Compile-time (nir_opt_dce):
Difference at 95.0% confidence
-319.51 +/- 5.67632
-12.0627% +/- 0.208076%
(Student's t, pooled s = 6.70399)
Compile-time (overall):
Difference at 95.0% confidence
-385.025 +/- 42.1124
-0.929489% +/- 0.10139%
(Student's t, pooled s = 49.7367)
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Acked-by: Daniel Schürmann <daniel@schuermann.dev>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7691>
Basically every pass in NIR uses nir_ssa_def_rewrite_uses which calls
nir_instr_rewrite_src which is fairly complex because it handles all
sorts of non-SSA cases. Since we already know a priori that every
source written by nir_ssa_def_rewrite_uses is SSA, we can check new_src
once at the top of the function and cut out all that complexity.
While we're at it, we expose a new SSA-only nir_ssa_def_rewrite_uses_ssa
helper which takes an SSA def which avoids the one SSA check. It's also
more convenient 90% of the time.
Compile time as tested by Rhys Perry <pendingchaos02@gmail.com>
Difference at 95.0% confidence
-797.166 +/- 418.649
-0.566174% +/- 0.296441%
(Student's t, pooled s = 325.459)
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8790>
The function is very convenient for lowering any type of instruction
that can be easily filtered, but so far instructions that didn't return
a value were siletly ignored.
Fix this by
- not requiring a return value in the instruction
- add a new special return value from the lowering implementation
function to indicated that an instruction that doesn't have a
return value must be removed, and
- don't try to collect and replace uses in this case.
Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8177>
With the block's end_ip accidentally being the ip of the next instruction,
contrary to the comment, you would end up doing end-of-block freeing early
and have the value missing when it came time to emit the next instruction.
Just expand the ips to have separate ones for start and end of block --
while it means that nir_instr->index is no longer incremented by 1 per
instruction, it makes sense for use in liveness because a backend is
likely to need to do other things at block boundaries (like emit the if
statement's code), and having an ip to identify that stuff is useful.
Fixes: a206b58157 ("nir: Add a block start/end ip to live instr index metadata.")
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7658>
v2: Fixup comment about bits in nir_intrinsics.py
v3: Use varying for primitive shading rate builtin (samuel)
v4: Reoder switch alphabetically
Make divergence of frag_shading_rate an option
v5: Remove stage check for frag_shading_rate in divergence (Samuel)
v6: s/frag_shading_rate_per_subgroup/single_frag_shading_rate_per_subgroup/ (Jason)
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7795>
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>
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>
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>
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>
I wanted it for the per-instruction live intervals metadata, and it's not
much to store in general. Make the ip explicitly 32-bit, on suggestion by
jekstrand.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3395>
live_index had two things going on: 0 meant the instr was an undef and
always dead, and otherwise ssa defs had increasing numbers by instruction
order. We already have a field in the instruction for storing instruction
order, and ssa defs don't need that number to be contiguous (if you want a
compact per-ssa-def number, use ssa->index after reindexing).
We don't use ssa->index for this, because reindexing those would change
nir_print, and that would be rude to people trying to track what's
happening in optimization passes.
This openend up a hole in nir_ssa_def, so we move nir_ssa_def->index
toward the end to shrink the struct from 64 bytes to 56.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3395>
These are based on the ones which already existed in the load/store
vectorization pass but I made some improvements while moving them. In
particular,
1. They're both faster if the bit sizes are equal
2. The check is faster if old_bit_size > new_bit_size
3. The check now fails if it would use more than NIR_MAX_VEC_COMPONENTS
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6871>
We're about to introduce conversion ops which are going to want two
different types. We may as well just split the one we have rather than
end up with three. There are a couple places where this is mildly
inconvenient but most of the time I find it to actually be nicer.
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6945>
In nir-to-tgsi, I want to free temps storing SSA values when they go dead,
and NIR liveness has most of the information I need. Hoever, when I reach
the end of a block, I need to free whatever temps were in liveout which
are dead at that point. If liveout is indexed by live_index, then I don't
know the maximum live_index for iterating the live_out bitset, and I also
don't have a way to map that index back to the def->index that my temps
are stored under.
We can use the more typical def->index for these bitsets, which resolves
both of those problems. The only cost is that ssa_undefs don't get merged
into a single bit in the bitfield, but there are generally 1-4 of them in
a shader and we don't track liveness for those anyway so splitting them
apart is fine.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6408>
New intrinsics are added for global invocation IDs and work group IDs to
deal with offsets in both. The only one of these that needs a system value
is global invocation offset, for CL's get_global_offset().
Note that CL requires very large work group sizes, so these intrinsics
are modified to be able to use 64bit values, for 64bit SPIR-V.
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5891>
Previously, this was left uninitialized. Let's initialize it to an
obviously bogus value so we notice if anyone ever tries to use stale
liveness data.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Jesse Natalie <jenatali@microsoft.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6210>
These are safe to call on either structured or unstructured NIR but
don't provide the nice ordering guarantees of nir_foreach_block and
friends. While we're here, we use them for a very small selection of
passes which are known to be safe for unstructured control-flow. The
most important such pass is nir_dominance which is required for
structurizing.
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>
v2 (Jason Ekstrand):
- Make "structured" a property of nir_function_impl not nir_shader
- More validation and asserts
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/2401>
We've hand-rolled this loop 10 places and those are just the ones I
found easily.
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5966>
Instead of having separate lists of variables, roughly sorted by mode,
use a single list for all shader-level NIR variables. This makes a few
list walks a bit longer here and there but list walks aren't a very
common thing in NIR at all. On the other hand, it makes a lot of things
like validation, printing, etc. way simpler. Also, there are a number
of cases where we move variables from inputs/outputs to globals and this
makes it way easier because we no longer have to move them between
lists. We only have to deal with that if moving them from the shader to
a nir_function_impl.
Reviewed-by: Rob Clark <robdclark@chromium.org>
Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Reviewed-by: Vasily Khoruzhick <anarsoul@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5966>
We also very slightly change the semantics. It no longer is one index
per list for global variables and is a single index over-all.
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5966>
We also add a new list iterator which takes a modes bitfield and
automatically figures out which list to use. In the future, this
iterator will work for multiple modes but today it assumes a single mode
thanks to the behavior of nir_variable_list_for_mode. This also doesn't
work for function_temp variables.
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5966>
The line coord is a coordinate along the axis perpendicular to the line.
It is in the range [0,1] between the two edges of the line. It is
available at least on Broadcom hardware.
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5624>
Previously all nir_intrinsic_load_uniform that were used as sources were
considered to be dynamically_uniform but when offsets of load_uniform
are indirect it can not be determined.
This fixes artefacts in Google Maps 3D view in V3D.
Fixes: 886d46b089 ("nir: Add a function to determine if a source is dynamically uniform")
Reviewed-by: Neil Roberts <nroberts@igalia.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5587>
This takes the same callback as nir_foreach_src except it walks all phi
sources which leave a given block.
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Rob Clark <robdclark@chromium.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5094>
This is a query that both Intel and freedreno need to do. We can simplify
it a lot with the new glsl_get_sampler_dim_coordinate_components()
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3728>
These will be useful in the upcoming load/store vectorizer.
v11: rebase
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Rhys Perry
Jason Ekstrand
Arcady Goldmints-Orlov
Gert Wollny
Eric Anholt
Pierre-Eric Pelloux-Prayer
Lionel Landwerlin
Jason Ekstrand
Marek Olšák
Jesse Natalie
Karol Herbst
Neil Roberts
Timothy Arceri
Jose Maria Casanova Crespo