fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2025-12-21 22:20:14 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions 1
121336 commits 64 branches 914 tags 1 GiB
Commit graph

265 commits

Author SHA1 Message Date
Ian Romanick
4e3d69ad07 nir/algebraic: Simplify a contradiction that can occur in __flt64_nonnan 
The pattern is added to opt_algebraic because, for example, comparisons
with constant 0.0 will produce (a1 < 0).

Even with a pass that optimized Boolean expressions, I think this would
be very difficult to automatically recognize and optimize.

Results on the 308 shaders extracted from the fp64 portion of the OpenGL
CTS:

Tiger Lake and Ice Lake had similar results. (Tiger Lake shown)
total instructions in shared programs: 933054 -> 929619 (-0.37%)
instructions in affected programs: 784041 -> 780606 (-0.44%)
helped: 59
HURT: 0
helped stats (abs) min: 2 max: 213 x̄: 58.22 x̃: 44
helped stats (rel) min: 0.02% max: 2.51% x̄: 0.72% x̃: 0.46%
95% mean confidence interval for instructions value: -70.80 -45.64
95% mean confidence interval for instructions %-change: -0.92% -0.53%
Instructions are helped.

total cycles in shared programs: 7304712 -> 7280180 (-0.34%)
cycles in affected programs: 7176260 -> 7151728 (-0.34%)
helped: 92
HURT: 0
helped stats (abs) min: 8 max: 1414 x̄: 266.65 x̃: 166
helped stats (rel) min: 0.04% max: 2.34% x̄: 0.43% x̃: 0.22%
95% mean confidence interval for cycles value: -333.05 -200.26
95% mean confidence interval for cycles %-change: -0.54% -0.31%
Cycles are helped.

Regular shader-db changes:

No changes on any Intel platform.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4142>
 2020-03-18 20:36:29 +00:00
Ian Romanick
e0cefc5a23 nir/algebraic: Constant reassociation for bitwise operations too 
Like 5886cd79a0, but for iand, ior, and ixor.

Results on the 308 shaders extracted from the fp64 portion of the OpenGL
CTS:

Tiger Lake
total instructions in shared programs: 903108 -> 902830 (-0.03%)
instructions in affected programs: 654910 -> 654632 (-0.04%)
helped: 31
HURT: 5
helped stats (abs) min: 2 max: 31 x̄: 9.58 x̃: 7
helped stats (rel) min: 0.01% max: 0.23% x̄: 0.06% x̃: 0.04%
HURT stats (abs)   min: 1 max: 10 x̄: 3.80 x̃: 3
HURT stats (rel)   min: 0.01% max: 0.10% x̄: 0.03% x̃: 0.02%
95% mean confidence interval for instructions value: -10.55 -4.89
95% mean confidence interval for instructions %-change: -0.07% -0.03%
Instructions are helped.

total cycles in shared programs: 7059681 -> 7058006 (-0.02%)
cycles in affected programs: 5081309 -> 5079634 (-0.03%)
helped: 33
HURT: 12
helped stats (abs) min: 1 max: 444 x̄: 60.91 x̃: 18
helped stats (rel) min: <.01% max: 2.17% x̄: 0.25% x̃: 0.05%
HURT stats (abs)   min: 1 max: 288 x̄: 27.92 x̃: 2
HURT stats (rel)   min: <.01% max: 1.00% x̄: 0.23% x̃: 0.02%
95% mean confidence interval for cycles value: -68.32 -6.12
95% mean confidence interval for cycles %-change: -0.28% 0.03%
Inconclusive result (%-change mean confidence interval includes 0).

Ice Lake
total instructions in shared programs: 895384 -> 895159 (-0.03%)
instructions in affected programs: 658678 -> 658453 (-0.03%)
helped: 37
HURT: 0
helped stats (abs) min: 3 max: 16 x̄: 6.08 x̃: 4
helped stats (rel) min: <.01% max: 0.07% x̄: 0.04% x̃: 0.04%
95% mean confidence interval for instructions value: -7.46 -4.70
95% mean confidence interval for instructions %-change: -0.04% -0.03%
Instructions are helped.

total cycles in shared programs: 7092224 -> 7091195 (-0.01%)
cycles in affected programs: 5221666 -> 5220637 (-0.02%)
helped: 35
HURT: 11
helped stats (abs) min: 1 max: 247 x̄: 43.46 x̃: 12
helped stats (rel) min: <.01% max: 2.17% x̄: 0.23% x̃: 0.05%
HURT stats (abs)   min: 2 max: 432 x̄: 44.73 x̃: 5
HURT stats (rel)   min: <.01% max: 1.00% x̄: 0.25% x̃: 0.02%
95% mean confidence interval for cycles value: -49.00 4.26
95% mean confidence interval for cycles %-change: -0.27% 0.03%
Inconclusive result (value mean confidence interval includes 0).

Regular shader-db results:

All Haswell+ platforms had similar results. (Tiger Lake shown)
total instructions in shared programs: 17611408 -> 17611398 (<.01%)
instructions in affected programs: 1648 -> 1638 (-0.61%)
helped: 2
HURT: 0

total cycles in shared programs: 338366148 -> 338366124 (<.01%)
cycles in affected programs: 124048 -> 124024 (-0.02%)
helped: 2
HURT: 0

No changes on any earlier Intel platforms.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4142>
 2020-03-18 20:36:29 +00:00
Ian Romanick
1d36af9338 nir/algebraic: Generalize some and-of-shift-right patterns [v2] 
Generalizes some of the patterns from 76289fbfa8 and 905ff86198.  In
particular, some of the soft-fp64 code generates (a & 0x7fffffff) << 1
when constant 0.0 is compared (flt or feq).

v2: Reduce the set of added patterns to those that actually help
something.  This reduces the size of the state transition tables by
about 29k.  Suggested by Jason.  Remove the existing patterns that this
commit subsumes.

Results on the 308 shaders extracted from the fp64 portion of the OpenGL
CTS:

Tiger Lake
total instructions in shared programs: 903171 -> 903108 (<.01%)
instructions in affected programs: 635903 -> 635840 (<.01%)
helped: 25
HURT: 11
helped stats (abs) min: 1 max: 16 x̄: 5.04 x̃: 3
helped stats (rel) min: <.01% max: 0.15% x̄: 0.04% x̃: 0.03%
HURT stats (abs)   min: 2 max: 14 x̄: 5.73 x̃: 5
HURT stats (rel)   min: <.01% max: 0.11% x̄: 0.04% x̃: 0.02%
95% mean confidence interval for instructions value: -3.91 0.41
95% mean confidence interval for instructions %-change: -0.03% <.01%
Inconclusive result (value mean confidence interval includes 0).

total cycles in shared programs: 7059527 -> 7059681 (<.01%)
cycles in affected programs: 5249401 -> 5249555 (<.01%)
helped: 41
HURT: 9
helped stats (abs) min: 2 max: 76 x̄: 11.90 x̃: 10
helped stats (rel) min: <.01% max: 11.86% x̄: 0.99% x̃: 0.01%
HURT stats (abs)   min: 2 max: 380 x̄: 71.33 x̃: 12
HURT stats (rel)   min: <.01% max: 0.22% x̄: 0.04% x̃: 0.01%
95% mean confidence interval for cycles value: -14.93 21.09
95% mean confidence interval for cycles %-change: -1.40% -0.20%
Inconclusive result (value mean confidence interval includes 0).

Ice Lake
total instructions in shared programs: 895506 -> 895384 (-0.01%)
instructions in affected programs: 658800 -> 658678 (-0.02%)
helped: 37
HURT: 0
helped stats (abs) min: 2 max: 8 x̄: 3.30 x̃: 2
helped stats (rel) min: <.01% max: 0.03% x̄: 0.02% x̃: 0.02%
95% mean confidence interval for instructions value: -4.00 -2.59
95% mean confidence interval for instructions %-change: -0.02% -0.02%
Instructions are helped.

total cycles in shared programs: 7092748 -> 7092224 (<.01%)
cycles in affected programs: 5272008 -> 5271484 (<.01%)
helped: 36
HURT: 14
helped stats (abs) min: 2 max: 440 x̄: 21.67 x̃: 8
helped stats (rel) min: <.01% max: 11.86% x̄: 1.12% x̃: 0.02%
HURT stats (abs)   min: 2 max: 122 x̄: 18.29 x̃: 6
HURT stats (rel)   min: <.01% max: 0.07% x̄: 0.01% x̃: <.01%
95% mean confidence interval for cycles value: -29.24 8.28
95% mean confidence interval for cycles %-change: -1.40% -0.21%
Inconclusive result (value mean confidence interval includes 0).

Regular shader-db results:

All Haswell+ platforms had similar results. (Tiger Lake shown)
total instructions in shared programs: 17611489 -> 17611408 (<.01%)
instructions in affected programs: 21188 -> 21107 (-0.38%)
helped: 23
HURT: 1
helped stats (abs) min: 1 max: 16 x̄: 3.78 x̃: 3
helped stats (rel) min: 0.03% max: 5.82% x̄: 1.13% x̃: 0.85%
HURT stats (abs)   min: 6 max: 6 x̄: 6.00 x̃: 6
HURT stats (rel)   min: 0.60% max: 0.60% x̄: 0.60% x̃: 0.60%
95% mean confidence interval for instructions value: -5.27 -1.48
95% mean confidence interval for instructions %-change: -1.70% -0.42%
Instructions are helped.

total cycles in shared programs: 338418502 -> 338366148 (-0.02%)
cycles in affected programs: 2289052 -> 2236698 (-2.29%)
helped: 18
HURT: 3
helped stats (abs) min: 4 max: 18000 x̄: 2909.67 x̃: 38
helped stats (rel) min: 0.09% max: 4.07% x̄: 0.96% x̃: 0.43%
HURT stats (abs)   min: 2 max: 14 x̄: 6.67 x̃: 4
HURT stats (rel)   min: 0.22% max: 1.13% x̄: 0.66% x̃: 0.64%
95% mean confidence interval for cycles value: -5204.00 217.91
95% mean confidence interval for cycles %-change: -1.31% -0.14%
Inconclusive result (value mean confidence interval includes 0).

Ivy Bridge
total instructions in shared programs: 11875617 -> 11875615 (<.01%)
instructions in affected programs: 1339 -> 1337 (-0.15%)
helped: 2
HURT: 0

No changes on any earlier Intel platforms.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Matt Turner <mattst88@gmail.com> [v1]
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4142>
 2020-03-18 20:36:29 +00:00
Ian Romanick
d6d63aec18 nir/algebraic: optimize ior(ine(a, 0), ine(b, 0)) to ine(ior(a, b), 0) 
Like 70f9e2589e.  Also scrub the unnecessary size qualifier in both
replacement patterns.

This occurs in a handful of places in the soft-fp64 code, and that is
the primary reason for the change.

Perhaps the patterns that generate umin should be conditioned on
something, but I'm not sure what.  lower_bitops might cover the cases
that matter, but it seems ugly.

