Alyssa Rosenzweig
e2f7871bcf
pan/bi: Extract bi_finalize_nir
...
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14154 >
2021-12-20 18:21:41 +00:00
Alyssa Rosenzweig
52fe998aa6
panfrost: Track preloaded registers
...
We already collect this information. We may as well make use of it.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14154 >
2021-12-20 18:21:40 +00:00
Dave Airlie
d051854cca
treewide: drop mtypes/macros includes from main
...
These aren't required in lots of places, so remove them.
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14127 >
2021-12-08 22:14:45 +00:00
Alyssa Rosenzweig
a4d3a29647
pan/bi: Enable dual texture fusing pass
...
Everything is in place for it now -- ship it! Our Bifrost cycle model is
coarse, so take the shader-db results with a massive spoonful of salt:
total instructions in shared programs: 107504 -> 107252 (-0.23%)
instructions in affected programs: 39692 -> 39440 (-0.63%)
helped: 191
HURT: 1
helped stats (abs) min: 1.0 max: 20.0 x̄: 1.32 x̃: 1
helped stats (rel) min: 0.11% max: 9.52% x̄: 1.21% x̃: 0.98%
HURT stats (abs) min: 1.0 max: 1.0 x̄: 1.00 x̃: 1
HURT stats (rel) min: 2.04% max: 2.04% x̄: 2.04% x̃: 2.04%
95% mean confidence interval for instructions value: -1.60 -1.02
95% mean confidence interval for instructions %-change: -1.37% -1.01%
Instructions are helped.
total tuples in shared programs: 89864 -> 89664 (-0.22%)
tuples in affected programs: 27787 -> 27587 (-0.72%)
helped: 146
HURT: 6
helped stats (abs) min: 1.0 max: 17.0 x̄: 1.54 x̃: 1
helped stats (rel) min: 0.14% max: 15.38% x̄: 1.83% x̃: 1.25%
HURT stats (abs) min: 1.0 max: 11.0 x̄: 4.17 x̃: 2
HURT stats (rel) min: 0.54% max: 3.55% x̄: 1.29% x̃: 0.87%
95% mean confidence interval for tuples value: -1.64 -0.99
95% mean confidence interval for tuples %-change: -2.06% -1.36%
Tuples are helped.
total clauses in shared programs: 18253 -> 18044 (-1.15%)
clauses in affected programs: 5127 -> 4918 (-4.08%)
helped: 164
HURT: 1
helped stats (abs) min: 1.0 max: 19.0 x̄: 1.28 x̃: 1
helped stats (rel) min: 0.78% max: 28.57% x̄: 6.73% x̃: 5.88%
HURT stats (abs) min: 1.0 max: 1.0 x̄: 1.00 x̃: 1
HURT stats (rel) min: 2.94% max: 2.94% x̄: 2.94% x̃: 2.94%
95% mean confidence interval for clauses value: -1.50 -1.04
95% mean confidence interval for clauses %-change: -7.42% -5.91%
Clauses are helped.
total cycles in shared programs: 8118.54 -> 8103.88 (-0.18%)
cycles in affected programs: 414.96 -> 400.29 (-3.53%)
helped: 43
HURT: 27
helped stats (abs) min: 0.041665999999999315 max: 4.375 x̄: 0.41 x̃: 0
helped stats (rel) min: 0.24% max: 50.00% x̄: 11.49% x̃: 5.26%
HURT stats (abs) min: 0.041665999999999315 max: 1.1666639999999973 x̄:
0.10 x̃: 0
HURT stats (rel) min: 0.43% max: 4.71% x̄: 1.42% x̃: 1.28%
95% mean confidence interval for cycles value: -0.35 -0.07
95% mean confidence interval for cycles %-change: -9.50% -3.53%
Cycles are helped.
