GitLab

We need to iterate the whole row, we can't be clever and only look at
one side, the symmetry doesn't work like that. See the original paper.

total instructions in shared programs: 69392 -> 69322 (-0.10%)
instructions in affected programs: 9002 -> 8932 (-0.78%)
helped: 82
HURT: 28
Instructions are helped.

total bundles in shared programs: 32225 -> 32155 (-0.22%)
bundles in affected programs: 4286 -> 4216 (-1.63%)
helped: 82
HURT: 28
Bundles are helped.

total quadwords in shared programs: 56102 -> 56102 (0.00%)
quadwords in affected programs: 0 -> 0
helped: 0
HURT: 0

total registers in shared programs: 4552 -> 4572 (0.44%)
registers in affected programs: 298 -> 318 (6.71%)
helped: 18
HURT: 38
Registers are HURT.

total threads in shared programs: 3772 -> 3775 (0.08%)
threads in affected programs: 84 -> 87 (3.57%)
helped: 15
HURT: 14
Inconclusive result (value mean confidence interval includes 0).

total spills in shared programs: 0 -> 0
spills in affected programs: 0 -> 0
helped: 0
HURT: 0

total fills in shared programs: 0 -> 0
fills in affected programs: 0 -> 0
helped: 0
HURT: 0

Fixes: 66ad64d7 ("pan/midgard: Implement linearly-constrained register allocation")
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <mesa/mesa!9338>

libpanfrost_lib depends on libpanfrost_bifrost for 'bifrost_compile_shader_nir' symbol
libpanfrost_lib depends on libpanfrost_bifrost_disasm for 'disassemble_bifrost' symbol
LOCAL_STATIC_LIBRARIES requires proper ordering to make the symbols available

Fixes the following building error happening with Android P:

FAILED: out/target/product/x86_64/obj/SHARED_LIBRARIES/gallium_dri_intermediates/LINKED/gallium_dri.so
external/mesa/src/panfrost/lib/decode.c:534: error: undefined reference to 'disassemble_bifrost'
external/mesa/src/panfrost/lib/pan_shader.c:145: error: undefined reference to 'bifrost_compile_shader_nir'

Cc: 20.3 21.0 <mesa-stable@lists.freedesktop.org>
Fixes: 166630f7 ("android: pan/bi: Separate disasm/compiler targets")
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9265>

We don't support it yet.

Fixes: 61d3ae6e ("panfrost: Initial stub for Panfrost driver")
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <!9164>

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <!9164>

Fixes dEQP-GLES31.functional.copy_image.non_compressed.viewclass_32_bits.r32ui_rgb10_a2.texture2d_to_texture2d
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <!9164>

fp16 loses precision.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <!9164>

dEQP-GLES31.functional.texture.multisample.samples_16.sample_position is
failing on Bifrost, and we already weren't advertising on Midgard. Until
someone gets spare cycles to debug all the problems, keep it hidden
behind a debug flag to avoid introducing bugs.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <!9164>

The blob advertises 64 for this, so let's use the same value. Small
alignments are observed to raise an IMPRECISE_FAULT at least on Bifrost.
As a bonus this forces cache line alignment which will help perf. Fixes

dEQP-GLES31.functional.texture.texture_buffer.render.as_vertex_texture.offset_1_alignments

Fixes: 5f7bafa3 ("panfrost: Enable ARB_texture_buffer_object")
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <!9164>

The last of the nr_cbufs audit changes.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <!9164>

Auditing nr_cbufs. Note we have to augment the blending logic a bit to
use the same 'no blend' case for missing RTs as we do for depth-only
passes.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <!9164>

Auditing all uses of nr_cbufs
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <!9164>

Already checked in the callee, no need to check in the caller.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <!9164>

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <!9164>

At least this way failed RA will crash (by having no spill node to pick)
instead? Seen in
dEQP-GLES31.functional.ssbo.layout.random.all_shared_buffer.21 on
Bifrost.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <!9164>

I cannot reproduce the issue in local runs.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

Passes dEQP-GLES31.functional.fbo.no_attachments.*, it doesn't look like
we need to do anything special.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

Pending on kernel work.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

Don't hardcode the RT to 0. Affects ES3.0 which already exposes MRT --
despite no dEQP coverage of this particular corner case, apps could hit
this in the wild on 21.0. Fixes

dEQP-GLES31.functional.draw_buffers_indexed.overwrite_indexed.common_blend_func_buffer_blend_func

Fixes: c7e1ef7c ("panfrost: Advertise ES3.0 on Bifrost")
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

Already checked in format_supported (only difference being cube map
arrays but the extension isn't advertised yet anyway due to missing
geometry shaders.)
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

Passes dEQP-GLES31.functional.texture.border_clamp.*
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

This appears to be architecture dependent.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

Fixes debug builds of dEQP-GLES31.functional.stencil_texturing.render.*
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

Contains the InternalConversionDescriptor corresponding to the selected
render target, ready to be fed into LD_TILE.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

Will be used to lower EXT_shader_framebuffer_fetch on Bifrost.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

A few cases missed, probably some more bugs hidden here!
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

We can use the constant format anywhere safely as a 0-component format,
which will map fine for PIPE_FORMAT_NONE. Used so
ARB_framebuffer_no_attachments can probe.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

Add some notes on what the useful combinations do and flesh out the
logic. Fixes fails in
dEQP-GLES31.functional.image_load_store.early_fragment_tests.*

It's still not entirely clear to me what the difference between "force
early" and "weak early" is from the driver's perspective, though I
suspect there are performance implications to getting it wrong. Follow
what the blob does in case of doubt..
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

These extra pieces of data are needed to correctly compute pixel kill /
zs update settings.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

Helps debug pandecode issues.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!9105>

The compiler ABI specifies push uniforms at a 4-byte granularity (like
Bifrost), but Midgard require a 16-byte granularity. As such if the
number of pushed words is not a multiple of 4, there is a buffer overrun
at shader load time. Ordinarily this is inaccessible so the garbage is
ignored.

However, there was a great deal of confusion around the `uniform_cutoff`
variable. In some cases (such a full glmark2 run on a 64-bit processor),
the push uniforms would be at the end of a BO and the overrun would
cause a page fault.

Remove uniform_cutoff entirely and work with count directly to avoid
faulting, and round the count up to be defensive.

Closes: #4289
Fixes: ed810eb0 ("panfrost: Don't truncate uniform_count")
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Robin Murphy <robin.murphy@arm.com>
Part-of: <!9109>

Most of it is API-independent, so let's move it out of the gallium
driver so it can be shared with the Vulkan driver.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!8963>

Will be used by the pan_blend library.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!8963>

We are about to add a pan_blend.h in src/panfrost/lib. Rename the
existing pan_blend.h so we can include both.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!8963>

This implies late preparation of the fragment shader RSD, but given the
simplicity of pan_shader_prepare_rsd() (it's basically a 1:1 translation
between shader info and the RSD fields), it's unlikely to make a
difference. If we really want to optimize the time spent preparing the
RSD, we should consider caching a packed version at the batch level and
re-using it when nothing changed.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!8963>

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!8963>

While at it, rework the code to avoid copies between intermediate
structures: the pan_shader_info is passed to the compiler context so
the compiler can fill shader information directly.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!8963>

So we don't have to copy data around.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!8963>

So we can re-use the panfrost_sysvals definition outside of the
compiler without dragging the sysval_to_id hash table.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!8963>

Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <!8963>