Results on the 308 shaders extracted from the fp64 portion of the OpenGL
CTS:

Tiger Lake and Ice Lake had similar results. (Tiger Lake shown)
total instructions in shared programs: 936505 -> 933388 (-0.33%)
instructions in affected programs: 925719 -> 922602 (-0.34%)
helped: 154
HURT: 1
helped stats (abs) min: 1 max: 211 x̄: 35.45 x̃: 16
helped stats (rel) min: 0.34% max: 9.30% x̄: 2.28% x̃: 0.96%
HURT stats (abs)   min: 2342 max: 2342 x̄: 2342.00 x̃: 2342
HURT stats (rel)   min: 2.28% max: 2.28% x̄: 2.28% x̃: 2.28%
95% mean confidence interval for instructions value: -51.21 10.99
95% mean confidence interval for instructions %-change: -2.61% -1.89%
Inconclusive result (value mean confidence interval includes 0).

total cycles in shared programs: 7323502 -> 7306184 (-0.24%)
cycles in affected programs: 7220376 -> 7203058 (-0.24%)
helped: 126
HURT: 1
helped stats (abs) min: 2 max: 946 x̄: 159.10 x̃: 95
helped stats (rel) min: 0.01% max: 9.62% x̄: 0.80% x̃: 0.37%
HURT stats (abs)   min: 2728 max: 2728 x̄: 2728.00 x̃: 2728
HURT stats (rel)   min: 0.37% max: 0.37% x̄: 0.37% x̃: 0.37%
95% mean confidence interval for cycles value: -192.07 -80.66
95% mean confidence interval for cycles %-change: -1.07% -0.51%
Cycles are helped.

total spills in shared programs: 635 -> 817 (28.66%)
spills in affected programs: 635 -> 817 (28.66%)
helped: 0
HURT: 3

total fills in shared programs: 2065 -> 2438 (18.06%)
fills in affected programs: 2019 -> 2392 (18.47%)
helped: 0
HURT: 2

Regular shader-db results:

All Haswell+ platforms had similar results. (Tiger Lake shown)
total instructions in shared programs: 17611506 -> 17611489 (<.01%)
instructions in affected programs: 33442 -> 33425 (-0.05%)
helped: 32
HURT: 6
helped stats (abs) min: 1 max: 6 x̄: 1.69 x̃: 1
helped stats (rel) min: 0.08% max: 1.90% x̄: 0.27% x̃: 0.11%
HURT stats (abs)   min: 1 max: 15 x̄: 6.17 x̃: 5
HURT stats (rel)   min: 0.09% max: 1.50% x̄: 0.65% x̃: 0.55%
95% mean confidence interval for instructions value: -1.70 0.80
95% mean confidence interval for instructions %-change: -0.30% 0.05%
Inconclusive result (value mean confidence interval includes 0).

total cycles in shared programs: 338419218 -> 338418502 (<.01%)
cycles in affected programs: 385795 -> 385079 (-0.19%)
helped: 42
HURT: 3
helped stats (abs) min: 2 max: 192 x̄: 24.57 x̃: 16
helped stats (rel) min: 0.04% max: 2.09% x̄: 0.33% x̃: 0.22%
HURT stats (abs)   min: 64 max: 164 x̄: 105.33 x̃: 88
HURT stats (rel)   min: 0.77% max: 1.58% x̄: 1.09% x̃: 0.93%
95% mean confidence interval for cycles value: -29.76 -2.06
95% mean confidence interval for cycles %-change: -0.40% -0.07%
Cycles are helped.

Ivy Bridge and Sandy Bridge had similar results. (Ivy Bridge shown)
total instructions in shared programs: 11875620 -> 11875617 (<.01%)
instructions in affected programs: 421 -> 418 (-0.71%)
helped: 2
HURT: 0

total cycles in shared programs: 178245336 -> 178245326 (<.01%)
cycles in affected programs: 3425 -> 3415 (-0.29%)
helped: 2
HURT: 0

No changes on Gen4 or Gen5.

Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4142>
 2020-03-18 20:36:29 +00:00
Ian Romanick
88eb8f190b nir/algebraic: Simplify logic to detect sign of an integer 
This occurs in a handful of places in the soft-fp64 code, and that is
the primary reason for the change.

v2: Fix a typo in a comment.  Noticed by Matt.  Copy the correct fp64
shader-db results to the commit message.  I realized that I used
accidentally used the results from the next commit.

Results on the 308 shaders extracted from the fp64 portion of the OpenGL
CTS:

Tiger Lake and Ice Lake had similar results. (Tiger Lake shown)
total instructions in shared programs: 906235 -> 906149 (<.01%)
instructions in affected programs: 353966 -> 353880 (-0.02%)
helped: 31
HURT: 2
helped stats (abs) min: 1 max: 8 x̄: 3.03 x̃: 3
helped stats (rel) min: 0.01% max: 1.59% x̄: 0.10% x̃: 0.04%
HURT stats (abs)   min: 3 max: 5 x̄: 4.00 x̃: 4
HURT stats (rel)   min: 0.02% max: 0.02% x̄: 0.02% x̃: 0.02%
95% mean confidence interval for instructions value: -3.51 -1.70
95% mean confidence interval for instructions %-change: -0.19% <.01%
Inconclusive result (%-change mean confidence interval includes 0).

total cycles in shared programs: 7076552 -> 7076173 (<.01%)
cycles in affected programs: 2878361 -> 2877982 (-0.01%)
helped: 37
HURT: 2
helped stats (abs) min: 2 max: 48 x̄: 10.81 x̃: 6
helped stats (rel) min: <.01% max: 2.17% x̄: 0.47% x̃: 0.01%
HURT stats (abs)   min: 1 max: 20 x̄: 10.50 x̃: 10
HURT stats (rel)   min: <.01% max: 0.01% x̄: <.01% x̃: <.01%
95% mean confidence interval for cycles value: -13.96 -5.48
95% mean confidence interval for cycles %-change: -0.72% -0.16%
Cycles are helped.

total fills in shared programs: 2064 -> 2065 (0.05%)
fills in affected programs: 45 -> 46 (2.22%)
helped: 0
HURT: 1

Regular shader-db results:

All Gen7+ platforms had similar results. (Tiger Lake shown)
total instructions in shared programs: 17611530 -> 17611506 (<.01%)
instructions in affected programs: 5934 -> 5910 (-0.40%)
helped: 10
HURT: 0
helped stats (abs) min: 1 max: 5 x̄: 2.40 x̃: 2
helped stats (rel) min: 0.14% max: 1.24% x̄: 0.47% x̃: 0.34%
95% mean confidence interval for instructions value: -3.53 -1.27
95% mean confidence interval for instructions %-change: -0.78% -0.17%
Instructions are helped.

total cycles in shared programs: 338419178 -> 338419218 (<.01%)
cycles in affected programs: 19244 -> 19284 (0.21%)
helped: 4
HURT: 2
helped stats (abs) min: 2 max: 4 x̄: 3.00 x̃: 3
helped stats (rel) min: 0.05% max: 0.11% x̄: 0.08% x̃: 0.08%
HURT stats (abs)   min: 26 max: 26 x̄: 26.00 x̃: 26
HURT stats (rel)   min: 1.20% max: 1.20% x̄: 1.20% x̃: 1.20%
95% mean confidence interval for cycles value: -9.08 22.41
95% mean confidence interval for cycles %-change: -0.35% 1.04%
Inconclusive result (value mean confidence interval includes 0).

No changes on any earlier Intel platform.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4142>
 2020-03-18 20:36:29 +00:00
Juan A. Suarez Romero
90550b2a3e nir/algebraic: coalesce fmod lowering 
As fmod for 16/32/64 bits lowering does the same, let's merge all of
them in a single case.

Fixes dEQP-VK.glsl.builtin.precision_double.mod.compute.* on ACO.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Signed-off-by: Juan A. Suarez Romero <jasuarez@igalia.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4118>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4118>
 2020-03-12 16:42:52 +00:00
Jason Ekstrand
b20693be41 nir: Flush to zero with OOB low exponents in ldexp 
Reviewed-by: Arcady Goldmints-Orlov <agoldmints@igalia.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
 2020-03-04 11:39:50 -06:00
Hyunjun Ko
9e8466a866 nir: Add optimization for doing removing f16/f32 conversions 
This eliminates conversions between f16 and f32 where possible. We can
always remove an upcast followed by a down cast, that is:

  f2f16 ( f2f32 (a) )  ->  a
  f2fmp ( f2f32 (a) )  ->  a

In the other direction, f2f16 loses precision and can't be undone by a
f2f32.  However, by definition it's always safe to elminate f2fmp:

  f2f32 ( f2fmp (a) )  ->  a

v2. [Neil Roberts (nroberts@igalia.com)]

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3822>
 2020-02-24 17:24:13 +00:00
Neil Roberts
125f867d3d nir/opcodes: Add nir_op_f2fmp 
This opcode is the same as the f2f16 opcode except that it comes with
a promise that it is safe to optimise it out if the result is
immediately converted back to a 32-bit float again. Normally this
would be a lossy conversion and so it would be visible to the
application, but if the conversion is generated as part of the mediump
lowering process then this removal doesn’t matter. The opcode is
eventually replaced with a regular f2f16 in the late optimisations so
the backends don’t need to handle it.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3822>
 2020-02-24 17:24:13 +00:00
Erik Faye-Lund
0c1ba69a27 Revert "nir: Add a couple trivial abs optimizations" 
These were already added in 9fdaeb7776 ("nir: add min/max optimisation"),
and there's no point in doing them twice.

This reverts commit e4d346c86d.