total arith in shared programs: 3375.67 -> 3376.42 (0.02%)
arith in affected programs: 345.29 -> 346.04 (0.22%)
helped: 24
HURT: 32
helped stats (abs) min: 0.041665999999999315 max: 0.5833329999999997 x̄:
0.09 x̃: 0
helped stats (rel) min: 0.24% max: 14.29% x̄: 2.82% x̃: 1.50%
HURT stats (abs) min: 0.041665999999999315 max: 1.1666639999999973 x̄:
0.09 x̃: 0
HURT stats (rel) min: 0.43% max: 4.71% x̄: 1.42% x̃: 1.28%
95% mean confidence interval for arith value: -0.04 0.07
95% mean confidence interval for arith %-change: -1.19% 0.39%
Inconclusive result (value mean confidence interval includes 0).
total texture in shared programs: 1275 -> 1157 (-9.25%)
texture in affected programs: 725.50 -> 607.50 (-16.26%)
helped: 192
HURT: 0
helped stats (abs) min: 0.5 max: 10.0 x̄: 0.61 x̃: 0
helped stats (rel) min: 2.86% max: 50.00% x̄: 25.20% x̃: 25.00%
95% mean confidence interval for texture value: -0.72 -0.51
95% mean confidence interval for texture %-change: -27.12% -23.27%
Texture are helped.
total vary in shared programs: 537.88 -> 536.12 (-0.33%)
vary in affected programs: 2.75 -> 1 (-63.64%)
helped: 1
HURT: 0
total quadwords in shared programs: 79762 -> 79681 (-0.10%)
quadwords in affected programs: 10261 -> 10180 (-0.79%)
helped: 59
HURT: 18
helped stats (abs) min: 1.0 max: 14.0 x̄: 1.88 x̃: 1
helped stats (rel) min: 0.38% max: 8.20% x̄: 1.95% x̃: 1.43%
HURT stats (abs) min: 1.0 max: 4.0 x̄: 1.67 x̃: 1
HURT stats (rel) min: 0.46% max: 8.89% x̄: 2.22% x̃: 1.21%
95% mean confidence interval for quadwords value: -1.57 -0.53
95% mean confidence interval for quadwords %-change: -1.59% -0.37%
Quadwords are helped.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13723 >
2021-11-12 16:30:02 +00:00
Alyssa Rosenzweig
98c1b3e7e1
pan/bi: Use BIFROST_TEXTURE_OPERATION_SINGLE enum
...
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13723 >
2021-11-12 16:30:02 +00:00
Alyssa Rosenzweig
7dc90b68d9
pan/bi: Add second destination to TEXC
...
Used to model dual texturing, which writes to separate sets of staging
registers.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13723 >
2021-11-12 16:30:02 +00:00
Alyssa Rosenzweig
c00e7b729f
pan/bi: Optimize abs(derivative)
...
We implement fine derivatives as:
broadcast(x, (lane & ~1) + 1) - broadcast(x, lane & ~1)
Most of the complexity is to get the right sign. If we can ignore the
sign, we can generate the simpler code:
broadcast(x, lane ^ 1) - lane
This is a particular win on v7+ where the broadcast instruction (CLPER)
can do `lane ^ value` for free. However, even on v6 where we lower to an
explicit XOR instruction, it's still a win.
