copy_deref for wildcard dereferences requires the same
arrays lengths otherwise it leads to a crash in optimizations
like 'nir_opt_copy_prop_vars' because these optimizations expect
'copy_deref' just for arrays with the same lengths.
v2: check was moved to 'try_match_deref' to fix aoa cases
(Jason Ekstrand <jason@jlekstrand.net>)
v3: -fixed comment
-the condition merged with other one
(Jason Ekstrand <jason@jlekstrand.net>)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111286
Signed-off-by: Andrii Simiklit <andrii.simiklit@globallogic.com>
The OpenGL and OpenGL ES specs require that implementations clamp the
value of gl_PointSize to an implementation-depedent range. This pass
is useful for any GPU hardware that doesn't do this automatically
for either one or both sides of the range, such as V3D.
v2:
- Turn into a generic NIR pass (Eric).
- Make the pass work before lower I/O so we can use the deref variable
to inspect if we are writing to gl_PointSize (Eric).
- Make the pass take the range to clamp as parameter and allow it
to clamp to both sides of the range or just one side.
- Make the pass report progress.
v3:
- Fix copyright header (Eric)
- use fmin/fmax instead of bcsel to clamp (Eric)
Reviewed-by: Eric Anholt <eric@anholt.net>
v2: add to series
v3: update Makefile.sources
v4: don't remove a comment and break statement
v4: use nir_can_move_instr
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
This is mostly the same as nir_move_load_const() but can also move
undef instructions, comparisons and some intrinsics (being careful with
loops).
v2: actually delete nir_move_load_const.c
v3: fix nir_opt_sink() usage in freedreno
v3: update Makefile.sources
v4: replace get_move_def with nir_can_move_instr and nir_instr_ssa_def
v4: handle if uses
v4: fix handling of nested loops
v5: re-write adjust_block_for_loops
v5: re-write setting of use_block for if uses
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Co-authored-by: Daniel Schürmann <daniel@schuermann.dev>
Reviewed-by: Eric Anholt <eric@anholt.net>
v2: use glsl_type_size_align_func
v2: move get_explicit_type() to glsl_types.cpp/nir_types.cpp
v2: use align() instead of util_align_npot()
v2: pack arrays a bit tighter
v2: rename mem_* to field_*
v2: don't attempt to handle when struct offsets are already set
v2: use column_type() instead of recreating it
v2: use a branch instead of |= in nir_lower_to_explicit_impl()
v2: assign locations to variables and update shared_size and num_shared
v2: allow the pass to be used with nir_var_{shader_temp,function_temp}
v4: rebase
v5: add TODO
v5: small formatting changes
v5: remove incorrect assert in get_explicit_type()
v5: rename to nir_lower_vars_to_explicit_types
v5: correctly update progress when only variables are updated
v5: rename get_explicit_type() to get_explicit_shared_type()
v5: add comment explaining how get_explicit_shared_type() is different
v5: update cast strides
v6: update progress when lowering nir_var_function_temp variables
v6: formatting changes
v6: add more detailed documentation comment for get_explicit_shared_type
v6: rename get_explicit_shared_type to get_explicit_type_for_size_align
v7: fix comment in nir_lower_vars_to_explicit_types_impl()
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com> (v5)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
v2: require nir_address_format_32bit_offset instead
v3: don't call nir_intrinsic_set_access() for shared atomics
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
This just eliminates tautological / contradictory compares that are used
for bcsel and other non-if-statement cases. If-statements are not
affected because removing flow control can cause the i965 instrution
scheduler to create some very long live ranges resulting in unncessary
spilling. This causes some shaders to fall of a performance cliff.
Since many small if-statements are already flattened to bcsel, this
optimization covers more than 68% of the possible cases (2417 shaders
helped for instructions on Skylake vs. 3554).
v2: Reorder and add whitespace to make the relationship between the
patterns more obvious. Suggested by Caio.
All Gen7+ platforms had similar results. (Ice Lake shown)
total instructions in shared programs: 16333474 -> 16322028 (-0.07%)
instructions in affected programs: 438559 -> 427113 (-2.61%)
helped: 1765
HURT: 0
helped stats (abs) min: 1 max: 275 x̄: 6.48 x̃: 4
helped stats (rel) min: 0.20% max: 36.36% x̄: 4.07% x̃: 1.82%
95% mean confidence interval for instructions value: -6.87 -6.10
95% mean confidence interval for instructions %-change: -4.30% -3.84%
Instructions are helped.
total cycles in shared programs: 367608554 -> 367511103 (-0.03%)
cycles in affected programs: 8368829 -> 8271378 (-1.16%)
helped: 1541
HURT: 129
helped stats (abs) min: 1 max: 4468 x̄: 66.78 x̃: 39
helped stats (rel) min: 0.01% max: 45.69% x̄: 4.10% x̃: 2.17%
HURT stats (abs) min: 1 max: 973 x̄: 42.25 x̃: 10
HURT stats (rel) min: 0.02% max: 64.39% x̄: 2.15% x̃: 0.60%
95% mean confidence interval for cycles value: -64.90 -51.81
95% mean confidence interval for cycles %-change: -3.89% -3.36%
Cycles are helped.