Fixes: e4d346c86d ("nir: Add a couple trivial abs optimizations")
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3786>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3786>
 2020-02-13 09:18:27 +00:00
Samuel Pitoiset
8e77280774 nir: do not use De Morgan's Law rules for flt and fge 
In presence of NaNs, "!(flt(a, b) && flt(c, d))" is NOT EQUAL
to "fge(a, b) || fge(c, d)". These optimizations are unsafe for
apps that rely on NaN behaviour.

pipeline-db (GFX9/LLVM):
Totals from affected shaders:
SGPRS: 3176 -> 3136 (-1.26 %)
VGPRS: 2188 -> 2144 (-2.01 %)
Spilled SGPRs: 227 -> 169 (-25.55 %)
Code Size: 150572 -> 151800 (0.82 %) bytes
Max Waves: 307 -> 310 (0.98 %)

pipeline-db (GFX9/ACO):
Totals from affected shaders:
SGPRS: 18744 -> 18744 (0.00 %)
VGPRS: 15576 -> 15580 (0.03 %)
Spilled SGPRs: 164 -> 164 (0.00 %)
Code Size: 1573012 -> 1576492 (0.22 %) bytes
Max Waves: 1534 -> 1532 (-0.13 %)

Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/2127
Fixes: d1ed4ffe0b ("nir: Use De Morgan's Law on logic compounded comparisons")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3696>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3696>
 2020-02-11 08:36:00 +01:00
Rhys Perry
1f72857739 nir/algebraic: add some half packing optimizations 
pipeline-db (ACO):
Totals from affected shaders:
SGPRS: 29200 -> 29200 (0.00 %)
VGPRS: 17372 -> 17372 (0.00 %)
Spilled SGPRs: 105 -> 105 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Private memory VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 1406576 -> 1389256 (-1.23 %) bytes
LDS: 83 -> 83 (0.00 %) blocks
Max Waves: 3976 -> 3976 (0.00 %)

pipeline-db (LLVM):
Totals from affected shaders:
SGPRS: 21320 -> 21320 (0.00 %)
VGPRS: 17056 -> 17036 (-0.12 %)
Spilled SGPRs: 22 -> 22 (0.00 %)
Spilled VGPRs: 503 -> 487 (-3.18 %)
Private memory VGPRs: 0 -> 0 (0.00 %)
Scratch size: 396 -> 396 (0.00 %) dwords per thread
Code Size: 1441244 -> 1423292 (-1.25 %) bytes
LDS: 463 -> 463 (0.00 %) blocks
Max Waves: 3609 -> 3611 (0.06 %)

v2: add pattern for ishr

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/2271>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/2271>
 2020-01-29 14:30:33 +00:00
Rhys Perry
5476d18183 nir/algebraic: add patterns for a >> #b << #b 
Fixes compilation of a Battlefront 2 shader with ACO by removing VGPR
spilling. The reassociation makes it worse on LLVM though.

pipeline-db (ACO):
Totals from affected shaders:
SGPRS: 10704 -> 10688 (-0.15 %)
VGPRS: 18736 -> 18528 (-1.11 %)
Spilled SGPRs: 70 -> 70 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Private memory VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 909696 -> 885796 (-2.63 %) bytes
LDS: 225 -> 225 (0.00 %) blocks
Max Waves: 1115 -> 1129 (1.26 %)

pipeline-db (LLVM):
Totals from affected shaders:
SGPRS: 8472 -> 8424 (-0.57 %)
VGPRS: 14284 -> 14368 (0.59 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 442 -> 503 (13.80 %)
Private memory VGPRs: 0 -> 0 (0.00 %)
Scratch size: 268 -> 396 (47.76 %) dwords per thread
Code Size: 862568 -> 853028 (-1.11 %) bytes
LDS: 0 -> 0 (0.00 %) blocks
Max Waves: 971 -> 964 (-0.72 %)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/2271>
 2020-01-29 14:30:33 +00:00
Ian Romanick
b065d8fb8c nir/algebraic: Optimize some 64-bit integer comparisons involving zero 
I noticed that we can do better for these kinds of comparisons while
working on the lowering for iadd_sat@64 and isub_sat@64.  This
eliminated 11 instruction from the fs-addSaturate-int64.shader_test.

My hope is that this will improve the run-time of int64 tests on Ice
Lake.  I have no data to support or refute this.

Unsurprisingly, no changes on shader-db.

v2: Condition the min and max patterns with nir_lower_minmax64.
Suggested by Caio.  Very long discussion in the MR. :)

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/767>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/767>
 2020-01-23 00:18:57 +00:00
Ian Romanick
d3d970166c nir/algebraic: Add lowering for 64-bit iadd_sat and isub_sat 
v2: Rearranged and expand the comment about the optimizations applied to
the lowering.  Suggested by Caio.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/767>
 2020-01-23 00:18:57 +00:00
Ian Romanick
dcadbd2dd2 nir/algebraic: Add lowering for 64-bit uadd_sat 
Fixes piglit fs-addsaturate-uint64 and vs-addsaturate-uint64 on Ice
Lake.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/767>
 2020-01-23 00:18:57 +00:00
Ian Romanick
1bdfc6d7cb nir/algebraic: Add lowering for 64-bit usub_sat 
v2: Rebase on 272e927d0e ("nir/spirv: initial handling of OpenCL.std
extension opcodes")

v3: Add a new lower_usub_sat64 flag that only applies to the 64-bit
version of the nir_op_usub_sat instruction.

v4: Also enable the lowering when nir_lower_iadd64 is set.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com> [v3]
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/767>
 2020-01-23 00:18:57 +00:00
Ian Romanick
a483771045 nir/algebraic: Add lowering for 64-bit hadd and rhadd 
v2: Rebase on 272e927d0e ("nir/spirv: initial handling of OpenCL.std
extension opcodes")

v3: Add a new lower_hadd64 flag that only applies to the 64-bit versions
of the instructions.

v4: Also enable the lowering when nir_lower_iadd64 is set.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com> [v3]
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/767>
 2020-01-23 00:18:57 +00:00
Ian Romanick
ea435560ee nir/algebraic: Add lowering for uabs_usub and uabs_isub 
v2: Remove some rebase failures noticed by Caio.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/767>
 2020-01-23 00:18:57 +00:00
Elie Tournier
22c5c54a4f nir/algebraic: sqrt(x)*sqrt(x) -> fabs(x) 
total instructions in shared programs: 12840840 -> 12839341 (-0.01%)
instructions in affected programs: 122581 -> 121082 (-1.22%)
helped: 559
HURT: 0

total cycles in shared programs: 302505756 -> 302490031 (<.01%)
cycles in affected programs: 2022900 -> 2007175 (-0.78%)
helped: 1090
HURT: 130

Signed-off-by: Elie Tournier <elie.tournier@collabora.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/948>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/948>
 2020-01-15 00:30:52 +00:00
Elie Tournier
6f394343b1 nir/algebraic: i2f(f2i()) -> trunc() 
total instructions in shared programs: 12840968 -> 12840784 (<.01%)
instructions in affected programs: 17886 -> 17702 (-1.03%)
helped: 77
HURT: 0

total cycles in shared programs: 302508917 -> 302505592 (<.01%)
cycles in affected programs: 249964 -> 246639 (-1.33%)
helped: 70
HURT: 7

Signed-off-by: Elie Tournier <elie.tournier@collabora.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/948>
 2020-01-15 00:30:52 +00:00
Rhys Perry
cc3ef3643a nir/algebraic: a & ~(a >> 31) -> imax(a, 0) 
Found in some Doom shaders

Totals from affected shaders:
SGPRS: 30056 -> 30064 (0.03 %)
VGPRS: 28024 -> 28024 (0.00 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Code Size: 4278648 -> 4270852 (-0.18 %) bytes
Max Waves: 1476 -> 1476 (0.00 %)
Instructions: 835287 -> 833338 (-0.23 %)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3089>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3089>
 2020-01-14 17:54:40 +00:00
Jonathan Marek
004797002f nir: add option to lower half packing opcodes 
Signed-off-by: Jonathan Marek <jonathan@marek.ca>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3106>
 2019-12-16 19:20:07 -05:00
Ian Romanick
fbd5359a0a nir/algebraic: Rearrange bcsel sequences generated by nir_opt_peephole_select 
Reviewed-by: Matt Turner <mattst88@gmail.com>

All Intel platforms had similar results. (Ice Lake shown)
total instructions in shared programs: 14660366 -> 14653437 (-0.05%)
instructions in affected programs: 316166 -> 309237 (-2.19%)
helped: 905
HURT: 10
helped stats (abs) min: 1 max: 36 x̄: 7.67 x̃: 6
helped stats (rel) min: 0.13% max: 18.75% x̄: 4.28% x̃: 3.60%
HURT stats (abs)   min: 1 max: 1 x̄: 1.00 x̃: 1
HURT stats (rel)   min: 0.10% max: 1.33% x̄: 0.70% x̃: 0.97%
95% mean confidence interval for instructions value: -7.91 -7.23
95% mean confidence interval for instructions %-change: -4.46% -3.99%
Instructions are helped.

total cycles in shared programs: 228571646 -> 228549759 (<.01%)
cycles in affected programs: 56239919 -> 56218032 (-0.04%)
helped: 681
HURT: 216
helped stats (abs) min: 1 max: 5156 x̄: 45.49 x̃: 10
helped stats (rel) min: <.01% max: 10.45% x̄: 1.29% x̃: 0.65%
HURT stats (abs)   min: 1 max: 320 x̄: 42.09 x̃: 14
HURT stats (rel)   min: <.01% max: 37.04% x̄: 1.38% x̃: 0.49%
95% mean confidence interval for cycles value: -41.51 -7.29
95% mean confidence interval for cycles %-change: -0.80% -0.49%
Cycles are helped.

LOST:   1
GAINED: 0
 2019-12-02 16:46:20 -08:00
Ian Romanick
780b5c1037 nir/algebraic: Simplify some Inf and NaN avoidance code 
Since a is non-negative, neither fsqrt nor frsq should return NaN.  frsq
should only return Inf when fsqrt returns 0.

The changes are pretty small, but this turns a few hundred hurt shaders
in the next patch into helped shaders.

An alternative to the intBitsToFloat is to import numpy and do
np.finfo(np.float32).max.  That's more explicit, but we may also want to
have specific bit encodings of float values later.  I could be convinced
either way, but intBitsToFloat(0x7f7fffff) was what I implemented first.

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

All Gen7+ platforms had similar results. (Ice Lake shown)
total instructions in shared programs: 14661140 -> 14661104 (<.01%)
instructions in affected programs: 7520 -> 7484 (-0.48%)
helped: 36
HURT: 0
helped stats (abs) min: 1 max: 1 x̄: 1.00 x̃: 1
helped stats (rel) min: 0.32% max: 0.61% x̄: 0.49% x̃: 0.52%
95% mean confidence interval for instructions value: -1.00 -1.00
95% mean confidence interval for instructions %-change: -0.52% -0.47%
Instructions are helped.

total cycles in shared programs: 228585416 -> 228584806 (<.01%)
cycles in affected programs: 56321 -> 55711 (-1.08%)
helped: 32
HURT: 0
helped stats (abs) min: 2 max: 98 x̄: 19.06 x̃: 10
helped stats (rel) min: 0.08% max: 6.41% x̄: 1.09% x̃: 0.65%
95% mean confidence interval for cycles value: -28.32 -9.80
95% mean confidence interval for cycles %-change: -1.63% -0.54%
Cycles are helped.

Sandy Bridge
total cycles in shared programs: 152991077 -> 152991075 (<.01%)
cycles in affected programs: 11525 -> 11523 (-0.02%)
helped: 2
HURT: 2
helped stats (abs) min: 2 max: 4 x̄: 3.00 x̃: 3
helped stats (rel) min: 0.07% max: 0.11% x̄: 0.09% x̃: 0.09%
HURT stats (abs)   min: 2 max: 2 x̄: 2.00 x̃: 2
HURT stats (rel)   min: 0.08% max: 0.08% x̄: 0.08% x̃: 0.08%
95% mean confidence interval for cycles value: -5.27 4.27
95% mean confidence interval for cycles %-change: -0.16% 0.15%
Inconclusive result (value mean confidence interval includes 0).

