State parameters are sometimes not perfectly sorted.
This optimizes the number of iterations we have to do for fetch_state.
Reviewed-by: Zoltán Böszörményi <zboszor@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8183>
This reduces the constant buffer size by eliminating unused elements
because it's no longer a uniform array that the compiler can't split.
This looks silly, but there is no other way because all elements must be
globally declared, which means they can't be generated by a loop.
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8183>
Flattening continue to get optimal code in fetch_state.
This merges the "face" field with the "attrib" field using the combined
MAT_ATTRIB_* enums. The outcome is that the inner switch statements can
be flatten because we can use MAT_ATTRIB_* to index into the attrib array
directly.
With LightSource attributes that don't have two sides, more math is
involved to get the correct index but it works out nicely too.
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8183>
Let's flatten the tokens to generate optimal code for fetch_state.
There was only one name conflict: STATE_NORMAL_SCALE was used both as
internal and non-internal.
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8183>
This limits the exposure of these functions to when the extension is
available. Prevents crashes otherwise, as the rest of the infrastructure
doesn't necessarily expect these functions when the extension is not
available.
Fixes: 40c1f9883e ("mesa,glsl: add support for GL_NV_shader_atomic_int64")
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8533>
This patch also replaces lower_negate with lower_ineg / lower_fneg.
The fneg semantics have been clarified as of Version 1.5, Revision 1
of the SPIR-V specification, which means that the previous lowering
to fsub is not a viable solution anymore, and is replaced with
lowering to fmul(x, -1.0).
Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6597>
If the instruction being coalesced would be vectorized but the target
doesn't support vectorizing that op, skip coalescing.
Reuse the callbacks from alu_to_scalar to describe which ops should not
be vectorized.
Signed-off-by: Erico Nunes <nunes.erico@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Eric Anholt <eric@anholt.net>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6506>
To match hash_alu_src(), this should return false if both are different
non-const ssa defs.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8391>
Cleaned up a bit of addressing math in the shader I just had to debug.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Gert Wollny <gert.wollny@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8373>
Not sure why I thought this was correct, but we should consider them for
optimization purposes.
Fixes: ce9205c03b ('nir: add a load/store vectorization pass')
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4202>
Like SPIR-V and GL_ARB_sparse_texture2, these return a residency code. It
is placed in the destination after the rest of the result. If it's zero,
then the texel is resident. Otherwise, it's not resident.
Besides the larger destination and the residency code, sparse fetches
work the same as normal fetches.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7774>
These will be useful for sparse texture instructions and image load
intrinsics.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7774>
The mantissa for a float doesn't contain enough data to accurately represent
the min/max values for some destination types. Instead of clamping before
converting, clamp after converting when coming from floats. This improves
conformance of CL conversions, specifically for float -> long/ulong with
int64 emulation enabled.
Refactors the limit determination from the clamp, so we can determine
limits for the dest type (int/uint) in both the source (float) and dest
type. The limit as a float is used for comparison, while the limit as a
dest type is used for bcsel.
Important note is that the comparison is inverted to fge instead of flt,
so the bcsel chooses the direct int/uint over the converted float in the
case where the comparison comes up equal, but the conversion can't produce
the exact min/max value.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8256>
* int64 is a core type on Haiku (and potentially other platforms)
* rename to int64_avail matching other similar calls
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
If multiple rules could match, the rule that appears first in the file
is used.
Only Tiger Lake and Ice Lake are affected. Other platforms either have
a LRP instruction or can't run any shaders from shader-db that would
benefit.
v2: Fix issues created when this commit was rebased on top of
3c8934a644 ("nir/algebraic: add flrp patterns for 16 and 64 bits").
Noticed by Caio.
Tiger Lake and Ice Lake had similar results.
total instructions in shared programs: 20908672 -> 20908661 (<.01%)
instructions in affected programs: 419 -> 408 (-2.63%)
helped: 5
HURT: 0
helped stats (abs) min: 1 max: 3 x̄: 2.20 x̃: 3
helped stats (rel) min: 1.85% max: 3.19% x̄: 2.49% x̃: 2.65%
95% mean confidence interval for instructions value: -3.56 -0.84
95% mean confidence interval for instructions %-change: -3.24% -1.73%
Instructions are helped.
total cycles in shared programs: 473513940 -> 473513793 (<.01%)
cycles in affected programs: 7176 -> 7029 (-2.05%)
helped: 12
HURT: 0
helped stats (abs) min: 5 max: 22 x̄: 12.25 x̃: 12
helped stats (rel) min: 0.84% max: 3.24% x̄: 2.09% x̃: 1.80%
95% mean confidence interval for cycles value: -15.43 -9.07
95% mean confidence interval for cycles %-change: -2.57% -1.61%
Cycles are helped.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358>
This prevents other transformations from converting them to 'a != 0'.