The limiting case is fwidth. The fragment shader
gl_FragColor = fwidth(vec4_varying);
has the following results on v6, v7, and v9:
G72 (-26% instructions, -43% cycles):
38 inst, 30 tuples, 5 clauses, 1.166667 cycles, 1.166667 arith, 28 quadwords
28 inst, 19 tuples, 4 clauses, 0.666667 cycles, 0.666667 arith, 19 quadwords
G76 (-37% instructions, -54% cycles):
38 inst, 30 tuples, 5 clauses, 1.166667 cycles, 1.166667 arith, 28 quadwords
24 inst, 16 tuples, 4 clauses, 0.541667 cycles, 0.541667 arith, 18 quadwords
G78 (-40% instructions, -56% cycles):
40 inst, 1.125000 cycles, 0.250000 fma, 0.109375 cvt, 1.125000 sfu, 20 quadwords
24 inst, 0.500000 cycles, 0.250000 fma, 0.015625 cvt, 0.500000 sfu, 12 quadwords
shader-db tells a similar story -- most shaders are unaffected, but a
shader that uses fwidth has a 20% reduction in cycle count:
instructions helped: shaders/tesseract/488.shader_test MESA_SHADER_FRAGMENT: 264 -> 262 (-0.76%)
instructions helped: shaders/chromeos/109-1.shader_test MESA_SHADER_FRAGMENT: 36 -> 28 (-22.22%)
tuples helped: shaders/chromeos/109-1.shader_test MESA_SHADER_FRAGMENT: 27 -> 22 (-18.52%)
tuples HURT: shaders/tesseract/488.shader_test MESA_SHADER_FRAGMENT: 211 -> 212 (0.47%)
clauses HURT: shaders/tesseract/488.shader_test MESA_SHADER_FRAGMENT: 32 -> 33 (3.12%)
cycles helped: shaders/chromeos/109-1.shader_test MESA_SHADER_FRAGMENT: 1 -> 0.79 (-20.83%)
arith helped: shaders/chromeos/109-1.shader_test MESA_SHADER_FRAGMENT: 1 -> 0.79 (-20.83%)
quadwords helped: shaders/chromeos/109-1.shader_test MESA_SHADER_FRAGMENT: 31 -> 28 (-9.68%)
quadwords HURT: shaders/tesseract/488.shader_test MESA_SHADER_FRAGMENT: 176 -> 178 (1.14%)
total instructions in shared programs: 148370 -> 148360 (<.01%)
instructions in affected programs: 300 -> 290 (-3.33%)
helped: 2
HURT: 0
total tuples in shared programs: 124188 -> 124184 (<.01%)
tuples in affected programs: 238 -> 234 (-1.68%)
helped: 1
HURT: 1
helped stats (abs) min: 5.0 max: 5.0 x̄: 5.00 x̃: 5
helped stats (rel) min: 18.52% max: 18.52% x̄: 18.52% x̃: 18.52%
HURT stats (abs) min: 1.0 max: 1.0 x̄: 1.00 x̃: 1
HURT stats (rel) min: 0.47% max: 0.47% x̄: 0.47% x̃: 0.47%
total clauses in shared programs: 25692 -> 25693 (<.01%)
clauses in affected programs: 32 -> 33 (3.12%)
helped: 0
HURT: 1
total cycles in shared programs: 12132.04 -> 12131.83 (<.01%)
cycles in affected programs: 1 -> 0.79 (-20.83%)
helped: 1
HURT: 0
total arith in shared programs: 4623.75 -> 4623.54 (<.01%)
arith in affected programs: 1 -> 0.79 (-20.83%)
helped: 1
HURT: 0
total quadwords in shared programs: 110386 -> 110385 (<.01%)
quadwords in affected programs: 207 -> 206 (-0.48%)
helped: 1
HURT: 1
helped stats (abs) min: 3.0 max: 3.0 x̄: 3.00 x̃: 3
helped stats (rel) min: 9.68% max: 9.68% x̄: 9.68% x̃: 9.68%
HURT stats (abs) min: 2.0 max: 2.0 x̄: 2.00 x̃: 2
HURT stats (rel) min: 1.14% max: 1.14% x̄: 1.14% x̃: 1.14%
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12332 >
2021-10-06 00:40:57 +00:00
Boris Brezillon
2f68c58767
pan/blend: Allow passing blend constants through a sysval
...
This is needed to allow lowering blend operations in fragment shaders
when the blend operation uses dynamic blend constants.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13060 >
2021-09-30 16:54:42 +02:00
Boris Brezillon
f256ec2a88
pan/bi: Fix 1DArray image coordinate retrieval
...