No changes on Iron Lake or GM45.
 2019-12-02 16:46:20 -08:00
Ian Romanick
9be4a422a0 nir/algebraic: Mark other comparison exact when removing a == a 
This prevents some additional optimizations that would change the
original result.  This includes things like (b < a && b < c) => b <
min(a, c) and !(a < b) => b >= a.  Both of these optimizations were
specifically observed in the piglit tests added in piglit!160.

This was discovered while investigating
https://gitlab.freedesktop.org/mesa/mesa/issues/1958.  However, the
problem in that issue was Chrome or Angle is replacing calls to isnan()
with some stuff that we (correctly) optimize to false.  If they had left
the calls to isnan() alone, everything would have just worked.

No shader-db changes on any Intel platform.

I also tried marking the comparison generated by the isnan() function
precise.  The precise marker "infects" every computation involved in
calculating the parameter to the isnan() function, and this severely
hurt all of the (few) shaders in shader-db that use isnan().

I also considered adding a new ir_unop_isnan opcode that would implement
the functionality.  During GLSL IR-to-NIR translation, the resulting
comparison operation would be marked exact (and the samething would need
to happen in SPIR-V translation).

This approach taken by this patch seemed easier, but we may want to do
the ir_unop_isnan thing anyway.

Fixes: d55835b8bd ("nir/algebraic: Add optimizations for "a == a && a CMP b"")
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
 2019-11-04 14:05:49 -08:00
Ian Romanick
ea19f2fb68 nir/algebraic: Add the ability to mark a replacement as exact 
Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
 2019-11-04 14:05:49 -08:00
Rob Clark
5e08f070f0 nir: add nir_lower_amul pass 
Lower amul to either imul or imul24, depending on whether 24b is enough
bits to calculate an offset within the thing being dereferenced.

Signed-off-by: Rob Clark <robdclark@chromium.org>
 2019-10-18 15:08:54 -07:00
Rob Clark
1bdde31392 nir: add address calc related opt rules 
Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>
 2019-10-18 15:08:54 -07:00
Rob Clark
6320e37d4b nir: add amul instruction 
Used for address/offset calculation (ie. array derefs), where we can
potentially use less than 32b for the multiply of array idx by element
size.  For backends that support `imul24`, this gives a lowering pass
an easy way to find multiplies that potentially can be converted to
`imul24`.

Signed-off-by: Rob Clark <robdclark@chromium.org>
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>
 2019-10-18 15:08:54 -07:00
Daniel Schürmann
239423d234 nir: Remove unnecessary subtraction optimizations 
These optimizations are already covered after lowering.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
 2019-09-30 09:44:10 +00:00
Daniel Schürmann
99848a57b7 nir: recombine nir_op_*sub when lower_sub = false 
There are some optimizations which are only implemented for additions
and some optimizations which assume that subtractions have been lowered.
By lowering all subtractions first and later recombine for backends
which prefer this option, we don't have to implement them twice.

This patch also moves lower_negate to nir_opt_algebraic_late() to enable
these optimizations for backends which make use of it.

Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
 2019-09-30 09:44:10 +00:00
Ian Romanick
317a88b920 nir/algebraic: Additional D3D Boolean optimization 
I observed this pattern in several shaders in Hand of Fate 2 while
investigating bugzilla #111490.  This also led to the related
bugzilla #111578.  The shaders from HoF2 are *not* in shader-db.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

Skylake and Ice Lake had similar results. (Ice Lake shown)
total instructions in shared programs: 16222621 -> 16205419 (-0.11%)
instructions in affected programs: 798418 -> 781216 (-2.15%)
helped: 548
HURT: 0
helped stats (abs) min: 2 max: 158 x̄: 31.39 x̃: 35
helped stats (rel) min: 0.45% max: 28.64% x̄: 2.83% x̃: 2.09%
95% mean confidence interval for instructions value: -33.22 -29.56
95% mean confidence interval for instructions %-change: -3.11% -2.56%
Instructions are helped.

total cycles in shared programs: 364676209 -> 363345763 (-0.36%)
cycles in affected programs: 112810504 -> 111480058 (-1.18%)
helped: 546
HURT: 7
helped stats (abs) min: 2 max: 118913 x̄: 2439.77 x̃: 2340
helped stats (rel) min: 0.08% max: 37.56% x̄: 1.46% x̃: 1.08%
HURT stats (abs)   min: 2 max: 770 x̄: 238.00 x̃: 43
HURT stats (rel)   min: 0.02% max: 11.24% x̄: 3.71% x̃: 0.35%
95% mean confidence interval for cycles value: -2884.33 -1927.41
95% mean confidence interval for cycles %-change: -1.59% -1.21%
Cycles are helped.

total spills in shared programs: 8870 -> 8514 (-4.01%)
spills in affected programs: 1230 -> 874 (-28.94%)
helped: 161
HURT: 0

total fills in shared programs: 21901 -> 21348 (-2.52%)
fills in affected programs: 2120 -> 1567 (-26.08%)
helped: 155
HURT: 5

Broadwell and Haswell had similar results. (Broadwell shown)
total instructions in shared programs: 14994910 -> 14975495 (-0.13%)
instructions in affected programs: 839033 -> 819618 (-2.31%)
helped: 548
HURT: 0
helped stats (abs) min: 2 max: 299 x̄: 35.43 x̃: 49
helped stats (rel) min: 0.39% max: 19.89% x̄: 2.91% x̃: 2.22%
95% mean confidence interval for instructions value: -37.46 -33.40
95% mean confidence interval for instructions %-change: -3.12% -2.70%
Instructions are helped.

total cycles in shared programs: 386032453 -> 384450722 (-0.41%)
cycles in affected programs: 117807357 -> 116225626 (-1.34%)
helped: 547
HURT: 6
helped stats (abs) min: 2 max: 22096 x̄: 2892.01 x̃: 3926
helped stats (rel) min: 0.17% max: 10.34% x̄: 1.56% x̃: 1.31%
HURT stats (abs)   min: 4 max: 60 x̄: 32.83 x̃: 29
HURT stats (rel)   min: 0.38% max: 12.79% x̄: 5.86% x̃: 4.65%
95% mean confidence interval for cycles value: -3060.28 -2660.27
95% mean confidence interval for cycles %-change: -1.59% -1.37%
Cycles are helped.

total spills in shared programs: 23372 -> 21869 (-6.43%)
spills in affected programs: 11730 -> 10227 (-12.81%)
helped: 352
HURT: 0

total fills in shared programs: 34747 -> 35351 (1.74%)
fills in affected programs: 11013 -> 11617 (5.48%)
helped: 3
HURT: 347

Ivy Bridge and Sandybridge had similar results. (Ivy Bridge shown)
total instructions in shared programs: 11956420 -> 11956126 (<.01%)
instructions in affected programs: 14898 -> 14604 (-1.97%)
helped: 98
HURT: 0
helped stats (abs) min: 3 max: 3 x̄: 3.00 x̃: 3
helped stats (rel) min: 1.30% max: 3.57% x̄: 2.08% x̃: 2.00%
95% mean confidence interval for instructions value: -3.00 -3.00
95% mean confidence interval for instructions %-change: -2.18% -1.98%
Instructions are helped.

total cycles in shared programs: 178791217 -> 178790792 (<.01%)
cycles in affected programs: 149763 -> 149338 (-0.28%)
helped: 91
HURT: 7
helped stats (abs) min: 3 max: 107 x̄: 20.63 x̃: 16
helped stats (rel) min: 0.13% max: 6.91% x̄: 1.40% x̃: 1.18%
HURT stats (abs)   min: 3 max: 322 x̄: 207.43 x̃: 322
HURT stats (rel)   min: 0.14% max: 19.85% x̄: 12.73% x̃: 17.41%
95% mean confidence interval for cycles value: -18.94 10.27
95% mean confidence interval for cycles %-change: -1.28% 0.49%
Inconclusive result (value mean confidence interval includes 0).
 2019-09-19 14:22:22 -07:00
Ian Romanick
92f70df8c3 nir/algebraic: Do not apply late DPH optimization in vertex processing stages 
Some shaders do not use 'invariant' in vertex and (possibly) geometry
shader stages on some outputs that are intended to be invariant.  For
various reasons, this optimization may not be fully applied in all
shaders used for different rendering passes of the same geometry.  This
can result in Z-fighting artifacts (at best).  For now, disable this
optimization in these stages.

In tessellation stages applications seem to use 'precise' when
necessary, so allow the optimization in those stages.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111490
Fixes: 09705747d7 ("nir/algebraic: Reassociate fadd into fmul in DPH-like pattern")

All Gen8+ platforms had similar results. (Ice Lake shown)
total instructions in shared programs: 16194726 -> 16344745 (0.93%)
instructions in affected programs: 2855172 -> 3005191 (5.25%)
helped: 6
HURT: 20279
helped stats (abs) min: 1 max: 3 x̄: 1.33 x̃: 1
helped stats (rel) min: 0.44% max: 1.00% x̄: 0.54% x̃: 0.44%
HURT stats (abs)   min: 1 max: 32 x̄: 7.40 x̃: 7
HURT stats (rel)   min: 0.14% max: 42.86% x̄: 8.58% x̃: 6.56%
95% mean confidence interval for instructions value: 7.34 7.45
95% mean confidence interval for instructions %-change: 8.48% 8.67%
Instructions are HURT.

total cycles in shared programs: 364471296 -> 365014683 (0.15%)
cycles in affected programs: 32421530 -> 32964917 (1.68%)
helped: 2925
HURT: 16144
helped stats (abs) min: 1 max: 403 x̄: 18.39 x̃: 5
helped stats (rel) min: <.01% max: 22.61% x̄: 1.97% x̃: 1.15%
HURT stats (abs)   min: 1 max: 18471 x̄: 36.99 x̃: 15
HURT stats (rel)   min: 0.02% max: 52.58% x̄: 5.60% x̃: 3.87%
95% mean confidence interval for cycles value: 21.58 35.41
95% mean confidence interval for cycles %-change: 4.36% 4.52%
Cycles are HURT.
 2019-09-19 14:21:31 -07:00
Andres Gomez
3f782cdd25 nir/algebraic: mark float optimizations returning one parameter as inexact 
With the arrival of VK_KHR_shader_float_controls algebraic
optimizations for float types of the form (('fop', a, b), a) become
inexact depending on the execution mode.

For example, if we have activated SHADER_DENORM_FLUSH_TO_ZERO, in case
of a denorm value for the "a" parameter, we cannot return it still as
a denorm, it needs to be flushed to zero. Therefore, we mark now all
those operations as inexact.