For example, both of these transformations can do this:
(('~flt', 0.0, ('fabs', a)), ('fne', a, 0.0)),
(('~flt', ('fneg', ('fabs', a)), 0.0), ('fne', a, 0.0)),
Both fsign(fabs(NaN)) and fsign(fneg(fabs(NaN))) should produce zero,
but, since 'NaN != 0.0' is true, cascading these transformations could
cause them to generate 1.0 or -1.0 respecively.
No shader-db or fossil-db changes on any Intel platform.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358>
No shader-db or fossil-db changes on any Intel platform.
v2: Add a coding line to fix SCons build problems caused by the ±
character.
Fixes: 25bfba3335 ("nir/algebraic: Recognize open-coded copysign(1.0, a)")
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358>
OpenGL GLSL, OpenGL ARB assembly shaders, and DX9 are pretty loose about
the behavior in the presence of NaNs. Many GPUs that implement these
specifications do not even have a representation of NaN. However,
OpenCL and Vulkan SPIR-V are not so lax. Both actually have some
required behavior in the presence of NaN, and, of the two, OpenCL is the
most strict.
For years we have implemented SPIR-V by using the same comparison
opcodes as we use for OpenGL GLSL and OpenGL assembly shaders. This has
repeatedly caused problems where an optimization that is valid in the
NaN-relaxed world is not valid in Vulkan or OpenCL. To fix this, set
the "exact" flag on comparisons instructions generated from SPIR-V.
This will block optimizations that may have different NaN behavior.
v2: Set the exact flag in the nir_builder, not in the vtn_builder.
v3: Add an assertion in vtn_handle_constant that the exact flag wasn't
set (because it's ignored). Rebase on 80163bbec3 ("nir/vtn: Support
OpOrdered and OpUnordered opcodes"). Mark the NIR generated for those
opcodes as exact as well.
v4: s/unused_exact/exact/ in a couple places, and assert that exact has
the expected value (true in one place, false in the other). Suggested
by Caio.
Closes: #3345
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Tested-by: Danylo Piliaiev <danylo.piliaiev@globallogic.com>
Fixes: 8513b12590 ("nir/opt_if: split ALU from Phi more aggressively")
This commit doesn't really fix anything in 8513b12590. However,
without 8513b12590, a regression is triggered in RADV on No Man's
Sky. I want to ensure that this change is only applied on top of
8513b12590, and Fixes: seems the safest way to do that.
No shader-db changes on any Intel platform. This only affects SPIR-V,
and we have no OpenGL SPIR-V shaders in shader-db.
124 shaders in Shadow of the Tomb Raider (Steam "native") were hurt by 1
spill and 1 fill each.
All Intel platforms had similar results. (Tiger Lake shown)
Instructions in all programs: 155668276 -> 155685764 (+0.0%)
SENDs in all programs: 6474570 -> 6474570 (+0.0%)
Loops in all programs: 35271 -> 35271 (+0.0%)
Cycles in all programs: 3198055373 -> 3198628031 (+0.0%)
Spills in all programs: 231522 -> 231646 (+0.1%)
Fills in all programs: 347571 -> 347695 (+0.0%)
Vega
Totals:
SGPRs: 20955712 -> 20956756 (+0.00%); split: -0.02%, +0.03%
VGPRs: 13476920 -> 13473132 (-0.03%); split: -0.07%, +0.04%
CodeSize: 613371940 -> 613339348 (-0.01%); split: -0.06%, +0.05%
MaxWaves: 3111886 -> 3112481 (+0.02%); split: +0.02%, -0.00%
Instrs: 120723785 -> 120746991 (+0.02%); split: -0.04%, +0.06%
Cycles: 626658992 -> 626862708 (+0.03%); split: -0.05%, +0.08%
VMEM: 216330854 -> 216343196 (+0.01%); split: +0.04%, -0.04%
SMEM: 32079391 -> 32081972 (+0.01%); split: +0.05%, -0.04%
VClause: 2688784 -> 2688789 (+0.00%); split: -0.03%, +0.03%
SClause: 6554669 -> 6556251 (+0.02%); split: -0.01%, +0.03%
Copies: 5356667 -> 5353283 (-0.06%); split: -0.36%, +0.29%
Branches: 954466 -> 954716 (+0.03%); split: -0.01%, +0.04%
PreSGPRs: 9078300 -> 9081626 (+0.04%); split: -0.01%, +0.05%