total spills in shared programs: 8867 -> 8868 (0.01%)
spills in affected programs: 18 -> 19 (5.56%)
helped: 0
HURT: 1
total fills in shared programs: 21900 -> 21903 (0.01%)
fills in affected programs: 78 -> 81 (3.85%)
helped: 0
HURT: 1
All Gen6 and earlier platforms had similar results. (Sandy Bridge shown)
total instructions in shared programs: 10829877 -> 10829247 (<.01%)
instructions in affected programs: 30240 -> 29610 (-2.08%)
helped: 177
HURT: 0
helped stats (abs) min: 1 max: 15 x̄: 3.56 x̃: 3
helped stats (rel) min: 0.37% max: 17.39% x̄: 2.68% x̃: 1.94%
95% mean confidence interval for instructions value: -3.93 -3.18
95% mean confidence interval for instructions %-change: -3.04% -2.32%
Instructions are helped.
total cycles in shared programs: 154036580 -> 154035437 (<.01%)
cycles in affected programs: 352402 -> 351259 (-0.32%)
helped: 96
HURT: 28
helped stats (abs) min: 1 max: 128 x̄: 14.73 x̃: 6
helped stats (rel) min: 0.03% max: 24.00% x̄: 1.51% x̃: 0.46%
HURT stats (abs) min: 1 max: 117 x̄: 9.68 x̃: 4
HURT stats (rel) min: 0.03% max: 2.24% x̄: 0.43% x̃: 0.23%
95% mean confidence interval for cycles value: -13.40 -5.03
95% mean confidence interval for cycles %-change: -1.62% -0.53%
Cycles are helped.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
A similar technique could be used for fmin3, fmax3, and fmid3.
This could be squashed with the previous commit. I kept it separate to
ease review.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
This could be squashed with the previous commit. I kept it separate to
ease review.
v2: Add some missing cases. Use nir_src_is_const helper. Both
suggested by Caio. Use a table for mapping source ranges to a result
range.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
This could be squashed with the previous commit. I kept it separate to
ease review.
v2: Use a switch statement and add more comments. Both suggested by
Caio.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Most integer operations are omitted because dealing with integer
overflow is hard. There are a few things that could be smarter if there
was a small amount more tracking of ranges of integer types (i.e.,
operands are Boolean, operand values fit in 16 bits, etc.).
The changes to nir_search_helpers.h are included in this patch to
simplify reordering the changes to nir_opt_algebraic.py.
v2: Memoize range analysis results. Without this, some shaders appear
to get stuck in infinite loops.
v3: Rebase on many months of Mesa changes, including 1-bit Boolean
changes.
v4: Rebase on "nir: Drop imov/fmov in favor of one mov instruction".
v5: Use nir_alu_srcs_equal for detecting (a*a). Previously just the SSA
value was compared, and this incorrectly matched (a.x*a.y).
v6: Many code improvements including (but not limited to) better names,
more comments, and better use of helper functions. All suggested by
Caio. Rework the handling of several opcodes to use a table for mapping
source ranges to a result range. This change fixed a bug that caused
fmax(gt_zero, ge_zero) to be incorrectly recognized as ge_zero.
Slightly tighten the range of fmul by recognizing that x*x is gt_zero if
x is gt_zero. Add similar handling for -x*x.
v7: Use _______ in the tables as an alias for unknown. Suggested by
Caio.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
This automates the include_directories and dependencies tracking so that
all users of libmesa_util don't need to add them manually.
Next commit will remove the ones that were only added for that reason.
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Acked-by: Eric Anholt <eric@anholt.net>
Tested-by: Vinson Lee <vlee@freedesktop.org>
When we detect a scalar/vector copy through load_deref/store_deref, we
have to be careful since those can bitcast an int to a float and
vice-versa even though copy_deref can't.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111251
Fixes: 156306e5e6 ("nir/find_array_copies: Handle wildcards and overlapping copies")
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
These were left after a rebase and happen to make
NIR_INTRINSIC_SWIZZLE_MASK == NIR_INTRINSIC_SRC_ACCESS, which is how it
was noticed.
Fixes: 6f20643b47 ("nir: Allow qualifiers on copy_deref and image instructions")
Cc: Connor Abbott <cwabbott0@gmail.com>
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Optimizations that insert bitshift or bitwise operations should not be
applied on GPUs that don't support integer operations.
The .lower_bitshift could be used to control the bitshift related ones,
but there was also one bitwise optimization uncovered.
Since only lima and freedreno use this option and the use case is that
no bit operations are wanted, let's rename it to .lower_bitops and use
it to control all bitops related optimizations.
Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Jonathan Marek <jonathan@marek.ca>
The Mali400 pp doesn't implement fdot but has fsum3 and fsum4, which can
be used to optimize fdot lowering. fsum2 is not implemented and can be
further lowered to an add with the vector components.
Currently lima ppir handles this lowering internally, however this
happens in a very late stage and requires a big chunk of code compared
to a nir_opt_algebraic lowering.
By having fsum in nir, we can reduce ppir complexity and enable the
lowered ops to be part of other nir optimizations in the optimization
loop.
Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Suggested-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
MAYBE_UNUSED is going away, so let's replace legitimate uses of it with
UNUSED, which the former aliased to so far anyway.
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
instr->type is the type of the array element, not the type of the array
being dereferenced. Rather than fishing out the parent type, just use
parent->num_children which should be the length plus 1. While we're here
add another assert for the issue fixed by the previous commit.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111251
Fixes: 156306e5e6 ("nir/find_array_copies: Handle wildcards and overlapping copies")
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
There was an off-by-one error.
Fixes: 156306e5e6 ("nir/find_array_copies: Handle wildcards and overlapping copies")
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
It's kind-of an anomaly that the Intel drivers are still treating
gl_FragCoord as an input. It also makes zero sense because we have to
special-case it in the back-end.
Because ANV is the only user of nir_lower_wpos_center, we go ahead and
just update it to look for nir_intrinsic_load_frag_coord as part of this
patch.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Even if the data race wasn't real (I'm not great at reasoning about
this), helgrind is a nice enough tool that keeping noise out of it is
probably worthwhile. Besides, typing out the numbers keeps the data
in the read-only data section instead of emitting code to initialize
it every time.
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
This commit rewrites opt_find_array_copies to be able to handle
an array copy sequence with other intervening operations in between. In
particular, this handles the case where we OpLoad an array of structs
and then OpStore it, which generates code like:
foo[0].a = bar[0].a
foo[0].b = bar[0].b
foo[1].a = bar[1].a
foo[1].b = bar[1].b
...
that wasn't recognized by the previous pass.
In order to correctly handle copying arrays of arrays, and in particular
to correctly handle copies involving wildcards, we need to use a tree
structure similar to lower_vars_to_ssa so that we can walk all the
partial array copies invalidated by a particular write, including
ones where one of the common indices is a wildcard. I actually think
that when factoring in the needed hashing/comparing code, a hash table
based approach wouldn't be a lot smaller anyways.
All of the changes come from tessellation control shaders in Strange
Brigade, where we're able to remove the DXVK-inserted copy at the
beginning of the shader. These are the result for radv:
Totals from affected shaders:
SGPRS: 4576 -> 4576 (0.00 %)
VGPRS: 13784 -> 5560 (-59.66 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Private memory VGPRs: 0 -> 0 (0.00 %)
Scratch size: 8696 -> 6876 (-20.93 %) dwords per thread
Code Size: 329940 -> 263268 (-20.21 %) bytes
LDS: 0 -> 0 (0.00 %) blocks
Max Waves: 330 -> 898 (172.12 %)
Wait states: 0 -> 0 (0.00 %)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
We don't have calculate final quotient in order to calculate unsigned
modulo result. Once we are done with error correction we have partial
result which can be used to find out modulo operation result
Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
SPIRV added the ability to access variables and have expressions non
dynamically uniform and because spirv_to_nir generates deref
instructions, we'll need to have that access there.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Cc: <mesa-stable@lists.freedesktop.org>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
When 'x' is the result of a scmp op:
x != 0.0 or x == 1.0: passthrough
x == 0.0 or x != 1.0: invert
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Add generic lowerings for fall_equalN/fany_nequalN. These should be optimal
for vec4 backends that doesn't have any special instructions for it, as
long as they support saturate.
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Add simple fdot2 optimizations that are missing.
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Thomas Helland <thomashelland90@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
For backends that don't have a 'fdph' instructions
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Thomas Helland <thomashelland90@gmail.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
This version has less ops for the same precision.
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Vasily Khoruzhick <anarsoul@gmail.com>
Acked-by: Matt Turner <mattst88@gmail.com>
This is to allow optimizations in nir_opt_algebraic not otherwise possible
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Acked-by: Matt Turner <mattst88@gmail.com>
This will effectively enable the optimization in anv.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Instead of building a constant mask (which depends on knowing the
subgroup size), we build an expression. Because the pass uses the
nir_shader_lower_instructions helper, subgroup lowering will be run on
any newly emitted instructions as well as the previously existing
instructions. In particular, if the subgroup size is known, the newly
emitted subgroup_size intrinsic will get turned into a constant and a
later constant folding pass will clean it up.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Tested on Gen > 9.
v2: 1) Fix lowering
2) Keep a consistent i/u order (Matt Turner)
Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
I can't find a single place where nir_lower_io is called after going out
of SSA which is the only real reason why you wouldn't do this. Returning
SSA defs is more idiomatic and is required for the next commit.
Reviewed-by: Matt Turner <mattst88@gmail.com>