Suggested-by: Samuel Iglesias Gonsálvez <siglesias@igalia.com>
Signed-off-by: Andres Gomez <agomez@igalia.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
 2019-09-17 23:39:18 +03:00
Ian Romanick
e07248d2a8 nir/algebraic: Clean up value range analysis-based optimizations 
Fix the a / b ordering in some compares.  Delete duplicate patterns.
Add a table explaining things.  While I was cleaning this up, I managed
to confuse myself.  The table helped sort that out.

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
 2019-08-29 13:15:52 -07:00
Ian Romanick
ccb236d1bc nir/algebraic: Mark some value range analysis-based optimizations imprecise 
This didn't fix bug #111308, but it was found will trying to find the
actual cause of that bug.

Fixes piglit tests (new in piglit!110):

    - fs-fract-of-NaN.shader_test
    - fs-lt-nan-tautology.shader_test
    - fs-ge-nan-tautology.shader_test

No shader-db changes on any Intel platform.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=111308
Fixes: b77070e293 ("nir/algebraic: Use value range analysis to eliminate tautological compares")
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
 2019-08-29 13:15:52 -07:00
Ian Romanick
d3fd1c761a nir/algrbraic: Don't optimize open-coded bitfield reverse when lowering is enabled 
This caused a problem on Sandybridge where an open-coded
bitfieldReverse() function could be optimized to a
nir_op_bitfield_reverse that would generate an unsupported BFREV
instruction in the backend.  This was encountered in some Unreal4 tech
demos in shader-db.  The bug was not previously noticed because we don't
actually try to run those demos on Sandybridge.

The fixes tag is a bit a lie.  The actual bug was introduced about
26,000 commits earlier in 371c4b3c48 ("nir: Recognize open-coded
bitfield_reverse.").  Without the NIR lowering pass, the flag needed to
avoid the optimization does not exist.  Hopefully nobody will care to
fix this on an earlier Mesa release.

Reviewed-by: Matt Turner <mattst88@gmail.com>
Fixes: 7afa26d4e3 ("nir: Add lowering for nir_op_bitfield_reverse.")
 2019-08-28 11:38:51 -07:00
Daniel Schürmann
7fa1740035 nir/algebraic: some subtraction optimizations 
Changes with RADV/ACO:
Totals from affected shaders:
SGPRS: 444087 -> 455543 (2.58 %)
VGPRS: 436468 -> 436768 (0.07 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Private memory VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 13448928 -> 13353520 (-0.71 %) bytes
LDS: 0 -> 0 (0.00 %) blocks
Max Waves: 68060 -> 67979 (-0.12 %)
Wait states: 0 -> 0 (0.00 %)

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
 2019-08-21 08:51:49 +00:00
Rhys Perry
0a790c3019 nir/algebraic: add a few masking-before-unpack optimizations 
Helps some Dawn of War 3 and F1 2017 shaders with ACO:
Totals from affected shaders:
SGPRS: 2136 -> 2128 (-0.37 %)
VGPRS: 1624 -> 1628 (0.25 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Private memory VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 168068 -> 164332 (-2.22 %) bytes
LDS: 44 -> 44 (0.00 %) blocks
Max Waves: 222 -> 221 (-0.45 %)
Wait states: 0 -> 0 (0.00 %)

Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
 2019-08-16 12:13:01 +01:00
Ian Romanick
0e6581b87d nir/algebraic: Reassociate shift-by-constant of shift-by-constant 
v2: After some review discussion with Alyssa, the replacements now
correct account for cases where (b+c) >= bitsize.

v3: Use a temporary to simplify the Python code quite a bit.  Suggested
by Jason.

Haswell and all Gen8+ platforms had similar results. (Ice Lake shown)
total instructions in shared programs: 16251155 -> 16249576 (<.01%)
instructions in affected programs: 232627 -> 231048 (-0.68%)
helped: 547
HURT: 1
helped stats (abs) min: 1 max: 15 x̄: 2.89 x̃: 3
helped stats (rel) min: 0.04% max: 7.84% x̄: 1.14% x̃: 1.06%
HURT stats (abs)   min: 2 max: 2 x̄: 2.00 x̃: 2
HURT stats (rel)   min: 0.12% max: 0.12% x̄: 0.12% x̃: 0.12%
95% mean confidence interval for instructions value: -3.12 -2.65
95% mean confidence interval for instructions %-change: -1.20% -1.06%
Instructions are helped.

total cycles in shared programs: 365924392 -> 365372103 (-0.15%)
cycles in affected programs: 59207053 -> 58654764 (-0.93%)
helped: 497
HURT: 34
helped stats (abs) min: 1 max: 29300 x̄: 1118.16 x̃: 16
helped stats (rel) min: <.01% max: 10.59% x̄: 1.82% x̃: 1.82%
HURT stats (abs)   min: 2 max: 424 x̄: 101.03 x̃: 63
HURT stats (rel)   min: 0.07% max: 46.17% x̄: 4.72% x̃: 2.06%
95% mean confidence interval for cycles value: -1426.41 -653.77
95% mean confidence interval for cycles %-change: -1.66% -1.15%
Cycles are helped.

total spills in shared programs: 8870 -> 8871 (0.01%)
spills in affected programs: 104 -> 105 (0.96%)
helped: 0
HURT: 1

Ivy Bridge and all pre-Gen7 platforms had similar results. (Ivy Bridge shown)
total instructions in shared programs: 11956236 -> 11955635 (<.01%)
instructions in affected programs: 94110 -> 93509 (-0.64%)
helped: 106
HURT: 0
helped stats (abs) min: 1 max: 14 x̄: 5.67 x̃: 4
helped stats (rel) min: 0.12% max: 4.71% x̄: 1.96% x̃: 0.76%
95% mean confidence interval for instructions value: -6.62 -4.72
95% mean confidence interval for instructions %-change: -2.27% -1.64%
Instructions are helped.

total cycles in shared programs: 179296340 -> 178788044 (-0.28%)
cycles in affected programs: 51009603 -> 50501307 (-1.00%)
helped: 82
HURT: 7
helped stats (abs) min: 5 max: 27820 x̄: 6199.00 x̃: 16
helped stats (rel) min: 0.30% max: 8.16% x̄: 2.58% x̃: 3.11%
HURT stats (abs)   min: 2 max: 8 x̄: 3.14 x̃: 2
HURT stats (rel)   min: 0.02% max: 1.40% x̄: 0.34% x̃: 0.10%
95% mean confidence interval for cycles value: -7649.38 -3773.00
95% mean confidence interval for cycles %-change: -2.71% -1.99%
Cycles are helped.

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com> [v2]
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
 2019-08-14 11:15:37 -07:00
Ian Romanick
73aaeac0a3 nir/algebraic: Reassociate add-and-shift to be shift-and-add 
A common thing in many shaders:

    uniform vs { vec4 bones[...]; };

    ...

    x = some_calculation(bones[i + 0]);
    y = some_calculation(bones[i + 1]);
    z = some_calculation(bones[i + 2]);

This turns into stuff like

    vec1 32 ssa_12 = iadd ssa_11, ssa_0
    vec1 32 ssa_13 = ishl ssa_12, ssa_3
    vec1 32 ssa_14 = intrinsic load_ssbo (ssa_7, ssa_13) (16, 4, 0)
    vec1 32 ssa_15 = iadd ssa_11, ssa_1
    vec1 32 ssa_16 = ishl ssa_15, ssa_3
    vec1 32 ssa_17 = intrinsic load_ssbo (ssa_7, ssa_16) (16, 4, 0)
    vec1 32 ssa_18 = iadd ssa_11, ssa_2
    vec1 32 ssa_19 = ishl ssa_18, ssa_3
    vec1 32 ssa_20 = intrinsic load_ssbo (ssa_7, ssa_19) (16, 4, 0)

By reassociating the shift and the add, we can reduce this to

    vec1 32 ssa_12 = ishl ssa_11, ssa_3
    vec1 32 ssa_13 = iadd ssa_12, ssa_0
    vec1 32 ssa_14 = intrinsic load_ssbo (ssa_7, ssa_13) (16, 4, 0)
    vec1 32 ssa_16 = iadd ssa_12, ssa_1
    vec1 32 ssa_17 = intrinsic load_ssbo (ssa_7, ssa_16) (16, 4, 0)
    vec1 32 ssa_19 = iadd ssa_12, ssa_2
    vec1 32 ssa_20 = intrinsic load_ssbo (ssa_7, ssa_19) (16, 4, 0)

v2: Add some commentary from Rhys Perry's nearly identical patch.

All Intel platforms had similar results. (Ice Lake shown)
total instructions in shared programs: 16277758 -> 16250704 (-0.17%)
instructions in affected programs: 1440284 -> 1413230 (-1.88%)
helped: 4920
HURT: 6
helped stats (abs) min: 1 max: 69 x̄: 5.50 x̃: 4
helped stats (rel) min: 0.10% max: 18.33% x̄: 2.21% x̃: 1.79%
HURT stats (abs)   min: 1 max: 12 x̄: 4.50 x̃: 3
HURT stats (rel)   min: 0.18% max: 3.23% x̄: 1.91% x̃: 2.55%
95% mean confidence interval for instructions value: -5.67 -5.31
95% mean confidence interval for instructions %-change: -2.26% -2.16%
Instructions are helped.

total cycles in shared programs: 367118526 -> 365895358 (-0.33%)
cycles in affected programs: 93504145 -> 92280977 (-1.31%)
helped: 2754
HURT: 1269
helped stats (abs) min: 1 max: 47039 x̄: 460.66 x̃: 16
helped stats (rel) min: <.01% max: 34.93% x̄: 3.77% x̃: 1.12%
HURT stats (abs)   min: 1 max: 1500 x̄: 35.85 x̃: 9
HURT stats (rel)   min: 0.01% max: 17.35% x̄: 2.18% x̃: 0.75%
95% mean confidence interval for cycles value: -387.31 -220.78
95% mean confidence interval for cycles %-change: -2.11% -1.68%
Cycles are helped.

LOST:   1
GAINED: 1

Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
 2019-08-14 11:15:32 -07:00
Ian Romanick
5544b2cbbd nir/algebraic: Use value range analysis to eliminate useless unary ops 
Sandy Bridge is the big winner because it lies at something of a
crossroads.  It supports a fairly high OpenGL version, and it still has
the old style math box.  The high OpenGL version means a lot more
shaders can run on it.  The old style math box means extra moves are
necessary to resolve source modifiers on operands to complex math
instructions like COS, SQRT, and RCP.

v2: Remove a couple patterns that are now redundant.