In NIR image instructions, the array index always follows the last
image coordinate, meaning that array index is in coord.y for 1D
arrays. But the current panfrost ABI wants it in coord.z regardless of
the image dimension.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13078 >
2021-09-28 18:44:05 +00:00
Boris Brezillon
2dd13257c8
pan/bi: Allow passing RT conversion descriptors to fragment shaders
...
The Vulkan copy shaders sometimes need to partially write a texel and
issue a load on the FRAGOUT variable in that case, but they do know
the format of the tile buffer in advance in that case. Let's not add an
RT_CONVERSION sysval if we can avoid it.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Acked-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12095 >
2021-09-21 14:59:38 +02:00
Boris Brezillon
04d6c593d0
pan/bi: Relax check on 8bit swizzles
...
Allow extracting components Y, Z or W from an 8bit vector.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12095 >
2021-09-21 14:58:54 +02:00
Alyssa Rosenzweig
61c8e39649
pan/bi: Use CLPER_V6 on Mali G31
...
Apparently, CLPER_V7 is missing from Mali G31, but CLPER_V6 works. Fixes
INSTR_INVALID_ENC faults and failures in
dEQP-GLES3.functional.shaders.derivate.* on Dvalin.
Technically not an errata but an implementation difference. I suspect
Mali G51 will need this as well, should we ever allowlist it.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12478 >
2021-08-20 20:13:27 +00:00
Alyssa Rosenzweig
bfd3ae35c9
pan/bi: Use ST_TILE for multisampled blend output
...
ST_TILE lets us specify an explicit sample, whereas BLEND replicates to
all samples. This fully fixes the interaction between blend shaders and
multisampling on Bifrost, manifesting as
dEQP-GLES3.functional.fragment_ops.random.* failures with the
configuration rgba8888d24s8ms4.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12478 >
2021-08-20 20:13:27 +00:00
Alyssa Rosenzweig
16394dc71a
pan/bi: Set the sample ID for blend shader LD_TILE
...
Use the explicit sample mode and set the sample ID in the pixel indices
structure to the current sample ID. This fixes tilebuffer loads in blend
shaders on multisampled framebuffers.
Make sure the new routine is broken out to a helper for use with ST_TILE
in the next commit.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12478 >
2021-08-20 20:13:27 +00:00
Alyssa Rosenzweig
9f19a883bc
pan/bi: Extract load_sample_id to a helper
...
Will be reused in the next commit.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12478 >
2021-08-20 20:13:27 +00:00
Icecream95
a9ab168e16
pan/bi,pan/mdg: Fix memory leak of hash tables
...
Despite being created with a ralloc context, some memory is still
leaked when not manually destroying hash tables.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12376 >
2021-08-16 14:18:36 +00:00
Alyssa Rosenzweig
d74ab1e4d9
pan/bi: Fuse DISCARD with conditions
...
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12205 >
2021-08-11 14:59:26 -04:00
Alyssa Rosenzweig
ac636f5adb
pan/bi: Use FCLAMP pseudo op for clamp prop
...
Map nir_op_fsat/etc to FCLAMP pseudo ops, instead of FADD. There are
significantly fewer knobs on FCLAMP, meaning significantly fewer things
to get wrong.
This fixes two(!) classes of bugs:
* Swizzles (failing to lower/compose swizzles on clamps)
* Numerical bugs (incorrectly treating +0.0 as an additive identity)
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12205 >
2021-08-11 14:59:26 -04:00
Alyssa Rosenzweig
89e452883a
pan/bi: Use FABSNEG pseudo ops for modifier prop
...
Simplifies pattern matching. This commit by itself fixes multiple
numerical issues -- the previous fabsneg check failed to check the round
mode or the sign of the zero. That will break Vulkan/OpenCL.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Cc: mesa-stable
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12205 >
2021-08-11 14:59:26 -04:00
Alyssa Rosenzweig
ec76119dfb
pan/bi: Constant fold texturing lowerings
...
