pass to sort/pack values on the stack
The previous pass shrinking values stored on the stack might have left
some gaps on the stack (a vec4 turned into a vec3 for instance).
This pass reorders variables on the stack, by component bit size and
by ssa value number. The component size is useful to pack smaller
values together. The ssa value number is also important because if we
have 2 calls spilling the same values, then we can avoid reemiting the
spillings if the values are stored in the same location.
v2: Remove unused sorting function (Konstantin)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16556>
Lionel Landwerlin [Wed, 17 Aug 2022 13:23:51 +0000 (16:23 +0300)]
nir/lower_shader_calls: add a pass to trim scratch values
For example, if we store to scratch a vec4 but only a subset of
components are used after the load operation.
v2: Use nir_intrinsic_write_mask (Konstantin)
Use u_foreach_bit() instead of u_bit_scan() (Konstantin)
Fix mask building loop (Konstantin)
v3: Fix reswizzle (Konstantin)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16556>
Lionel Landwerlin [Wed, 19 Oct 2022 13:33:20 +0000 (16:33 +0300)]
nir/lower_shader_calls: add an option structure for future optimizations
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16556>
Lionel Landwerlin [Wed, 24 Aug 2022 14:15:58 +0000 (17:15 +0300)]
nir/lower_shader_calls: cleanup shaders a bit more post split
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16556>
Lionel Landwerlin [Mon, 22 Aug 2022 07:25:14 +0000 (10:25 +0300)]
nir/lower_shader_calls: add NIR_PASS_V internally
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16556>
Lionel Landwerlin [Wed, 24 Aug 2022 20:24:57 +0000 (23:24 +0300)]
nir/lower_shader_calls: rematerialize values in more complex cases
Previously when considering whether to rematerialize or spill/fill
ssa_1954, we would go for a spill/fill :
vec4 32 ssa_388 = (float32)txf ssa_387 (texture_handle), ssa_86 (coord), ssa_23 (lod), 0 (texture), 0 (sampler)
...
vec1 32 ssa_1953 = load_const (0xbd23d70a = -0.040000)
vec1 32 ssa_1954 = fadd ssa_388.x, ssa_1953
vec1 32 ssa_1955 = fneg ssa_1954
This is because when looking at ssa_1955 the first time, we would
consider ssa_388 unrematerialiable, and therefore all values built on
top of it would be considered unrematerialiable as well.
The missing piece when considering whether to rematerialize ssa_1954
is that we should look at filled values. Now that ssa_388 has been
spilled/filled, we can rebuild ssa_1955 on top of the filled value and
avoid spilling/filling ssa_1955 at all.
This requires a bit more work though. We can't just look at an
instruction in isolation, we need to go through the ssa chains until
we find values we can rematerialize or not.
In this change we build a list of all ssa values involved in building
a given value, up to the point there we find a filled or a
rematerializable value.
In this particular case, looking at ssa_1955 :
* We can rematerialize ssa_388 from its filled value
* We can rematerialize ssa_1953 trivially
* We can rematerialize ssa_1954 because its 2 inputs are rematerializable
* We can rematerialize ssa_1955 because ssa_1954 is rematerializable
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16556>
Lionel Landwerlin [Mon, 16 May 2022 13:23:02 +0000 (16:23 +0300)]
nir/lower_shader_calls: avoid respilling values
Currently we do something like this :
ssa_0 = ...
ssa_1 = ...
* spill ssa_0, ssa_1
call1()
* fill ssa_0, ssa_1
ssa_2 = ...
ssa_3 = ...
* spill ssa_0, ssa_1, ssa_2, ssa_3
call2()
* fill ssa_0, ssa_1, ssa_2, ssa_3
If we assign the same possition to ssa_0 & ssa_1 in the spilling
stack, then on call2(), we know that those values are already present
in memory at the right location and we can avoid respilling them.
The result would be something like this :
ssa_0 = ...
ssa_1 = ...
* spill ssa_0, ssa_1
call1()
* fill ssa_0, ssa_1
ssa_2 = ...
ssa_3 = ...
* spill ssa_2, ssa_3
call2()
* fill ssa_0, ssa_1, ssa_2, ssa_3
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16556>
Lionel Landwerlin [Wed, 18 May 2022 15:29:10 +0000 (18:29 +0300)]
nir/lower_shader_calls: lower scratch access to format internally
For a follow up optimization, we would like to track scratch loads.
