Consider the following snippet from a Trine shader
20: RCP temp[40].z, temp[39].__w_;
21: MOV temp[40].xy, temp[34].-x-y__;
22: DP3 temp[41].x, temp[40].xyz_, temp[29].xyz_;
...
33: DP3 temp[52].x, temp[40].xyz_, temp[51].xyz_;
34: MAX temp[53].x, temp[52].x___, none.0___;
35: MUL temp[54].xy, temp[40].xy__, const[8].ww__;
36: MUL temp[55].xy, temp[54].xy__, temp[41].xx__;
37: MUL temp[56].x, temp[40].z___, const[8].w___;
When we search for writers for temp[40] so that we can check if we can
convert the MUL to omod, the corresponding variable actually contains
the RCP temp[40].z first and the MOV temp[40].xy is marked as friend.
However the current logic only checks the first instruction of variable,
so we fail to find the writer. Just search recursivelly also the
variable friends.
Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34462>
This moves per-patch output VMEM stores to the end of the shader where they
execute only once. They are skipped if the whole workgroup discards
all patches.
If tcs_vertices_out == 1, per-patch output VMEM stores use the same lanes
as per-vertex output VMEM stores, which are aligned to 4 or 8 lanes to get
cached bandwidth for the stores.
Previously, per-patch outputs were stored to memory for every store_output
intrinsic in TCS.
Additionally, LDS is no longer allocated for per-patch outputs that are only
written and read by invocation 0, or they are written by all invocations
but not read, and don't have indirect indexing. This reduces LDS usage and
LDS traffic.
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34780>
This unifies the duplicated LDS output patch size computation between
hs_output_lds_offset and ac_nir_compute_tess_wg_info.
"+ 4" to the output patch stride minimizes LDS bank conflicts by making
the beginning of each patch start on a different LDS bank.
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34780>
Checking whether every compoment is valid in tess_level_has_effect() when
prim_mode is unknown generated too many SALU. Do this instead:
if (triangles) ...
subgroup vote for triangles
else if (quads) ..
subgroup vote for quads
else // isoline
subgroup vote for isolines
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34780>
This rewrites tess level value tracking to use the 2-bit masks, which
means LDS allocation is determined separately for outer and inner levels.
LDS is not allocated for tess levels that are only written by invocation 0
and never read or only read by invocation 0. If the number of output
patch vertices is 1, LDS is also not allocated for tess levels.
Tess level outputs for TES are always written as whole vec4 to get cached
bandwidth.
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34780>
This improves write throughput for TCS outputs. It follows the same idea
as attribute stores in hw GS. The improvement is easily measurable with
a microbenchmark.
It also has the advantage that multiple output stores to the same address
don't result in multiple memory stores. Each output components gets only
one memory store at the end of the shader.
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34780>
It's a stride of 1 output, which isn't 16. It's 16 * num_threads,
aligned to 256.
tcs_offchip_layout has 5 unused bits, so let's use them.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34780>
One is only used by TCS, the other is only used by TES.
Use the same field for both, call it PATCH_VERTICES_IN.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34780>
This just removes the if/endif wrapping for LLVM, and hopefully the ACO
change does the same thing.
ACO had redundant code in endif_merged_wave_info, which is removed here.
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/34780>
This would cause Mesa to print this message even if an Intel GPU is just
being enumerated by a Vulkan application. For example, `vulkaninfo
--summary`.
Fixes: 52f73db5b7 ("brw: implement read without format lowering")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35396>
Got a consecutive timeout in two different `a660-vk` jobs on
`dEQP-VK.subgroups.clustered.compute.subgroupclusteredmax_u64vec4_requiredsubgroupsize`
test.
Analyzed the CI quality stats dashboard [1] (via job name filter) and also
found a flaky
`dEQP-VK.clipping.user_defined.clip_cull_distance.vert_tess_geom.1_7` on
the same job
[1] https://ci-stats-grafana.freedesktop.org/goto/1ai94GLHR?orgId=1
Signed-off-by: Guilherme Gallo <guilherme.gallo@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35377>
Usage is not a bitmask. This was causing us to not pick cached-coherent
for staging read buffers (ie. clEnqueueReadBuffer() and friends).
Signed-off-by: Rob Clark <rob.clark@oss.qualcomm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35371>
If we've already picked sysmem (for example, non-draw batches) no need
to autotune. Avoids autotune for compute-only workloads, for example.
Cuts down on per-batch overhead when we get a lot of single grid batches.
Signed-off-by: Rob Clark <rob.clark@oss.qualcomm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35371>
fd_device_gen() calls fd_dev_info_raw() which loops over all the device
table entries. No need to do this for every batch. And we _especially_
don't want to do this per-batch when rusticl cl profiling mode triggers
a batch flush for every launched kernel.
Signed-off-by: Rob Clark <rob.clark@oss.qualcomm.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35371>
Prior to Gen 11, we had to upload a bunch of SGVs
(FirstVertex, BaseVertex, BaseInstance, DrawID) via
3DSTATE_VERTEX_BUFFERS.
For Gen11+, we upload via 3DSTATE_SGVS_2 instead.
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35376>
The intention here is to build a SIMD8 value, that will be expanded
as needed -- just like the SHL/ADD case, but with a single instruction.
Found when the was triggering invalid MAD with SIMD32 (that gets compressed)
*and* with overlapping destination and source *and* which would cause
conflict when divided into two SIMD16.
Fixes: 338273dedd ("brw/reg_allocate: Optimize spill offset calculation using integer MAD")
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35302>
We recently found a weird case where an image was re-interpreted with
a format that had a bigger size than the image format, which led the
resulting texture descriptor to have a size that was bigger than what
the plane descriptors were covering.
Let's make sure this sort of unexpected use don't slip through in the
future by introducing basic image view consistency checks.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35317>
There are several problems in the current logic:
- it doesn't account for the WSI stride, and just assumes things are
tightly packed and the resource extent is aligned on a tile
- the format re-interpretation done in panfrost_mtk_detile_compute()
where the view format and the image format have a different block size
is not supposed to be supported, and that's something we'd like
to enforce now
- we write to textures that may have no descriptors bound to. This works
thanks to the robustness behind image stores/loads, but we'd probably
rather discard a bunch of instructions when Y/UV planes are copied
separately
- the linear to tiled coordinates conversion can be simplified by using
a combination of local_invocation_id+workgroup_id instead of the
global_invocation_id
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Tested-by: Eric R. Smith <eric.smith@collabora.com>
Reviewed-by: Eric R. Smith <eric.smith@collabora.com>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35317>
Those two types are completely different, and they don't use the same
key attributes. Let's provide two different helpers, and place the key
bits that are AFBC-specific into an struct inside a union. We also add
the modifier to the key to prevent AFBC/MTK_TILED collisions.
Not only this make things clearer, but we also no longer compile
MTK detiling shaders when AFBC shaders are requested (or the other way
around).
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Mary Guillemard <mary.guillemard@collabora.com>
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35317>
Adds support for reading PrimitiveID in a fragment shader assuming the
previous shader was vertex.
If support for shaders that can modify PrimitiveID is added, additional
changes will be required.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35367>
For formats that do not support LogicOps, NOOP should be treated as
COPY. This was implemented in nir/lower_blend, but an optimization in
PanVK would trigger for all NOOP that stopped this from happening.
This commit limits the NOOP optimization to supported formats.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35367>