This ensures we can constant fold the ALU ops used to lower:
* explicit LOD calculations
* array textures
* texture offsets
* multisample indices
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12205 >
2021-08-11 14:59:24 -04:00
Alyssa Rosenzweig
3958f00215
pan/bi: Add a noopt debug option
...
To rule out buggy optimization passes when debugging.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12328 >
2021-08-11 18:00:45 +00:00
Alyssa Rosenzweig
ff03f096bf
pan/bi: Make bi_opt_push_ubo optional
...
It's an optimization pass -- omitting it should not cause MMU faults
(!). Make sure the UBO push mask is set regardless of whether the pass
is called, and just call the pass when required.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12328 >
2021-08-11 18:00:45 +00:00
Icecream95
ee2bb57f1e
pan/bi: Use the computed scale for fexp NaN propagation
...
This makes pow(NaN, x) return NaN rather than 1.0.
Fixes: 499397700c
("pan/bi: Don't lower fpow")
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5189
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12269 >
2021-08-10 22:42:08 +00:00
Timothy Arceri
a9ed4538ab
nir: add indirect loop unrolling to compiler options
...
This is where it should be rather than having to pass it into the
optimisation pass every time.
It also allows us to call the loop analysis pass without having to
duplicate these options which we will do later in this series.
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12064 >
2021-08-03 10:54:50 +00:00
Alyssa Rosenzweig
1e29f57b3a
pan/bi: Validate the live set starts empty
...
Otherwise there is an uninitialized read, and the register allocation
will fail. (In the sense of failing a precondition. This manifests as
synthetic interference leading to higher register pressure and useless
moves. The allocation itself is ok, but it indicates a real bug.)
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12130 >
2021-07-29 23:40:46 +00:00
Alyssa Rosenzweig
c4f8b52e06
pan/bi: Lower fragment output with <4 components
...
This avoids undefined behaviour in the shader, which will fail
validation added later in the series. shader-db results are neglible --
the extra moves required in a few cases are cancelled out by the extra
moves eliminated by allowing register allocation to work properly.
total instructions in shared programs: 146903 -> 146907 (<.01%)
instructions in affected programs: 33 -> 37 (12.12%)
helped: 0
HURT: 1
total tuples in shared programs: 123616 -> 123613 (<.01%)
tuples in affected programs: 764 -> 761 (-0.39%)
helped: 6
HURT: 4
helped stats (abs) min: 1.0 max: 4.0 x̄: 1.67 x̃: 1
helped stats (rel) min: 0.54% max: 5.88% x̄: 2.64% x̃: 1.86%
HURT stats (abs) min: 1.0 max: 2.0 x̄: 1.75 x̃: 2
HURT stats (rel) min: 4.55% max: 13.33% x̄: 8.57% x̃: 8.19%
95% mean confidence interval for tuples value: -1.73 1.13
95% mean confidence interval for tuples %-change: -2.72% 6.41%
Inconclusive result (value mean confidence interval includes 0).
total clauses in shared programs: 25656 -> 25654 (<.01%)
clauses in affected programs: 43 -> 41 (-4.65%)
helped: 2
HURT: 1
helped stats (abs) min: 1.0 max: 2.0 x̄: 1.50 x̃: 1
helped stats (rel) min: 6.25% max: 12.50% x̄: 9.38% x̃: 9.38%
HURT stats (abs) min: 1.0 max: 1.0 x̄: 1.00 x̃: 1
HURT stats (rel) min: 33.33% max: 33.33% x̄: 33.33% x̃: 33.33%
total cycles in shared programs: 12114.21 -> 12114.12 (<.01%)
cycles in affected programs: 27.42 -> 27.33 (-0.30%)
helped: 4
HURT: 3
helped stats (abs) min: 0.04166700000000034 max: 0.08333299999999966 x̄: 0.06 x̃: 0
helped stats (rel) min: 0.57% max: 1.59% x̄: 1.02% x̃: 0.96%
HURT stats (abs) min: 0.0416669999999999 max: 0.08333299999999999 x̄: 0.06 x̃: 0
HURT stats (rel) min: 4.17% max: 16.67% x̄: 8.80% x̃: 5.56%
95% mean confidence interval for cycles value: -0.07 0.05
95% mean confidence interval for cycles %-change: -2.90% 9.27%
Inconclusive result (value mean confidence interval includes 0).