This isn't possible with global load/store intrinsics. So use a couple
of special intrinsic in the pass and only lower it to global
intrinsics at the end.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16556>
Lionel Landwerlin [Tue, 9 Aug 2022 19:09:52 +0000 (22:09 +0300)]
nir/lower_shader_calls: rematerialize more trivial values
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16556>
Alejandro Piñeiro [Tue, 18 Oct 2022 12:25:14 +0000 (14:25 +0200)]
broadcom/compiler: call nir_opt_gcm with a custom strategy
nir_opt_gcm get us worse shader-db stats, but that is expected. But we
want to prevent to get worse values on spill/fills. Analyzing the
outcome with shader-db, this mostly happen with shaders that are
already complex, and are already spilling/filling.
So the best option here is adding a new strategy, that fall backs if
we get spill/fill using nir_opt_gcm.
It is not clear in which order we should disable gcm. For now we
disable it before loop unrolling.
We get a slight performance gain (in average) using nir_opt_gcm.
We don't show the shaderdb stats, as they are worse, but as mentioned,
this is expected.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17185>
Alejandro Piñeiro [Sun, 31 Jul 2022 00:51:05 +0000 (02:51 +0200)]
broadcom/compiler: pass a strategy struct to vir_compile_init
That allows to reduce the number of parameters of the method. And
after all, they were already filled using an existing strategy struct.
This would make easier adding new fields on a strategy.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17185>
Alejandro Piñeiro [Mon, 13 Jun 2022 12:04:55 +0000 (14:04 +0200)]
v3dv/pipeline: use v3d_optimize_nir
Instead of using a custom optimize_nir method, with the same purpose.
Running the fossils for the v3dv well know applications (ue4 demos,
Quake3d, etc) we got somewhat inconclusive outcome in general,
although slightly worse values:
Instrs: 265129 -> 265277 (+0.06%); split: -0.06%, +0.12%
Thread Count: 5504 -> 5506 (+0.04%)
Totals from 153 (10.23% of 1495) affected shaders:
Instrs: 84603 -> 84751 (+0.17%); split: -0.19%, +0.37%
Thread Count: 316 -> 318 (+0.63%)
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17185>
Alejandro Piñeiro [Mon, 13 Jun 2022 10:43:12 +0000 (12:43 +0200)]
broadcom/compiler: add more lowerings/optimizations on v3d_optimize_nir
Optimizations that we are already calling on the Vulkan driver. As
preparation to the Vulkan frontend to use v3d_optimize_nir too.
We need to add a new parameter to v3d_optimize_nir in order to know if
we can call nir_opt_find_array_copies. As we don't track if we are
calling nir_var_lower_copies, we explicitly call it when we create the
uncompiled shader create. So instead of tracking, we assume that each
driver (v3d/v3dv) would call it when the shader is created. So when
v3d_optimize_nir is called as part of the process to compile it at the
compiler, we call it with allow_copies as false.
We exclude on purpose nir_opt_gcm as it is a case of a optimization
that could help performance even if it hurts shader db stats.
shaderdb stats:
total instructions in shared programs:
11705923 ->
11705034 (<.01%)
instructions in affected programs: 88350 -> 87461 (-1.01%)
helped: 201
HURT: 80
Instructions are helped.
total threads in shared programs: 375552 -> 375558 (<.01%)
threads in affected programs: 6 -> 12 (100.00%)
helped: 3
HURT: 0
total uniforms in shared programs: 3486108 -> 3485789 (<.01%)
uniforms in affected programs: 7473 -> 7154 (-4.27%)
helped: 90
HURT: 1
Uniforms are helped.
total max-temps in shared programs: 2021860 -> 2021802 (<.01%)
max-temps in affected programs: 800 -> 742 (-7.25%)
helped: 21
HURT: 3
Max-temps are helped.
total sfu-stalls in shared programs: 19299 -> 19296 (-0.02%)
sfu-stalls in affected programs: 18 -> 15 (-16.67%)
helped: 10
HURT: 7
Inconclusive result (value mean confidence interval includes 0).
total inst-and-stalls in shared programs:
11725222 ->
11724330 (<.01%)
inst-and-stalls in affected programs: 88402 -> 87510 (-1.01%)
helped: 201
HURT: 80
Inst-and-stalls are helped.
total nops in shared programs: 269674 -> 269386 (-0.11%)
nops in affected programs: 3641 -> 3353 (-7.91%)
helped: 103
HURT: 29
Nops are helped.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17185>
Alejandro Piñeiro [Thu, 16 Jun 2022 21:26:22 +0000 (23:26 +0200)]
broadcom/compiler: update how we compute return_words_of_texture_data on non-ssa
For the non-ssa case, we were trying to use reg->num_components. But
this is not the same that nir_ssa_def_components_read. It is the
number of components of the destination register. And in the 16bit
case, even if nir_lower_tex packs the outcome, it doesn't update the
number of components, as nir_tex_instr_dest_size would still return
4. And nir validate would check that those values are the same.
So this change focuses on the last part of this comment at
nir_lower_tex:
* Note that we don't change the destination num_components, because
* nir_tex_instr_dest_size() will still return 4. The driver is just
* expected to not store the other channels, given that nothing at the
* NIR level will read them.
