NIR doesn't really support global instructions such as global val
initilisation. So here we add functionality to glsl_to_nir() to
put these instructions into a temporary function that will be
later inlined into main.
We give the function a name starting with gl_mesa_tmp_ as functions
starting with gl_ are reserved and will not have any clashes with
user functions, we finish the name with the blake3 of the shader
source to avoid conflicts with multiple shaders attached to a single
stage.
Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31137>
Most of our visitors have custom destructors because they wrap C types,
and need to manually call destructors for those types. Because of this,
the implicitly generated copy constructors and copy-assignment operators
are not safe due copying pointers rather than data. Since we don't need
these features just delete them, so any attempt to use them would be a
compilation error. This doesn't fix any existing issue except stopping
coverity from complaining, but it does prevent new issues in the future.
Acked-by: Caio Oliveira <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29662>
Upcoming more strict context vs glsl version checks will fail otherwise.
Since standalone compiler requires ARB_ES3_2_compatibility that requires
GL 4.5 we simply set that as the context version.
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31061>
Previously we didn't check if the glsl version supported overload resolution
at link time only compile time. Note with this change we just use the
has_implicit_int_to_uint_conversion bool as support for both is added under
the same conditions.
Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30752>
If doubles are not supported by the shader the compiler will throw
an error if it sees one, there is no need to check if they are
supported in this helper for implicit conversions.
Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30752>
The bindless spec has no language requiring functions params to be
defined as bindless so we need to be able to look at the values being
passed to functions to decide if they are bindless or not. This
intrinsic allows us to wait until function inlining is complete to make
this assessment.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/11535
Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30315>
Instead of having a hardcoded list of endian-independent format aliases
in the header, generate them from the format definitions.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29649>
GLSL 4.40 changed the relevant language in Section 8.13.2 (Interpolation
Functions) to:
"Component selection operators (e.g., .xy) may be used when specifying
interpolant."
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30239>
Ensure the following identities hold to match IEEE-754-2019 and upcoming NIR:
min(-0, +0) = -0
min(+0, -0) = -0
max(-0, +0) = +0
max(+0, -0) = +0
To implement, we specialize a version of flt64_nonnan. The regular flt64 has
extra logic to handle signed zero, so this version is actually simpler. So in
addition to the bug fix, this is an optimization. Compute shaders from
KHR-GL46.gpu_shader_fp64.builtin.max_dvec4 before and after:
before: 136 inst, 122 alu, 122 fscib, 4 ic, 1006 bytes, 39 regs, 28 uniforms
after: 104 inst, 90 alu, 90 fscib, 4 ic, 766 bytes, 39 regs, 28 uniforms
I will happy take a 24% reduction in instruction count as the cost of standards
conformance ^_^
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Konstantin Seurer <konstantin.seurer@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30075>
nir_lower_io_to_temporaries is now done for all stages except TCS, and
nir_lower_io_to_temporaries with a TCS is a no-op.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25590>
It seems only radeonsi and v3d sets
CAN_READ_OUTPUTS/SupportsReadingOutputs, and v3d has
lower_all_io_to_temps=true. It looks like radeonsi basically lowers the
outputs to temporaries in the backend.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Timothy Arceri <tarceri@itsqueeze.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25590>
This fixes test_gl_lower_mediump to properly test linking, which also
means we can drop all the custom nir calls as we are now simply passing
the tests directly through the real nir linking code.
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30034>
This updates the scaffolding to reflect init_shader_program() and
will be required in the following patch to avoid a segfault.
Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30034>
Driver with compact_arrays=false (i.e. radeonsi) is broken when
tess factor is accessed indirectly, for example:
gl_TessLevelOuter[gl_InvocationID] = xxx;
This fix use nir_vectorize_tess_levels to lower array tess factor
access into direct vector access before nir_lower_io() like clip
and cull distance way.
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/29799>
The standalone compiler previously ran these tests through a hacked up
partial linker. When this partial linker was recently removed from the
standalone compiler the --link option was turned on because some tests
are testing linking not just compilation. However in a future patchset
we will switch the standalone linker to use the nir linking code and
when this is done all of these shaders will need to pass full linking,
so here we update them to do so.
Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29991>
There are no more users of the glsl ir at this point in the standalone compiler
anymore for these optimisations. Later patches will also switch the
standalone compiler to the nir linker.
Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29991>
We move this later so that we can call glsl_to_nir() on glsl ir that
has not set the array size on unsized ifc members. Later patches will
move sizing of the arrays out of glsl ir and into the nir linker.
Acked-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29991>