PreVGPRs: 10972090 -> 10966576 (-0.05%); split: -0.06%, +0.01%
Totals from 48239 (12.08% of 399432) affected shaders:
SGPRs: 2713984 -> 2715028 (+0.04%); split: -0.16%, +0.19%
VGPRs: 1997804 -> 1994016 (-0.19%); split: -0.46%, +0.27%
CodeSize: 172094092 -> 172061500 (-0.02%); split: -0.21%, +0.19%
MaxWaves: 337327 -> 337922 (+0.18%); split: +0.20%, -0.02%
Instrs: 33053657 -> 33076863 (+0.07%); split: -0.15%, +0.22%
Cycles: 254961228 -> 255164944 (+0.08%); split: -0.12%, +0.20%
VMEM: 15165226 -> 15177568 (+0.08%); split: +0.59%, -0.51%
SMEM: 3304938 -> 3307519 (+0.08%); split: +0.49%, -0.41%
VClause: 766225 -> 766230 (+0.00%); split: -0.12%, +0.12%
SClause: 1332645 -> 1334227 (+0.12%); split: -0.04%, +0.16%
Copies: 2040651 -> 2037267 (-0.17%); split: -0.94%, +0.77%
Branches: 743668 -> 743918 (+0.03%); split: -0.01%, +0.05%
PreSGPRs: 1697667 -> 1700993 (+0.20%); split: -0.07%, +0.27%
PreVGPRs: 1718424 -> 1712910 (-0.32%); split: -0.39%, +0.07%
Polaris
Totals:
SGPRs: 21349172 -> 21354376 (+0.02%); split: -0.02%, +0.04%
VGPRs: 13690680 -> 13686920 (-0.03%); split: -0.07%, +0.04%
CodeSize: 613745824 -> 613704988 (-0.01%); split: -0.06%, +0.05%
MaxWaves: 2775012 -> 2775189 (+0.01%); split: +0.01%, -0.00%
Instrs: 120735079 -> 120756209 (+0.02%); split: -0.04%, +0.06%
Cycles: 627906100 -> 628076156 (+0.03%); split: -0.05%, +0.08%
VMEM: 216623065 -> 216641838 (+0.01%); split: +0.04%, -0.04%
SMEM: 32295618 -> 32299338 (+0.01%); split: +0.05%, -0.04%
VClause: 2711025 -> 2711141 (+0.00%); split: -0.03%, +0.04%
SClause: 6545185 -> 6546769 (+0.02%); split: -0.01%, +0.03%
Copies: 5387723 -> 5383249 (-0.08%); split: -0.37%, +0.29%
Branches: 953775 -> 953954 (+0.02%); split: -0.01%, +0.03%
PreSGPRs: 9148814 -> 9153211 (+0.05%); split: -0.01%, +0.06%
PreVGPRs: 11029429 -> 11023915 (-0.05%); split: -0.06%, +0.01%
Totals from 48239 (12.00% of 402052) affected shaders:
SGPRs: 2682056 -> 2687260 (+0.19%); split: -0.16%, +0.35%
VGPRs: 1994436 -> 1990676 (-0.19%); split: -0.46%, +0.27%
CodeSize: 170857060 -> 170816224 (-0.02%); split: -0.21%, +0.19%
MaxWaves: 295429 -> 295606 (+0.06%); split: +0.07%, -0.01%
Instrs: 32808802 -> 32829932 (+0.06%); split: -0.16%, +0.22%
Cycles: 254633252 -> 254803308 (+0.07%); split: -0.13%, +0.20%
VMEM: 14897934 -> 14916707 (+0.13%); split: +0.65%, -0.52%
SMEM: 3289726 -> 3293446 (+0.11%); split: +0.53%, -0.42%
VClause: 775318 -> 775434 (+0.01%); split: -0.11%, +0.13%
SClause: 1304867 -> 1306451 (+0.12%); split: -0.04%, +0.16%
Copies: 2026334 -> 2021860 (-0.22%); split: -0.99%, +0.77%
Branches: 742554 -> 742733 (+0.02%); split: -0.02%, +0.04%
PreSGPRs: 1690887 -> 1695284 (+0.26%); split: -0.07%, +0.33%
PreVGPRs: 1717709 -> 1712195 (-0.32%); split: -0.40%, +0.07%
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358>
This prevents some fossil-db regressions in "spir-v: Mark floating point
comparisons exact".
v2: Note that the patterns and replacements produce the same value when
isnan(b). Suggested by Caio.
v3: Use C99 isfinite() instead of (obsolete) BSD finite(). Fixes
various Windows builds.
No fossil-db changes on any Inetl platform, Vega, or Polaris10.
All Intel platforms had similar results. (Tiger Lake shown)
total instructions in shared programs: 20908670 -> 20908672 (<.01%)
instructions in affected programs: 69 -> 71 (2.90%)
helped: 0
HURT: 1
total cycles in shared programs: 473515288 -> 473513940 (<.01%)
cycles in affected programs: 4942 -> 3594 (-27.28%)
helped: 2
HURT: 0
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358>
This also prevents some fossil-db regressions in "spir-v: Mark floating
point comparisons exact".
v2: Mark the fmin / fmax in the replacement exact to prevent other
optimizations from ruining the NaN-clensing property of the fmin / fmax.