All Gen7+ platforms had similar results.  (Ice Lake shown)
total instructions in shared programs: 16282006 -> 16278207 (-0.02%)
instructions in affected programs: 174555 -> 170756 (-2.18%)
helped: 661
HURT: 0
helped stats (abs) min: 1 max: 36 x̄: 5.75 x̃: 3
helped stats (rel) min: 0.06% max: 23.68% x̄: 2.81% x̃: 1.94%
95% mean confidence interval for instructions value: -6.16 -5.34
95% mean confidence interval for instructions %-change: -3.02% -2.60%
Instructions are helped.

total cycles in shared programs: 367168597 -> 367134284 (<.01%)
cycles in affected programs: 1105276 -> 1070963 (-3.10%)
helped: 460
HURT: 150
helped stats (abs) min: 1 max: 568 x̄: 96.60 x̃: 82
helped stats (rel) min: 0.02% max: 32.50% x̄: 7.99% x̃: 4.27%
HURT stats (abs)   min: 1 max: 901 x̄: 67.49 x̃: 39
HURT stats (rel)   min: 0.07% max: 20.00% x̄: 4.90% x̃: 4.22%
95% mean confidence interval for cycles value: -65.68 -46.82
95% mean confidence interval for cycles %-change: -5.59% -4.05%
Cycles are helped.

Sandy Bridge
total instructions in shared programs: 10824272 -> 10802557 (-0.20%)
instructions in affected programs: 1237988 -> 1216273 (-1.75%)
helped: 8199
HURT: 0
helped stats (abs) min: 1 max: 41 x̄: 2.65 x̃: 2
helped stats (rel) min: 0.12% max: 20.00% x̄: 2.04% x̃: 1.73%
95% mean confidence interval for instructions value: -2.70 -2.59
95% mean confidence interval for instructions %-change: -2.07% -2.00%
Instructions are helped.

total cycles in shared programs: 154009894 -> 153843598 (-0.11%)
cycles in affected programs: 10650486 -> 10484190 (-1.56%)
helped: 4973
HURT: 1533
helped stats (abs) min: 1 max: 3904 x̄: 40.20 x̃: 20
helped stats (rel) min: 0.02% max: 41.72% x̄: 2.63% x̃: 1.67%
HURT stats (abs)   min: 1 max: 453 x̄: 21.94 x̃: 8
HURT stats (rel)   min: 0.02% max: 41.91% x̄: 1.54% x̃: 0.58%
95% mean confidence interval for cycles value: -28.02 -23.10
95% mean confidence interval for cycles %-change: -1.74% -1.56%
Cycles are helped.

LOST:   0
GAINED: 2

GM45 and Iron Lake had similar results. (Iron Lake shown)
total instructions in shared programs: 8135196 -> 8134888 (<.01%)
instructions in affected programs: 31920 -> 31612 (-0.96%)
helped: 169
HURT: 0
helped stats (abs) min: 1 max: 12 x̄: 1.82 x̃: 2
helped stats (rel) min: 0.43% max: 3.23% x̄: 1.23% x̃: 1.16%
95% mean confidence interval for instructions value: -2.01 -1.64
95% mean confidence interval for instructions %-change: -1.32% -1.15%
Instructions are helped.

total cycles in shared programs: 188575724 -> 188574092 (<.01%)
cycles in affected programs: 406840 -> 405208 (-0.40%)
helped: 169
HURT: 0
helped stats (abs) min: 4 max: 72 x̄: 9.66 x̃: 10
helped stats (rel) min: 0.07% max: 2.16% x̄: 0.57% x̃: 0.47%
95% mean confidence interval for cycles value: -10.72 -8.59
95% mean confidence interval for cycles %-change: -0.63% -0.50%
Cycles are helped.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
 2019-08-05 20:14:14 -07:00
Ian Romanick
8d14380971 nir/algebraic: Use value range analysis to convert fmin to fsat 
All Gen8+ platforms had similar results. (Ice Lake shown)
total instructions in shared programs: 16297320 -> 16282006 (-0.09%)
instructions in affected programs: 2434498 -> 2419184 (-0.63%)
helped: 8091
HURT: 1
helped stats (abs) min: 1 max: 51 x̄: 1.89 x̃: 2
helped stats (rel) min: 0.04% max: 14.29% x̄: 0.98% x̃: 0.95%
HURT stats (abs)   min: 7 max: 7 x̄: 7.00 x̃: 7
HURT stats (rel)   min: 0.28% max: 0.28% x̄: 0.28% x̃: 0.28%
95% mean confidence interval for instructions value: -1.94 -1.85
95% mean confidence interval for instructions %-change: -0.99% -0.96%
Instructions are helped.

total cycles in shared programs: 367221624 -> 367168597 (-0.01%)
cycles in affected programs: 126409635 -> 126356608 (-0.04%)
helped: 5612
HURT: 1023
helped stats (abs) min: 1 max: 2332 x̄: 31.11 x̃: 16
helped stats (rel) min: <.01% max: 30.31% x̄: 1.69% x̃: 1.42%
HURT stats (abs)   min: 1 max: 2372 x̄: 118.84 x̃: 16
HURT stats (rel)   min: <.01% max: 46.98% x̄: 1.46% x̃: 0.35%
95% mean confidence interval for cycles value: -11.52 -4.46
95% mean confidence interval for cycles %-change: -1.26% -1.14%
Cycles are helped.

total spills in shared programs: 8868 -> 8870 (0.02%)
spills in affected programs: 28 -> 30 (7.14%)
helped: 0
HURT: 1

total fills in shared programs: 21903 -> 21904 (<.01%)
fills in affected programs: 42 -> 43 (2.38%)
helped: 0
HURT: 1

Haswell
total instructions in shared programs: 13353925 -> 13338728 (-0.11%)
instructions in affected programs: 2265850 -> 2250653 (-0.67%)
helped: 8127
HURT: 5
helped stats (abs) min: 1 max: 51 x̄: 1.88 x̃: 2
helped stats (rel) min: 0.04% max: 20.00% x̄: 1.13% x̃: 1.07%
HURT stats (abs)   min: 5 max: 16 x̄: 9.00 x̃: 6
HURT stats (rel)   min: 0.19% max: 0.52% x̄: 0.35% x̃: 0.28%
95% mean confidence interval for instructions value: -1.91 -1.83
95% mean confidence interval for instructions %-change: -1.15% -1.11%
Instructions are helped.

total cycles in shared programs: 375535444 -> 375536343 (<.01%)
cycles in affected programs: 131206582 -> 131207481 (<.01%)
helped: 5590
HURT: 1055
helped stats (abs) min: 1 max: 2844 x̄: 34.15 x̃: 16
helped stats (rel) min: <.01% max: 21.57% x̄: 2.08% x̃: 1.60%
HURT stats (abs)   min: 1 max: 2487 x̄: 181.78 x̃: 21
HURT stats (rel)   min: <.01% max: 40.66% x̄: 1.96% x̃: 0.37%
95% mean confidence interval for cycles value: -4.74 5.01
95% mean confidence interval for cycles %-change: -1.51% -1.37%
Inconclusive result (value mean confidence interval includes 0).

total spills in shared programs: 23401 -> 23407 (0.03%)
spills in affected programs: 248 -> 254 (2.42%)
helped: 2
HURT: 5

total fills in shared programs: 34850 -> 34845 (-0.01%)
fills in affected programs: 383 -> 378 (-1.31%)
helped: 2
HURT: 5

Ivy Bridge
total instructions in shared programs: 11975423 -> 11968117 (-0.06%)
instructions in affected programs: 845703 -> 838397 (-0.86%)
helped: 4071
HURT: 0
helped stats (abs) min: 1 max: 51 x̄: 1.79 x̃: 1
helped stats (rel) min: 0.08% max: 8.21% x̄: 1.04% x̃: 0.93%
95% mean confidence interval for instructions value: -1.87 -1.71
95% mean confidence interval for instructions %-change: -1.06% -1.02%
Instructions are helped.

total cycles in shared programs: 179674318 -> 179635552 (-0.02%)
cycles in affected programs: 5100065 -> 5061299 (-0.76%)
helped: 2650
HURT: 611
helped stats (abs) min: 1 max: 900 x̄: 21.85 x̃: 16
helped stats (rel) min: <.01% max: 21.55% x̄: 2.39% x̃: 1.40%
HURT stats (abs)   min: 1 max: 1841 x̄: 31.33 x̃: 6
HURT stats (rel)   min: <.01% max: 58.71% x̄: 1.64% x̃: 0.37%
95% mean confidence interval for cycles value: -14.14 -9.64
95% mean confidence interval for cycles %-change: -1.75% -1.52%
Cycles are helped.

LOST:   3
GAINED: 7

Sandy Bridge
total instructions in shared programs: 10828844 -> 10824272 (-0.04%)
instructions in affected programs: 525678 -> 521106 (-0.87%)
helped: 2386
HURT: 0
helped stats (abs) min: 1 max: 51 x̄: 1.92 x̃: 2
helped stats (rel) min: 0.11% max: 7.96% x̄: 1.05% x̃: 0.94%
95% mean confidence interval for instructions value: -2.04 -1.80
95% mean confidence interval for instructions %-change: -1.08% -1.03%
Instructions are helped.

total cycles in shared programs: 154024591 -> 154009894 (<.01%)
cycles in affected programs: 4005766 -> 3991069 (-0.37%)
helped: 1245
HURT: 506
helped stats (abs) min: 1 max: 585 x̄: 21.07 x̃: 16
helped stats (rel) min: 0.02% max: 11.57% x̄: 1.98% x̃: 0.83%
HURT stats (abs)   min: 1 max: 639 x̄: 22.81 x̃: 6
HURT stats (rel)   min: 0.01% max: 26.21% x̄: 1.07% x̃: 0.26%
95% mean confidence interval for cycles value: -10.57 -6.21
95% mean confidence interval for cycles %-change: -1.23% -0.97%
Cycles are helped.

GM45 and Iron Lake had similar results. (Iron Lake shown)
total instructions in shared programs: 8137248 -> 8135196 (-0.03%)
instructions in affected programs: 148322 -> 146270 (-1.38%)
helped: 992
HURT: 0
helped stats (abs) min: 1 max: 32 x̄: 2.07 x̃: 2
helped stats (rel) min: 0.41% max: 9.73% x̄: 1.74% x̃: 1.51%
95% mean confidence interval for instructions value: -2.16 -1.98
95% mean confidence interval for instructions %-change: -1.80% -1.67%
Instructions are helped.

total cycles in shared programs: 188583424 -> 188575724 (<.01%)
cycles in affected programs: 4409620 -> 4401920 (-0.17%)
helped: 956
HURT: 6
helped stats (abs) min: 2 max: 168 x̄: 8.09 x̃: 8
helped stats (rel) min: 0.04% max: 6.76% x̄: 0.27% x̃: 0.18%
HURT stats (abs)   min: 6 max: 6 x̄: 6.00 x̃: 6
HURT stats (rel)   min: 0.10% max: 0.10% x̄: 0.10% x̃: 0.10%
95% mean confidence interval for cycles value: -8.41 -7.60
95% mean confidence interval for cycles %-change: -0.29% -0.25%
Cycles are helped.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
 2019-08-05 20:14:14 -07:00
Ian Romanick
b77070e293 nir/algebraic: Use value range analysis to eliminate tautological compares 
It's only one application on one platform (Haswell) that's affected,
but spills and fills increase quite dramatically. :(

All Gen8+ platforms had similar results. (Ice Lake shown)
total instructions in shared programs: 16320850 -> 16297320 (-0.14%)
instructions in affected programs: 448012 -> 424482 (-5.25%)
helped: 1938
HURT: 0
helped stats (abs) min: 2 max: 264 x̄: 12.14 x̃: 10
helped stats (rel) min: 0.35% max: 43.75% x̄: 5.85% x̃: 5.38%
95% mean confidence interval for instructions value: -12.80 -11.48
95% mean confidence interval for instructions %-change: -5.99% -5.72%
Instructions are helped.

total cycles in shared programs: 367496943 -> 367221624 (-0.07%)
cycles in affected programs: 8557232 -> 8281913 (-3.22%)
helped: 1907
HURT: 26
helped stats (abs) min: 4 max: 12802 x̄: 147.21 x̃: 48
helped stats (rel) min: 0.03% max: 75.85% x̄: 5.55% x̃: 3.94%
HURT stats (abs)   min: 4 max: 1870 x̄: 208.23 x̃: 20
HURT stats (rel)   min: 0.16% max: 32.11% x̄: 8.31% x̃: 0.79%
95% mean confidence interval for cycles value: -165.38 -119.48
95% mean confidence interval for cycles %-change: -5.68% -5.04%
Cycles are helped.

