In general we should only call it once, and then we should avoid to
call any lowering that introduce back copies. So far we were tracking
that manually out of the nir shader on several places.
Ideally we would like to add a nir_validate rule, but right now there
are some exceptions to this rule. For example right now the Intel
compiler calls nir_lower_io_to_temporaries as part of linking
tess_ctrl/mesh/task sahders.
One option would be to allow drivers to reset the value, but for now
let's not add that validation rule.
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19338>
This will be useful for RADV since it hashes the state.
v3dv changes:
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20731>
If we have multiple clip planes, rather than emit multiple discards we can just
OR together the discard criteria. Then a nir_opt_algebraic rule kicks in to
optimize out the flt/.../flt/ior/.../ior into fmin/.../fmin/flt, generating
much less code at the end.
Written while debugging an unrelated issue with the clip lowering.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21103>
These are an enum. Furthermore, their 0 state is INTERP_MODE_NONE which we
shouldn't bother printing at all.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21091>
Instead of printing the raw location number, which is pretty hard to interpret,
let's print the name of the location. Example output:
vec4 16 ssa_2 = intrinsic load_interpolated_input (ssa_0, ssa_1) (base=0,
component=0, dest_type=float16 /*144*/, io location=VARYING_SLOT_VAR0 slots=1
mediump /*8388768*/)
One of the "regressions" from moving to purely lowered I/O with all variables
removed is a lack of debuggability, since otherwise these location strings don't
show up anywhere in the printed shader! By contrast this should make the lowered
I/O nice to read like the early I/O.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21091>
Locations show up in two places: variables and lowered I/O semantics. We want to
reuse the logic in both places, so extract it out. The extracted logic is IMO
easier to read, too.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21091>
Builds on the work of !15121. This gets to delete even more code
because many drivers shared a lot of code for i2b and f2b.
No shader-db or fossil-db changes on any Intel platform.
v2: Rebase on 1a35acd8d9.
v3: Update a comment in nir_opcodes_c.py. Suggested by Konstantin.
v4: Another rebase. Remove f2b stuff from Midgard.
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20509>
There were only two users. Replace each with nir_fneu instead.
This is now a squash of what was two separate commits.
nir_lower_pstipple_block is called after nir_lower_bool_to_int32, so
nir_fneu32 has to be used here or there will be regresssions in stipple
tests on llvmpipe.
v2: Rebase on !20869.
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Suggested-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20509>
Add a second NIR pass for lowering point/texture coordinate replacement (i.e.
point sprites). Why a second one? The current pass works on derefs/variables,
which is good for drivers that don't lower I/O at all (like Zink, where the pass
originates). However, it is problematic for hardware drivers: the inputs to this
pass depend on the shader key, so we want to run the pass as late as possible to
minimize the cost of building/compiling the associated shader variants. In
particular, we need to be able to lower point sprites after lowering I/O if we
would like to lower I/O when preprocessing NIR.
The logic for early lowering and late lowering is considerably different (the
late lowering is a lot simpler), so I've split this out into a second pass
rather than trying to weld them together into one.
This pass will be used on Asahi, which currently uses the early pass. It may be
useful for other drivers as well. (Actually, it's been shipping on Asahi for a
little while now, just hasn't been sent upstream yet.)
Tested with Neverball.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Emma Anholt <emma@anholt.net>
Acked-by: Asahi Lina <lina@asahilina.net>
Acked-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21065>
while lowering image_samples to one, we need to take
nir_intrinsic_image_deref_samples and
nir_intrinsic_bindless_image_samples intrinsic into account.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/8211
Fixes: ab4c2990ed ("intel/compiler: use lower_image_samples_to_one")
Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21053>
When the unused channels were at the end and so no reswizzling was
needed, we wouldn't correctly mark the progress.
Fixes: 3305c960
Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21014>
When the unused channels were at the end and so no reswizzling was
needed, we wouldn't correctly mark the progress.
Fixes: cb7f2012
Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21014>
This can be used by multiple drivers that do not support ms images
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Reviewed-by: Rob Clark <robclark@freedesktop.org>
Reviewer-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Signed-off-by: Amber Amber <amber@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20813>
It's kosher to get load_preamble intrinsics ahead of time if the driver is
pushing sysvals. Handle them like load_uniform.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by-(with-sparkles): Asahi Lina <lina@asahilina.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20562>
Some backends may wish to reserve early uniforms for internal system values, and
use the remaining space for preamble storage. In this case, it's convenient to
teach nir_opt_preamble about a reserved offset. It's logical to treat the output
*size instead of an in/out variable that nir_opt_preamble adds to. This requires
a slight change to the consumers to zero the input.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Acked-by-(with-sparkles): Asahi Lina <lina@asahilina.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20562>
The Meson docs points out that it's better to use the files() function
when referring to files in the source tree than manually constructing
paths like this. Let's follow that advice, and get some neat cleanups.