total arith in shared programs: 4601.08 -> 4601.04 (<.01%)
arith in affected programs: 29 -> 28.96 (-0.14%)
helped: 6
HURT: 4
helped stats (abs) min: 0.04166700000000001 max: 0.08333299999999966 x̄: 0.06 x̃: 0
helped stats (rel) min: 0.57% max: 10.00% x̄: 3.63% x̃: 1.39%
HURT stats (abs) min: 0.04166700000000001 max: 0.08333399999999991 x̄: 0.07 x̃: 0
HURT stats (rel) min: 5.56% max: 16.67% x̄: 10.85% x̃: 10.60%
95% mean confidence interval for arith value: -0.05 0.05
95% mean confidence interval for arith %-change: -3.95% 8.28%
Inconclusive result (value mean confidence interval includes 0).
total quadwords in shared programs: 110008 -> 110002 (<.01%)
quadwords in affected programs: 1090 -> 1084 (-0.55%)
helped: 11
HURT: 8
helped stats (abs) min: 1.0 max: 7.0 x̄: 2.18 x̃: 1
helped stats (rel) min: 0.61% max: 13.16% x̄: 4.07% x̃: 1.82%
HURT stats (abs) min: 1.0 max: 6.0 x̄: 2.25 x̃: 1
HURT stats (rel) min: 3.70% max: 42.86% x̄: 12.55% x̃: 7.50%
95% mean confidence interval for quadwords value: -1.76 1.13
95% mean confidence interval for quadwords %-change: -2.95% 8.81%
Inconclusive result (value mean confidence interval includes 0).
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12130 >
2021-07-29 23:40:46 +00:00
Alyssa Rosenzweig
a5727909e1
pan/bi: Rename CLPER_V7 back to CLPER
...
v6 is really the oddball here. CLPER on v9 supports a superset of v7.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12026 >
2021-07-28 00:26:06 +00:00
Alyssa Rosenzweig
be95198de5
pan/bi: DCE after bifrost_nir_lower_algebraic_late
...
Needed for sat_signed to fuse, since we run modifier prop before backend
DCE.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12026 >
2021-07-28 00:26:06 +00:00
Alyssa Rosenzweig
8cbabbd532
pan/bi: Only call clause code on Bifrost
...
Valhall will have its own simpler code path.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12026 >
2021-07-28 00:26:06 +00:00
Jason Ekstrand
74ec2b12be
nir/lower_tex: Rework invalid implicit LOD lowering
...
Only fragment and some compute shaders support implicit derivatives.
They're totally meaningless without helper invocations and some
understanding of the dispatch pattern. We've got code to lower
nir_texop_tex in these shader stages to use an explicit derivative of 0
but it was pretty badly broken:
1. It only handled nir_texop_tex, not nir_texop_txb or nir_texop_lod.
2. It didn't take min_lod into account
3. It was conflated with adding a missing LOD parameter to opcodes
which expect one such as nir_texop_txf. While not really a bug,
this does make it way harder to reason about the code.
4. Unless you set a flag (which most drivers don't), it left the
opcode nir_texop_tex instead of nir_texop_txl which it should have
been.
This reworks it to go through roughly the same path as other LOD
lowering only with a constant lod of 0 instead of calling out to
nir_texop_lod. We also get rid of the lower_tex_without_implicit_lod
flag because most drivers set it and those that don't are probably
subtly broken. If someone really wants to get nir_texop_tex in their
vertex shaders, they can write a new patch to add the flag back in.