LOST:   1
GAINED: 0

Haswell
total instructions in shared programs: 13374211 -> 13353925 (-0.15%)
instructions in affected programs: 349868 -> 329582 (-5.80%)
helped: 1669
HURT: 1
helped stats (abs) min: 1 max: 264 x̄: 12.57 x̃: 10
helped stats (rel) min: 0.12% max: 46.81% x̄: 6.86% x̃: 6.49%
HURT stats (abs)   min: 700 max: 700 x̄: 700.00 x̃: 700
HURT stats (rel)   min: 64.34% max: 64.34% x̄: 64.34% x̃: 64.34%
95% mean confidence interval for instructions value: -13.25 -11.04
95% mean confidence interval for instructions %-change: -7.01% -6.63%
Instructions are helped.

total cycles in shared programs: 375763544 -> 375535444 (-0.06%)
cycles in affected programs: 6932686 -> 6704586 (-3.29%)
helped: 1622
HURT: 48
helped stats (abs) min: 2 max: 12229 x̄: 148.31 x̃: 68
helped stats (rel) min: 0.06% max: 74.03% x̄: 5.94% x̃: 4.12%
HURT stats (abs)   min: 3 max: 7451 x̄: 259.44 x̃: 41
HURT stats (rel)   min: 0.05% max: 54.99% x̄: 8.52% x̃: 2.88%
95% mean confidence interval for cycles value: -159.86 -113.31
95% mean confidence interval for cycles %-change: -5.86% -5.18%
Cycles are helped.

total spills in shared programs: 23258 -> 23401 (0.61%)
spills in affected programs: 54 -> 197 (264.81%)
helped: 4
HURT: 2

total fills in shared programs: 34775 -> 34850 (0.22%)
fills in affected programs: 52 -> 127 (144.23%)
helped: 4
HURT: 1

LOST:   5
GAINED: 0

Ivy Bridge
total instructions in shared programs: 11996051 -> 11977964 (-0.15%)
instructions in affected programs: 346679 -> 328592 (-5.22%)
helped: 1508
HURT: 0
helped stats (abs) min: 2 max: 198 x̄: 11.99 x̃: 10
helped stats (rel) min: 0.26% max: 19.83% x̄: 5.73% x̃: 5.43%
95% mean confidence interval for instructions value: -12.65 -11.34
95% mean confidence interval for instructions %-change: -5.86% -5.60%
Instructions are helped.

total cycles in shared programs: 179891389 -> 179691339 (-0.11%)
cycles in affected programs: 7869479 -> 7669429 (-2.54%)
helped: 1485
HURT: 23
helped stats (abs) min: 1 max: 12615 x̄: 136.16 x̃: 54
helped stats (rel) min: 0.02% max: 71.84% x̄: 4.69% x̃: 3.49%
HURT stats (abs)   min: 1 max: 403 x̄: 93.48 x̃: 6
HURT stats (rel)   min: 0.04% max: 34.01% x̄: 8.68% x̃: 0.81%
95% mean confidence interval for cycles value: -154.59 -110.73
95% mean confidence interval for cycles %-change: -4.79% -4.19%
Cycles are helped.

Sandy Bridge
total instructions in shared programs: 10829247 -> 10828844 (<.01%)
instructions in affected programs: 21258 -> 20855 (-1.90%)
helped: 88
HURT: 0
helped stats (abs) min: 2 max: 17 x̄: 4.58 x̃: 5
helped stats (rel) min: 0.52% max: 3.92% x̄: 2.05% x̃: 2.21%
95% mean confidence interval for instructions value: -5.03 -4.13
95% mean confidence interval for instructions %-change: -2.21% -1.89%
Instructions are helped.

total cycles in shared programs: 154035437 -> 154024591 (<.01%)
cycles in affected programs: 430176 -> 419330 (-2.52%)
helped: 78
HURT: 10
helped stats (abs) min: 2 max: 4649 x̄: 143.06 x̃: 32
helped stats (rel) min: 0.05% max: 6.02% x̄: 2.03% x̃: 1.07%
HURT stats (abs)   min: 3 max: 265 x̄: 31.30 x̃: 6
HURT stats (rel)   min: 0.10% max: 8.67% x̄: 1.03% x̃: 0.21%
95% mean confidence interval for cycles value: -232.53 -13.97
95% mean confidence interval for cycles %-change: -2.13% -1.23%
Cycles are helped.

Iron Lake and GM45 had similar results. (Iron Lake shown)
total instructions in shared programs: 8137402 -> 8137248 (<.01%)
instructions in affected programs: 2280 -> 2126 (-6.75%)
helped: 10
HURT: 0
helped stats (abs) min: 12 max: 19 x̄: 15.40 x̃: 15
helped stats (rel) min: 3.90% max: 11.73% x̄: 7.19% x̃: 6.95%
95% mean confidence interval for instructions value: -17.69 -13.11
95% mean confidence interval for instructions %-change: -8.99% -5.39%
Instructions are helped.

total cycles in shared programs: 188538716 -> 188583424 (0.02%)
cycles in affected programs: 69326 -> 114034 (64.49%)
helped: 0
HURT: 10
HURT stats (abs)   min: 2068 max: 7686 x̄: 4470.80 x̃: 4870
HURT stats (rel)   min: 27.20% max: 173.66% x̄: 69.55% x̃: 59.41%
95% mean confidence interval for cycles value: 2830.86 6110.74
95% mean confidence interval for cycles %-change: 39.18% 99.91%
Cycles are HURT.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
 2019-08-05 20:14:13 -07:00
Ian Romanick
96fcb3f95b nir/algebraic: Use value range analysis to eliminate tautological compares not used by if-statements 
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>
 2019-08-05 20:14:13 -07:00
Ian Romanick
d24edb4b8c nir/algebraic: Simplify some comparisons like a+constant < constant 
v2: Remove unsafe integer versions of the optimizations.  This change
had no effect on shader-db results.  Suggested by Caio.

All Gen6+ platforms had similar results. (Ice Lake shown)
total instructions in shared programs: 16333713 -> 16332631 (<.01%)
instructions in affected programs: 258112 -> 257030 (-0.42%)
helped: 1275
HURT: 407
helped stats (abs) min: 1 max: 7 x̄: 1.17 x̃: 1
helped stats (rel) min: 0.20% max: 8.33% x̄: 1.33% x̃: 0.86%
HURT stats (abs)   min: 1 max: 2 x̄: 1.00 x̃: 1
HURT stats (rel)   min: 0.11% max: 2.94% x̄: 0.98% x̃: 0.98%
95% mean confidence interval for instructions value: -0.70 -0.59
95% mean confidence interval for instructions %-change: -0.84% -0.70%
Instructions are helped.

total cycles in shared programs: 367596791 -> 367601268 (<.01%)
cycles in affected programs: 3420062 -> 3424539 (0.13%)
helped: 1553
HURT: 783
helped stats (abs) min: 1 max: 742 x̄: 24.36 x̃: 6
helped stats (rel) min: 0.05% max: 21.12% x̄: 1.47% x̃: 0.65%
HURT stats (abs)   min: 1 max: 557 x̄: 54.04 x̃: 14
HURT stats (rel)   min: 0.01% max: 33.66% x̄: 3.36% x̃: 1.43%
95% mean confidence interval for cycles value: -1.60 5.43
95% mean confidence interval for cycles %-change: -0.03% 0.33%
Inconclusive result (value mean confidence interval includes 0).

LOST:   0
GAINED: 2

Iron Lake
total instructions in shared programs: 8137992 -> 8137874 (<.01%)
instructions in affected programs: 17501 -> 17383 (-0.67%)
helped: 104
HURT: 2
helped stats (abs) min: 1 max: 2 x̄: 1.17 x̃: 1
helped stats (rel) min: 0.25% max: 2.63% x̄: 0.87% x̃: 0.72%
HURT stats (abs)   min: 2 max: 2 x̄: 2.00 x̃: 2
HURT stats (rel)   min: 0.45% max: 0.45% x̄: 0.45% x̃: 0.45%
95% mean confidence interval for instructions value: -1.22 -1.00
95% mean confidence interval for instructions %-change: -0.94% -0.76%
Instructions are helped.

total cycles in shared programs: 188540038 -> 188539650 (<.01%)
cycles in affected programs: 704574 -> 704186 (-0.06%)
helped: 125
HURT: 84
helped stats (abs) min: 2 max: 96 x̄: 6.45 x̃: 4
helped stats (rel) min: <.01% max: 3.47% x̄: 0.42% x̃: 0.25%
HURT stats (abs)   min: 2 max: 58 x̄: 4.98 x̃: 4
HURT stats (rel)   min: 0.01% max: 2.75% x̄: 0.36% x̃: 0.33%
95% mean confidence interval for cycles value: -3.20 -0.52
95% mean confidence interval for cycles %-change: -0.19% -0.03%
Cycles are helped.