Reviewed-by: Eric Engestrom <eric@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20907>
VKD3D-Proton DXBC f32 to f16 conversion implements a float conversion using PackHalf2x16.
Because the spec does not specify a rounding mode, it emits a sequence to ensure
D3D-like behaviour for infinity.
When we know the current backend has pack_half_2x16_rtz_split,
we can eliminate the extra sequence.
Fossil DB stats on GFX11:
Totals from 835 (0.62% of 134913) affected shaders:
VGPRs: 49368 -> 49224 (-0.29%)
CodeSize: 5341956 -> 5124564 (-4.07%)
Instrs: 1024062 -> 987041 (-3.62%)
Latency: 6530956 -> 6465120 (-1.01%); split: -1.01%, +0.00%
InvThroughput: 908189 -> 870253 (-4.18%)
VClause: 18704 -> 18702 (-0.01%); split: -0.02%, +0.01%
SClause: 33406 -> 33284 (-0.37%); split: -0.38%, +0.01%
Copies: 67440 -> 65992 (-2.15%); split: -2.15%, +0.00%
Branches: 18498 -> 18465 (-0.18%)
PreSGPRs: 38409 -> 38331 (-0.20%)
PreVGPRs: 44089 -> 43834 (-0.58%)
Note, some fossils are from before this pattern was added to VKD3D-Proton,
so the above may not reflect real-world impact.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15838>
Constant folding always uses RTNE for pack_half_2x16_split, but some
backends implement it with RTZ.
Lowering to RTZ when available ensures that the behaviour will be
consistent between constant folding and the backend.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15838>
Same as pack_half_2x16_rtz_split, but always uses RTZ mode.
Note that pack_half_2x16 rounding mode is unspecified.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Georg Lehmann <dadschoorse@gmail.com>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15838>
If the offset is negative like it's the case in
dEQP-VK.robustness.robustness2.bind.notemplate.r32i.unroll.volatile.storage_buffer_dynamic.readwrite.no_fmt_qual.len_256.samples_1.1d.comp
we end up passing the bounds checking condition because it's using
signed integers.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Suggested-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Cc: mesa-stable
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Reviewed-by: Emma Anholt <emma@anholt.net>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20762>
In this usual case, do a quick check to avoid generating 5 useless instructions
(mov/vec4 instructions). They'll get copypropped but that creates more work for
the optimizer and nir/lower_blend runs in a hot variant path on both Asahi and
Panfrost.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Acked-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20016>
nir/lower_blend asserts:
assert(nir_intrinsic_write_mask(store) ==
nir_component_mask(store->num_components));
For the special blend shaders used in Panfrost, this holds. But for arbitrary
shaders coming out of GLSL-to-NIR (as used with Asahi), this does not hold. In
particular, after nir_opt_undef runs, undefined components can be trimmed.
Concretely, if we have the shader:
gl_FragColor.xyz = foo;
Then this becomes in NIR
gl_FragColor = vec4(foo.xyz, undef);
and then opt_undef will give the store_deref a wrmask of xyz but 4 components.
Then lower_blend asserts out.
Found in a gfxbench shader on asahi.
Closes: #6982
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20016>
Per the spec.
Fixes arb_color_buffer_float-render on both Panfrost and Asahi (before/after
reproduced on Mali-T860 and AGX G13 respectively). Without that patch, that test
fails the assertion:
arb_color_buffer_float-render: ../src/compiler/nir/nir_lower_blend.c:259: nir_blend_logicop: Assertion `util_format_is_pure_integer(format)' failed.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20016>
The upper bits start correctly, there's no need to clear them as long as we keep
them zero'ed by using ixor with a valid bit mask instead of inot.
Makes the code generated for logic op slightly less ridiculous. I'm joking. It's
still ridiculous but I'm not in the mood to fix up the Midgard compiler and it's
just a little ALU for a feature almost nothing uses.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20016>
We need to sign extend. Incidentally this means the iand above is useless for
SNORM.
Fixes arb_color_buffer_float-render with GL_RGBA8_SNORM.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20016>
Particularly constant colours, but also (more obscurely) SNORM.
Fixes arb_color_buffer_float-render with SNORM framebuffers. Issue affects both
Asahi and Panfrost (the latter after we start advertising EXT_render_snorm).
v2: Check the blend factor to avoid unnecessary clamps. This avoids regressing
blend shader code quality on Panfrost.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com> [v1]
Acked-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20016>
In this case we have 4 components but the value of the fourth component
is undefined. Apply the fixup we already have.
