Without this patch, copy propagation pass can optimize out
buffer loads out of compare & swap loop, which then leads
to infinite loop.
Triggered by a change to atomicCompSwap float test in piglit.
Fixes: 8424cd8fbd ("nir: Account for atomics in copy propagation.")
Suggested-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Marcin Ślusarz <marcin.slusarz@intel.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7538>
It is UB to initialize unions on the stack and rely on bits not covered
by the initialized union member to be zero. Lets just simplify it and
move the entire nir_const_value off the stack.
While we're in there, sprinkle around some const.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3778
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7579>
This cleans up a bunch of gross sprintfs and keeps the caller from needing
to remember to ralloc_strdup. I added a couple of '"%s", name ? name :
""' to radv where I didn't fully trace through whether a non-null name was
being passed in.
I also took the liberty of adding a basic name to a few shaders (pan_blit,
unit tests)
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7323>
We can use num_blocks (if it's been initialized by some pass indexing
blocks) to pre-size our table, which helps on validating shaders with many
blocks which would otherwise reallocate the set several times.
No statistically significant performance difference on softpipe
KHR-GL33.texture_swizzle.functional runtime (n=15). A previous, similar
variant of this patch cut .3% of instructions in softpipe shader-db ./run
shaders/closed/steam/borderlands-2/35* (an arbitrary set of shaders that
completed in reasonable amount of time) according to callgrind.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7244>
Here's the code to emit DXIL code from NIR. It's big and bulky as-is,
and it needs to be split up a bit.
This is the combination of a lot of commits from our development branch,
containing code by several authors.
Co-authored-by: Bill Kristiansen <billkris@microsoft.com>
Co-authored-by: Boris Brezillon <boris.brezillon@collabora.com>
Co-authored-by: Daniel Stone <daniels@collabora.com>
Co-authored-by: Gert Wollny <gert.wollny@collabora.com>
Co-authored-by: Jesse Natalie <jenatali@microsoft.com>
Co-authored-by: Louis-Francis Ratté-Boulianne <lfrb@collabora.com>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7477>
Some ALU ops (comparisons being the primary example) have a fixed
bit-size destination and, in that case, we don't want to insert a
conversion on the destination.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7482>
GLSL requires that image atomics have formats and there are rules about
things matching properly. We should enforce those in NIR unless we have
reason to do otherwise.
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7509>
ctz is a CL2.0 opcode but 3.0 requires it as well so just add support
for it.
Tested against CTS integer_ops integer_ctz test.
(long line broken up)
Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7468>
We need to consider shader calls as potential writes to their payloads.
For other ray-tracing intrinsics, we may not have a shader payload
pointer and have to treat them more like a barrier. We also need to
ensure that global and SSBO reads/writes aren't propagated across shader
call intrinsics.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>
The SPV_KHR_ray_tracing extension adds 6 new storage classes which is a
bit on the ridiculous side. In order to avoid adding that many variable
modes to NIR, we make a few simplifying assumptions:
1. CallableData and RayPayload data actually lives on the stack
somewhere, presumably in the caller's stack. We assume that these
are no different from global variables and use nir_var_shader_temp
for them. We still need a separate storage class for the incoming
variants but only so we can figure out which one the incoming one
is and lower it to something useful.
2. There's no difference between incoming CallableData and RayPaolad
data. We can use a single storage class for both.
3. ShaderRecordBuffer data is just a global memory access. This lets
us avoid NIR variables entirely and just fetch the pointer via the
shader_record_ptr system value and it's accessed using a 64-bit
global memory pointer.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>
If we were desperate to reduce bits, we could probably also use
shader_in/out for hit attributes as they really are an output from
intersection shaders and read-only in any-hit and closest-hit shaders.
However, other passes such as nir_gether_info like to assume that
anything with nir_var_shader_in/out is indexed using vec4 locations for
interface matching. It's easier to just add a new variable mode.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>
These are a bit more tricky than most because they're matrix system
values. We make the intentional choice here to not bother with allowing
indirect addressing of columns for these. Since they're system values,
they may be magically constructed somehow or come from weird hardware so
it's easier on back-ends to just handle any indirects with bcsel.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>
This is an operation we have to do already for nir_vector_extract and
I'm about to do something very similar for matrix columns. Having a
more generic helper is useful.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>
Missing in this commit are NIR intrinsics for the ObjectToWorld and
WorldToObject built-ins. Those are matrices and so they take a bit more
work and justify a separate commit. For now, we add the enums and leave
the SYSTEM_VALUE <-> nir_intrinsic conversion commented out.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>
For now, we assume its a 64-bit global pointer.
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6479>
some drivers may have pre-lowered gl_ClipDistance to 2x vec4 to match hw
usage, so for those cases we'll be getting deref_var here and then components
will be stored to the deref at some point
fixesmesa/mesa#3480
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6563>
we can just check the bits using clip_distance_array_size here to simplify
everything and more easily determine if we need to be running this pass
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6563>
This also seems to be done by nir_opt_algebraic, but RADV will be moving
nir_lower_bit_size() to after that (so it doesn't create unsupported
8/16-bit instructions) and it doesn't seem worth creating a new pass just
for this simple optimization.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4791>
The shuffles provided by the SPV_INTEL_subgroups extension generate
bcsel(b, shuffle(x, i), shuffle(y, j))
In the case where x and y are the same, we can turn this into a shuffle
with the bcsel on the index which lets us drop a whole shuffle.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7366>
Typically, if we have one alu instruction, we call it "alu" and if we
have one intrinsic we call it "intrin".
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7366>
Fixes validation error in a Dota 2 shader.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Fixes: b031c64349 ("nir: Convert a bcsel with only phi node sources to a phi node")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7426>
Unlike most address formats, this address format is capable of handling
all of the fancy generic pointers stuff like is_global and friends.
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6332>