GM45
total instructions in shared programs: 5008889 -> 5008830 (<.01%)
instructions in affected programs: 8824 -> 8765 (-0.67%)
helped: 52
HURT: 1
helped stats (abs) min: 1 max: 2 x̄: 1.17 x̃: 1
helped stats (rel) min: 0.25% max: 2.38% x̄: 0.86% x̃: 0.72%
HURT stats (abs)   min: 2 max: 2 x̄: 2.00 x̃: 2
HURT stats (rel)   min: 0.45% max: 0.45% x̄: 0.45% x̃: 0.45%
95% mean confidence interval for instructions value: -1.27 -0.95
95% mean confidence interval for instructions %-change: -0.96% -0.71%
Instructions are helped.

total cycles in shared programs: 128969426 -> 128969128 (<.01%)
cycles in affected programs: 399798 -> 399500 (-0.07%)
helped: 74
HURT: 30
helped stats (abs) min: 2 max: 22 x̄: 6.76 x̃: 6
helped stats (rel) min: <.01% max: 1.83% x̄: 0.46% x̃: 0.29%
HURT stats (abs)   min: 2 max: 58 x̄: 6.73 x̃: 6
HURT stats (rel)   min: 0.06% max: 2.75% x̄: 0.42% x̃: 0.21%
95% mean confidence interval for cycles value: -4.60 -1.14
95% mean confidence interval for cycles %-change: -0.32% -0.08%
Cycles are helped.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
 2019-08-05 20:14:13 -07:00
Ian Romanick
7c64cbf49d nir/algebraic: Recognize (a < 0 || 0 < b) as min(a, -b) < 0 
Similar to commit 97e6c1b9 and f5cf74d8ba.

First apply 0 < b => -b < 0 to get (a < 0 || -b < 0), then apply some
pre-existing rules to get min(a, -b) < 0.

v2: Substantially update the comment explaining the use of is_used_once
and the duplication of patterns.  Suggested by Caio.  Also, while flt
and fge are not commutative, ior and iand are.  Half of the original
patterns were redundant, so delete them.  As alternate justification for
deleting them, fmin(a, -b) < 0 <=> 0 < fmax(-a, b).  Proof left as an
exercise for the reader.

All Gen7+ platforms had similar results. (Ice Lake shown)
total instructions in shared programs: 16333789 -> 16333713 (<.01%)
instructions in affected programs: 11424 -> 11348 (-0.67%)
helped: 32
HURT: 0
helped stats (abs) min: 1 max: 7 x̄: 2.38 x̃: 2
helped stats (rel) min: 0.20% max: 1.67% x̄: 0.76% x̃: 0.69%
95% mean confidence interval for instructions value: -3.03 -1.72
95% mean confidence interval for instructions %-change: -0.89% -0.62%
Instructions are helped.

total cycles in shared programs: 367598295 -> 367596791 (<.01%)
cycles in affected programs: 141414 -> 139910 (-1.06%)
helped: 23
HURT: 6
helped stats (abs) min: 3 max: 386 x̄: 72.52 x̃: 20
helped stats (rel) min: 0.15% max: 4.86% x̄: 1.01% x̃: 0.76%
HURT stats (abs)   min: 4 max: 88 x̄: 27.33 x̃: 12
HURT stats (rel)   min: 0.22% max: 3.95% x̄: 1.08% x̃: 0.59%
95% mean confidence interval for cycles value: -93.51 -10.21
95% mean confidence interval for cycles %-change: -1.10% -0.05%
Cycles are helped.

total instructions in shared programs: 10830836 -> 10830779 (<.01%)
instructions in affected programs: 6895 -> 6838 (-0.83%)
helped: 12
HURT: 0
helped stats (abs) min: 1 max: 14 x̄: 4.75 x̃: 1
helped stats (rel) min: 0.14% max: 1.61% x̄: 0.65% x̃: 0.33%
95% mean confidence interval for instructions value: -8.46 -1.04
95% mean confidence interval for instructions %-change: -1.03% -0.27%
Instructions are helped.

total cycles in shared programs: 154028477 -> 154032740 (<.01%)
cycles in affected programs: 178433 -> 182696 (2.39%)
helped: 3
HURT: 9
helped stats (abs) min: 3 max: 20 x̄: 11.00 x̃: 10
helped stats (rel) min: 0.07% max: 0.20% x̄: 0.12% x̃: 0.09%
HURT stats (abs)   min: 27 max: 1415 x̄: 477.33 x̃: 262
HURT stats (rel)   min: 0.22% max: 6.45% x̄: 2.49% x̃: 1.76%
95% mean confidence interval for cycles value: 28.68 681.82
95% mean confidence interval for cycles %-change: 0.37% 3.30%
Cycles are HURT.

Iron Lake
total instructions in shared programs: 8137966 -> 8137992 (<.01%)
instructions in affected programs: 3281 -> 3307 (0.79%)
helped: 0
HURT: 6
HURT stats (abs)   min: 3 max: 7 x̄: 4.33 x̃: 3
HURT stats (rel)   min: 0.63% max: 1.01% x̄: 0.76% x̃: 0.64%
95% mean confidence interval for instructions value: 2.17 6.50
95% mean confidence interval for instructions %-change: 0.56% 0.96%
Instructions are HURT.

total cycles in shared programs: 188539386 -> 188540038 (<.01%)
cycles in affected programs: 103826 -> 104478 (0.63%)
helped: 0
HURT: 7
HURT stats (abs)   min: 16 max: 218 x̄: 93.14 x̃: 80
HURT stats (rel)   min: 0.14% max: 0.95% x̄: 0.53% x̃: 0.46%
95% mean confidence interval for cycles value: 10.26 176.02
95% mean confidence interval for cycles %-change: 0.24% 0.81%
Cycles are HURT.

GM45
total instructions in shared programs: 5008876 -> 5008889 (<.01%)
instructions in affected programs: 1645 -> 1658 (0.79%)
helped: 0
HURT: 3
HURT stats (abs)   min: 3 max: 7 x̄: 4.33 x̃: 3
HURT stats (rel)   min: 0.63% max: 1.00% x̄: 0.76% x̃: 0.63%

total cycles in shared programs: 128968950 -> 128969426 (<.01%)
cycles in affected programs: 64854 -> 65330 (0.73%)
helped: 0
HURT: 4
HURT stats (abs)   min: 18 max: 218 x̄: 119.00 x̃: 120
HURT stats (rel)   min: 0.14% max: 0.95% x̄: 0.60% x̃: 0.66%
95% mean confidence interval for cycles value: -62.92 300.92
95% mean confidence interval for cycles %-change: -0.05% 1.26%
Inconclusive result (value mean confidence interval includes 0).

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
 2019-08-05 20:14:13 -07:00
Ian Romanick
92b75c126b nir/algebraic: Replace checks that a value is between (or not) [0, 1] 
v2: Add an extra line to one of the proofs.  Suggested by Caio.

All Gen7+ platforms had similar results. (Ice Lake shown)
total instructions in shared programs: 16329772 -> 16329427 (<.01%)
instructions in affected programs: 41980 -> 41635 (-0.82%)
helped: 110
HURT: 0
helped stats (abs) min: 1 max: 20 x̄: 3.14 x̃: 2
helped stats (rel) min: 0.19% max: 5.56% x̄: 1.12% x̃: 0.94%
95% mean confidence interval for instructions value: -4.10 -2.17
95% mean confidence interval for instructions %-change: -1.28% -0.96%
Instructions are helped.

total cycles in shared programs: 367551273 -> 367549979 (<.01%)
cycles in affected programs: 492462 -> 491168 (-0.26%)
helped: 76
HURT: 25
helped stats (abs) min: 1 max: 400 x̄: 42.86 x̃: 12
helped stats (rel) min: 0.06% max: 10.72% x̄: 1.23% x̃: 0.75%
HURT stats (abs)   min: 2 max: 730 x̄: 78.52 x̃: 16
HURT stats (rel)   min: 0.17% max: 6.89% x̄: 2.08% x̃: 1.23%
95% mean confidence interval for cycles value: -37.79 12.16
95% mean confidence interval for cycles %-change: -0.90% 0.07%
Inconclusive result (value mean confidence interval includes 0).

LOST:   0
GAINED: 2

Sandy Bridge
total instructions in shared programs: 10831115 -> 10830836 (<.01%)
instructions in affected programs: 37830 -> 37551 (-0.74%)
helped: 70
HURT: 0
helped stats (abs) min: 1 max: 20 x̄: 3.99 x̃: 2
helped stats (rel) min: 0.33% max: 7.14% x̄: 1.21% x̃: 0.97%
95% mean confidence interval for instructions value: -5.47 -2.50
95% mean confidence interval for instructions %-change: -1.49% -0.92%
Instructions are helped.

total cycles in shared programs: 154029323 -> 154028477 (<.01%)
cycles in affected programs: 247909 -> 247063 (-0.34%)
helped: 52
HURT: 6
helped stats (abs) min: 2 max: 254 x̄: 25.81 x̃: 4
helped stats (rel) min: 0.07% max: 4.39% x̄: 0.81% x̃: 0.19%
HURT stats (abs)   min: 4 max: 403 x̄: 82.67 x̃: 8
HURT stats (rel)   min: 0.18% max: 1.60% x̄: 0.71% x̃: 0.53%
95% mean confidence interval for cycles value: -34.83 5.65
95% mean confidence interval for cycles %-change: -0.98% -0.32%
Inconclusive result (value mean confidence interval includes 0).

Iron Lake
total instructions in shared programs: 8138007 -> 8137966 (<.01%)
instructions in affected programs: 4060 -> 4019 (-1.01%)
helped: 31
HURT: 0
helped stats (abs) min: 1 max: 2 x̄: 1.32 x̃: 1
helped stats (rel) min: 0.68% max: 8.33% x̄: 1.45% x̃: 0.90%
95% mean confidence interval for instructions value: -1.50 -1.15
95% mean confidence interval for instructions %-change: -2.11% -0.79%
Instructions are helped.

total cycles in shared programs: 188539492 -> 188539386 (<.01%)
cycles in affected programs: 26280 -> 26174 (-0.40%)
helped: 25
HURT: 0
helped stats (abs) min: 2 max: 8 x̄: 4.24 x̃: 4
helped stats (rel) min: 0.08% max: 2.11% x̄: 0.54% x̃: 0.50%
95% mean confidence interval for cycles value: -5.08 -3.40
95% mean confidence interval for cycles %-change: -0.70% -0.37%
Cycles are helped.

GM45
total instructions in shared programs: 5008897 -> 5008876 (<.01%)
instructions in affected programs: 2096 -> 2075 (-1.00%)
helped: 16
HURT: 0
helped stats (abs) min: 1 max: 2 x̄: 1.31 x̃: 1
helped stats (rel) min: 0.68% max: 7.69% x̄: 1.41% x̃: 0.89%
95% mean confidence interval for instructions value: -1.57 -1.06
95% mean confidence interval for instructions %-change: -2.32% -0.49%
Instructions are helped.

total cycles in shared programs: 128969020 -> 128968950 (<.01%)
cycles in affected programs: 18490 -> 18420 (-0.38%)
helped: 15
HURT: 0
helped stats (abs) min: 2 max: 8 x̄: 4.67 x̃: 4
helped stats (rel) min: 0.08% max: 2.11% x̄: 0.51% x̃: 0.48%
95% mean confidence interval for cycles value: -6.03 -3.30
95% mean confidence interval for cycles %-change: -0.78% -0.24%
Cycles are helped.

Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
 2019-08-05 20:14:13 -07:00
Erico Nunes
b3676a6548 nir/algebraic: rename lower_bitshift to lower_bitops 
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>
 2019-07-31 23:06:04 +02:00