We just limit how many channels we would use for the f16 case.
It is also worth to note, based on the CTS and different applications
we test, that this is a corner case.
This was detected when we experimented to enable nir_opt_gcm for v3d,
that lead to raise an assertion slightly below with some shaderdb
tests, but technically it could happen without it.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17185>
Alejandro Piñeiro [Mon, 13 Jun 2022 12:26:17 +0000 (14:26 +0200)]
broadcom/compiler: don't call nir_opt_load_store_vectorize on all v3d_optimize_nir calls
For compute shaders, to avoid a crash with that optimization, it requires
doing some optimizations and lowerings before. Example:
static void
lower_cs_shared(struct nir_shader *nir)
{
NIR_PASS_V(nir, nir_lower_vars_to_explicit_types,
nir_var_mem_shared, shared_type_info);
NIR_PASS_V(nir, nir_lower_explicit_io,
nir_var_mem_shared, nir_address_format_32bit_offset);
}
In the same way other drivers (like anv) calls
nir_opt_load_store_vectorize as part of their post-process-nir.
So one option would be to move nir_opt_load_store_vectorize outsize
the common v3d_nir_optimize, to a post-process nir method.
To make things simpler, this change calls that optimization only if we
have a v3d_compiler object, that is when each frontend has already
done their lowerings, and call the v3d_compiler to get the final
assembly (so we are already on a kind of post processing nir step).
This avoids dEQP-VK.memory_model.shared.basic_types.3 crashing if we
start to call v3d_optimize_nir on v3dv directly.
Slight shaderdb changes, but not significant.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/17185>
Yusuf Khan [Wed, 28 Sep 2022 01:01:50 +0000 (20:01 -0500)]
nouveau: put nv04 push macros in nouveau_winsys
Remove some of the duplicated code that comes from it
Signed-off-by: Yusuf Khan <yusisamerican@gmail.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18857>
Karol Herbst [Sun, 16 Oct 2022 21:02:01 +0000 (23:02 +0200)]
rusticl/mem: can only map staging textures directly
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19213>
Karol Herbst [Sun, 16 Oct 2022 20:59:47 +0000 (22:59 +0200)]
rusticl: force BIND_LINEAR on staging resources
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19213>
Gert Wollny [Tue, 25 Oct 2022 15:42:12 +0000 (17:42 +0200)]
r600/sfn: evaluate LDS location for color and clip-vertex too
This is required to support compatibility contexts
Related: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6926
Fixes:
3340c7ce359252ad09b3e4d338837944482fb248
r600/sfn: lower CLIPVERTEX to clip planes
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19300>
Gert Wollny [Tue, 25 Oct 2022 15:41:15 +0000 (17:41 +0200)]
r600: Account for color and clipvertex when evaluating LDS space
Related: https://gitlab.freedesktop.org/mesa/mesa/-/issues/6926
Fixes:
3340c7ce359252ad09b3e4d338837944482fb248
r600/sfn: lower CLIPVERTEX to clip planes
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19300>
Gert Wollny [Tue, 25 Oct 2022 15:29:16 +0000 (17:29 +0200)]
r600/sfn: elimiate dead registers too
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19300>
Gert Wollny [Tue, 11 Oct 2022 06:14:05 +0000 (08:14 +0200)]
r600/sfn: Increase scheduling priority on uniform reads and non-ssa writes
* Non-ssa values are pre-allocated, so assigning values to these is
unlikely to increase register pressure.
* Uniforms are often used early in the shader to evaluate dependend
values so, don't penalize their scheduling priority like literal
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19300>
Gert Wollny [Fri, 7 Oct 2022 18:36:43 +0000 (20:36 +0200)]
r600/sfn: improve scheduling of tex sources
Let tex sources switch the channel to unused channels.