Fixes
dEQP-GLES3.functional.draw_buffers_indexed.random.max_implementation_draw_buffers.0
on Asahi. That test blend with DST_ALPHA with its RGB565 attachment,
which is fine if RGB565 is preserved, but Asahi is demoting that
format to RGBX8 which means -- after lowering the tilebuffer access --
we blend with an ssa_undef.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20016>
Some drivers may encode constant offsets in the instruction, so
make it possible for the drivers to request lowering the atomic
uniform offset into the range_base variable of the intrinsic.
v2: drop patch to use build-in array offset evaluation, it makes
problems with zink, and update the code accordingly
v3: always initialize range base
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19980>
Add the intrinsic range_base value to the image intrinsics and add
the option to store the image array offset into range_base instead
of adding it to the image array index if the driver requests it.
v2: Always initialize range_base
v3: fix for bindless intrinsics
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19980>
As we've done for the AMD one, to prevent any codegen regression from switching
the Midgard lowering.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Italo Nicola <italonicola@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19350>
NIR has a fsin instruction that takes an argument in radians. Midgard instead
has an fsinpi argument that takes an argument in multiples of pi. So, we had a
NIR pass that would change fsin(x) to fsin(x / pi) and then map fsin to fsinpi
in the backend.
But that's invalid! In NIR, the opcode fsin is well-defined. fsin(x) means
something very different than fsin(x / pi). They won't usually be equal. The
transform fsin(x) -> fsin(x / pi) is fundamentally unsound.
It did work before, by accident. Most NIR passes don't care about the semantics
of ALU instructions. fsin(x) and fsin(x / pi) are both well-defined but
fundamentally different NIR shaders. So while rewriting is wrong -- the NIR we
get out is not equivalent to the NIR we put in, and the Midgard ops we generate
are not equivalent to the NIR -- but if we don't run any passes that care about
the definition of fsin the two wrongs will cancel out to make a right.
However, some NIR passes do care about the definitions of ALU instructions,
instead of treating them as named black boxes. In particular, constant folding
(nir_opt_constant_fold) evaluates ALU instructions when their inputs are
constants, according to the definition in nir_opcodes.py. So our little charade
will only work if we don't call nir_opt_constant_fold, or if all the fsin
instructions have non-constant inputs. At the beginning of this series, that is
the case. With the later scalarization change, that's no longer the case, and
the unsoundness translates to real failing tests rather than a quibble of NIR's
semantics.
To mitigate, we define a new NIR opcode with the semantics we want and translate
fsin(x) = fsin_mdg(x / pi), where that equivalence does hold mathematically. So
the new translation is sound and doesn't rely on lucky pass ordering.
This matches the approach already used for AMD and AGX, which have fsin_amd and
fsin_agx opcodes respectively.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Italo Nicola <italonicola@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19350>
load_preamble is intended to be almost free (costing at most a move), and it
does not have special bounds checking requirement, so it's ok to select with it.
With this, drivers that use nir_opt_preamble together with a late call to
peephole_select can optimize sequences like:
if (x) {
<uniform-on-uniform calculation>
} else {
<different uniform-on-uniform calculation>
}
to simply
bcsel(x, <uniform register 0>, <uniform register 1>)
rather than emitting needless control flow / branching over some moves.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Jason Ekstrand <jason.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20597>
This is heavily copy-pasted from a patch of Ian Romanick, including the
commit message.
Previously, this pass always generated fcsel for bcsel. This was the
only place that generate fcsel, so various drivers assumed (and needed!)
that src0 was a Boolean with 0.0 or 1.0 as the only values.
Specifically, many DX9 / GL_ARB_vertex_program platforms lack a CMP
instruction in vertex shaders. In those cases, they would use LRP to
implement fcsel. The bummer is that many plaforms have a real fcsel
instruction, and those platforms would benefit from other places
generating that opcode.
Instead of leaving assumptions in drivers about the sources of an opcode
that they can't really support, allow them to control the way the
lowering pass translates bcsel. Two flags are used to control this:
- If the driver sets has_fused_comp_and_csel in nir_options, fcsel_gt
will be used. Since the Boolean value is 0.0 or 1.0, this is
equivalent to fcsel.
- If the parameter has_fcsel_ne is set, fcsel will be used. This is the
old path.
- Otherwise, the lowering pass assumes we're on a crufty, old DX9 vertex
program, and it emits flrp.
With this, the assumptions about src0 of fcsel in NTT can be removed.
If a platform can't handle fcsel, it should ensure that the lowering
pass won't generate it.
No change in shader-db.
Signed-off-by: Pavel Ondračka <pavel.ondracka@gmail.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20162>