Suggested by Rhys. Don't assume that constants are not NaN because some
components of a vector might be NaN while others are numbers. Noticed
by Rhys. This causes ~8 more shaders in Age of Wonders III (dxvk) to
regress on cycles (not instructions) by less than 1% when "spir-v: Mark
floating point comparisons exact" is applied. This difference is too
small to care.
All Intel platforms had similar results. (Tiger Lake shown)
total instructions in shared programs: 20908668 -> 20908670 (<.01%)
instructions in affected programs: 9196 -> 9198 (0.02%)
helped: 10
HURT: 5
helped stats (abs) min: 1 max: 2 x̄: 1.40 x̃: 1
helped stats (rel) min: 0.02% max: 5.41% x̄: 2.20% x̃: 2.16%
HURT stats (abs) min: 2 max: 6 x̄: 3.20 x̃: 3
HURT stats (rel) min: 2.44% max: 16.67% x̄: 9.39% x̃: 12.50%
95% mean confidence interval for instructions value: -1.22 1.49
95% mean confidence interval for instructions %-change: -2.08% 5.41%
Inconclusive result (value mean confidence interval includes 0).
total cycles in shared programs: 473515330 -> 473515288 (<.01%)
cycles in affected programs: 67146 -> 67104 (-0.06%)
helped: 10
HURT: 7
helped stats (abs) min: 1 max: 36 x̄: 15.90 x̃: 17
helped stats (rel) min: 0.01% max: 1.29% x̄: 0.66% x̃: 0.89%
HURT stats (abs) min: 1 max: 48 x̄: 16.71 x̃: 4
HURT stats (rel) min: 0.08% max: 1.94% x̄: 0.87% x̃: 0.19%
95% mean confidence interval for cycles value: -13.88 8.94
95% mean confidence interval for cycles %-change: -0.56% 0.49%
Inconclusive result (value mean confidence interval includes 0).
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358>
GLSL and SPIR-V GLSL.std.450 don't have any requirements for fsign(NaN),
and both only require that FSign(-0.0) == 0.0. OpenCL, on the other
hand, requires sign(-0.0) be exactly -0.0. It also requires that
sign(NaN) be exactly 0.0.
In practice, this change is difficult to test. Our GLSL frontend
already constant folds sign(NaN) to 0.0 before even getting to NIR. As
far as I can tell, glslang does the same. I don't have a good way to
run an OpenCL SPIR-V test. Maybe SPIR-V GLSL.std.450 assembly?
No shader-db or fossil-db changes on any Intel platform.
Acked-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358>
The reordered are automatically considered by nir_algebraic rules for
commutative instructions.
No shader-db or fossil-db changes on any Intel platform.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358>
I originally noticed that 3b30814791 ("nir/algebraic: Optimize 1-bit
Booleans") caused this pattern no longer be matched by incorrectly
replacing b@32 with b@1. Making that correct had no effect on
shader-db. When this pattern originally was added, it only affected 4
shaders, so it's not worth the effort to debug further.
This reverts commit f50400cc80.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358>
The original comment was a little terse and a little incorrect. The
rearrangements are fine w.r.t. NaN. However, they produce incorrect
results if one operand is +Inf and the other is -Inf.
A later commit, "nir/algebraic: Add some compare-with-zero optimizations
that are exact", will add some more patterns here. It may be reasonable
to squash this commit (forward) into that commit.
v2: Fix some incorrect comparisons operators in the comment (<= vs >=).
Add commentary that subtraction works like addition w.r.t. NaN. Both
noticed / suggested by Caio.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358>
This commit only documents the current behavior, even if that behavior
is not the behavior preferred by the relevant specs.
In SPIR-V, there are two flavors of the sign instruction, and each lives
in an extended instruction set. The GLSL.std.450 FSign instruction is
defined as:
Result is 1.0 if x > 0, 0.0 if x = 0, or -1.0 if x < 0.
This also matches the GLSL 4.60 definition.
However, the OpenCL.ExtendedInstructionSet.100 sign instruction is
defined as:
Returns 1.0 if x > 0, -0.0 if x = -0.0, +0.0 if x = +0.0, or -1.0 if
x < 0. Returns 0.0 if x is a NaN.
There are two differences. Each treats -0.0 differently, and each also
treats NaN differently. Specifically, GLSL.std.450 FSign does not
define any specific behavior for NaN.
There has been some discussion in Khronos about the NaN behavior of
GLSL.std.450 FSign. As part of that discussion, I did some research
into how we treat NaN for nir_op_fsign, and this commit just captures
some of those notes.
v2: Document the expected behavior of nir_op_fsign more thoroughly.
Suggested by Rhys. Note that the current implementation of constant
folding does not produce the expected result for NaN. Suggested by
Caio.
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com> [v1]
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358>