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19300>
Gert Wollny [Fri, 7 Oct 2022 13:54:52 +0000 (15:54 +0200)]
r600/sfn: Trigger TEX CF based on max TEX CF size
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19300>
Gert Wollny [Fri, 7 Oct 2022 13:52:15 +0000 (15:52 +0200)]
r600/sfn: Allow copy-prop of group dest into origin
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19300>
Gert Wollny [Fri, 7 Oct 2022 18:09:18 +0000 (20:09 +0200)]
r600/sfn: Add test for channel changes in TEX source from opt
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19300>
Gert Wollny [Fri, 7 Oct 2022 14:21:42 +0000 (16:21 +0200)]
r600: Fix printing t-channel in diss-assambly
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19300>
Gert Wollny [Fri, 14 Oct 2022 13:29:55 +0000 (15:29 +0200)]
r600/sfn: Handle nir_op_seq and nir_op_sne
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19300>
Gert Wollny [Fri, 21 Oct 2022 18:43:23 +0000 (20:43 +0200)]
r600/sfn: Unify the handling of resource IDs in instruction
Fetch, GDS, Texture, and RAT instructions all use resources
with a possible offset defined by the index register. Unify
the handling of resource ID and the offset register for these
instruction types.
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19300>
Vinson Lee [Sun, 23 Oct 2022 19:28:44 +0000 (12:28 -0700)]
v3d: Fix initializer-overrides warning.
../src/gallium/drivers/v3d/v3d_screen.c:733:27: warning: initializer overrides prior initialization of this subobject [-Winitializer-overrides]
.lower_mul_high = true,
^~~~
/usr/lib/llvm-14/lib/clang/14.0.0/include/stdbool.h:16:14: note: expanded from macro 'true'
^
../src/gallium/drivers/v3d/v3d_screen.c:726:27: note: previous initialization is here
.lower_mul_high = true,
^~~~
/usr/lib/llvm-14/lib/clang/14.0.0/include/stdbool.h:16:14: note: expanded from macro 'true'
^
Fixes:
73e8fc3efbf ("broadcom/compiler: don't use imprecise_32bit_lowering for idiv lowering")
Signed-off-by: Vinson Lee <vlee@freedesktop.org>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19262>
Yonggang Luo [Sun, 23 Oct 2022 14:54:01 +0000 (22:54 +0800)]
xlib: Remove the usage of global constructor in xlib.c, so the xm_public.h are removed
The usage of global constructor should be limited, only in absolutely needed case.
The call style of xlib_create_screen referenced to osmesa_create_screen
Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19258>
Yonggang Luo [Sun, 23 Oct 2022 14:45:59 +0000 (22:45 +0800)]
xlib: Getting xmesa_strict_invalidate to be function
So that doesn't be called in global constructor
Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19258>
Yonggang Luo [Mon, 24 Oct 2022 15:39:01 +0000 (23:39 +0800)]
ci: Getting debian-clang-release to building with glx=xlib
Because this is just building, didn't not used by CTS runner,
So enable glx=xlib in this build variant
Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19258>
Yonggang Luo [Sat, 22 Oct 2022 09:58:41 +0000 (17:58 +0800)]
ci/macOS: Getting the installed binary to be artifacts
Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19258>
Yonggang Luo [Sun, 23 Oct 2022 14:47:36 +0000 (22:47 +0800)]
xlib: Indent with space and trim trailing spaces of xlib files
Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Ack-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19258>
Yonggang Luo [Sun, 23 Oct 2022 15:47:32 +0000 (23:47 +0800)]
gallium: Remove mgl* prefix in linker script osmesa.sym and libgl-xlib.sym
Seeing no mgl prefixed function in codebase, so removed it
Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Ack-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19258>
Yonggang Luo [Sun, 23 Oct 2022 15:00:09 +0000 (23:00 +0800)]
xlib: Remove the linkage hack in lib.c by use meson link_whole option
Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Ack-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19258>
Yonggang Luo [Sat, 22 Oct 2022 06:49:56 +0000 (14:49 +0800)]
xlib: Fixes compiling on linux
configure command line:
meson ../.. -Dglx=xlib -Dosmesa=true
link error:
virgl_screen.c.o: in function `virgl_is_video_format_supported':
/mnt/c/work/xemu/mesa/build/windows-wsl2/../../src/gallium/drivers/virgl/virgl_screen.c:885: undefined reference to `vl_video_buffer_is_format_supported'
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7545
Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Ack-by: Eric Engestrom <eric@igalia.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19258>
Alyssa Rosenzweig [Fri, 21 Oct 2022 15:43:15 +0000 (11:43 -0400)]
panfrost: Use proper formats for pntc varying
The formats of special attributes are supposed to match their architectural
definitions, and point coordinates are architecturally defined as RGBA32F. In
practice this doesn't seem to fix anything.
Cc: mesa-stable
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19237>
Alyssa Rosenzweig [Fri, 21 Oct 2022 14:57:55 +0000 (10:57 -0400)]
panfrost: Don't use lower_wpos_pntc on Midgard
gl_PointCoord is implemented via a special attribute descriptor on Midgard. This
descriptor has an orientation bit, the orientation is driver-controlled. That
means we can map rast->sprite_coord_mode to this bit, rather than lowering in
the shader.
This is a bug fix for point sprites, which are implemented natively on Midgard
for dubious reasons and need to be flipped this way. It is also an optimization
for apps reading gl_PointCoord, removing the extra arithmetic to flip, although
the value of this is somewhat dubious.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19237>
Emma Anholt [Tue, 25 Oct 2022 17:28:35 +0000 (10:28 -0700)]
ci/freedreno: Update known flakes/timeouts.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19304>
Emma Anholt [Tue, 25 Oct 2022 17:19:25 +0000 (10:19 -0700)]
ci/freedreno: Update CivV trace expectation.
Rendered scene still looks good, just a few pixels changed a bit.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19304>
Karol Herbst [Sat, 15 Oct 2022 21:00:37 +0000 (23:00 +0200)]
aux/trace: add set_global_binding
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19214>
Mike Blumenkrantz [Mon, 24 Oct 2022 20:46:59 +0000 (16:46 -0400)]
zink: elide a buffer samplerview update conditional
this is really part of the previous conditional blocks
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19306>
Mike Blumenkrantz [Mon, 24 Oct 2022 20:44:34 +0000 (16:44 -0400)]
zink: rescope/simplify some push descriptor code
this is only used in the push descriptor block, so move it there and
simplify redundant cases
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19306>
Mike Blumenkrantz [Mon, 24 Oct 2022 20:58:18 +0000 (16:58 -0400)]
zink: flag push state changed in batch changed conditional
this will matter after future refactoring
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19306>
Mike Blumenkrantz [Mon, 24 Oct 2022 18:30:18 +0000 (14:30 -0400)]
zink: split out uniform decriptor bindings in shader data
this is both more usable and fixes some descriptor util functionality
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19306>
Mike Blumenkrantz [Fri, 21 Oct 2022 19:05:16 +0000 (15:05 -0400)]
zink: simplify some program descriptor helpers
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19306>
Karol Herbst [Tue, 25 Oct 2022 22:02:15 +0000 (00:02 +0200)]
zink: fix spirv_builder_spec_const_uint
We need to create the type upfront otherwise the spir-v binary gets
corrupted.
Fixes:
c305a2c9625 ("zink: move spec constant emission to the types/consts block")
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19310>
Danylo Piliaiev [Wed, 19 Oct 2022 16:47:36 +0000 (18:47 +0200)]
freedreno: Add 'replay' tool which allows to replay cmdstreams
Replay command stream obtained from:
- /sys/kernel/debug/dri/0/rd
- /sys/kernel/debug/dri/0/hangrd
!!! Command stream capture should be done with ALL buffers:
- echo 1 > /sys/module/msm/parameters/rd_full
Requires kernel with MSM_INFO_SET_IOVA support.
This tool is intended for reproduction of various GPU issues:
- GPU hangs, note that command stream obtained from hangrd
may not reproduce (rarely) the hang, since the buffers are
snapshotted at the moment of the hang and not at the start
of the hanging command stream.
- TODO: Misrendering, would require marking framebuffer images
at each renderpass in order to fetch and decode them.
Code from Freedreno/Turnip is not re-used here since the relevant
pieces may introduce additional allocations which cannot be allowed
during the replay.
Signed-off-by: Danylo Piliaiev <dpiliaiev@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19171>
Samuel Pitoiset [Mon, 24 Oct 2022 02:15:06 +0000 (02:15 +0000)]
aco: add support for device clock on GFX11
According to LLVM, s_sendmsg_rtn(GET_REALTIME) should be used instead
of s_memrealtime.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19267>
Samuel Pitoiset [Tue, 25 Oct 2022 11:17:34 +0000 (13:17 +0200)]
aco: split the sendmsg enumeration into sendmsg_rtn
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19267>
Samuel Pitoiset [Mon, 24 Oct 2022 02:14:24 +0000 (02:14 +0000)]
aco: add support for s_sendmsg_rtn_b{32,64}
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19267>
Samuel Pitoiset [Fri, 21 Oct 2022 00:37:28 +0000 (00:37 +0000)]
ac/llvm: add support for device clock on GFX11
Signed-off-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/19267>