Fixes: e382890e25
"nir: set default lod to texture opcodes that..."
Fixes: d5ac5d6e83
"nir: Add option to lower tex to txl when..."
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11775 >
2021-07-23 15:53:57 +00:00
Jason Ekstrand
cdde108af5
panfrost: Don't handle nir_texop_txf_ms_mcs
...
It's an intel-specific opcode and will never be seen on panfrost
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Reviewed-by: Connor Abbott <cwabbott0@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11775 >
2021-07-23 15:53:57 +00:00
Alyssa Rosenzweig
b56f5c27de
pan/bi: Add explicit cast for lod_or_mode
...
The enums alias. Fixes the following warning under clang:
../src/panfrost/bifrost/bifrost_compile.c:2515:25: warning: implicit conversion from enumeration type 'enum bifrost_texture_fetch' to different enumeration type 'enum bifrost_lod_mode' [-Wenum-conversion]
BIFROST_TEXTURE_FETCH_TEXEL;
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12003 >
2021-07-22 17:55:49 +00:00
Alyssa Rosenzweig
84e2e5f23c
pan/bi: Clean up useless casts
...
Left over from removing inheritance with sed instead of coccinelle.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11936 >
2021-07-18 01:49:26 +00:00
Alyssa Rosenzweig
1fcea30295
pan/bi: Copy block bi_block
...
Gets rid of the silly inheritance everywhere, which has caused _far_
more problems in practice than it has fixed. It was an idea I tried
before the pandemic. It didn't work. I'm finally cleaning it up.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11936 >
2021-07-18 01:49:26 +00:00
Alyssa Rosenzweig
8bee48a635
pan/bi: Copy back add_successor
...
Trying to get back independent block types.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11936 >
2021-07-18 01:49:26 +00:00
Alyssa Rosenzweig
5996622050
pan/bi: Calculate dependency graph when bundling
...
Code is ported from Midgard, modified to be scalar, post-RA, and to put
the arrays on the worklist instead of the instruction to save memory.
This enables out-of-order scheduling.
total tuples in shared programs: 128691 -> 125304 (-2.63%)
tuples in affected programs: 114091 -> 110704 (-2.97%)
helped: 844
HURT: 377
helped stats (abs) min: 1.0 max: 150.0 x̄: 4.88 x̃: 3
helped stats (rel) min: 0.30% max: 26.42% x̄: 5.56% x̃: 4.35%
HURT stats (abs) min: 1.0 max: 8.0 x̄: 1.94 x̃: 1
HURT stats (rel) min: 0.20% max: 33.33% x̄: 6.84% x̃: 3.23%
95% mean confidence interval for tuples value: -3.16 -2.38
95% mean confidence interval for tuples %-change: -2.19% -1.27%
Tuples are helped.
total clauses in shared programs: 27579 -> 26059 (-5.51%)
clauses in affected programs: 20606 -> 19086 (-7.38%)
helped: 941
HURT: 39
helped stats (abs) min: 1.0 max: 21.0 x̄: 1.66 x̃: 1
helped stats (rel) min: 0.69% max: 44.44% x̄: 10.48% x̃: 9.09%
HURT stats (abs) min: 1.0 max: 2.0 x̄: 1.15 x̃: 1
HURT stats (rel) min: 1.89% max: 10.00% x̄: 4.73% x̃: 4.55%
95% mean confidence interval for clauses value: -1.63 -1.47
95% mean confidence interval for clauses %-change: -10.27% -9.48%
Clauses are helped.
total cycles in shared programs: 12262.54 -> 12154.79 (-0.88%)
cycles in affected programs: 2210.54 -> 2102.79 (-4.87%)
helped: 374
HURT: 56
helped stats (abs) min: 0.041665999999999315 max: 6.25 x̄: 0.30 x̃: 0
helped stats (rel) min: 0.42% max: 26.00% x̄: 6.90% x̃: 7.14%
HURT stats (abs) min: 0.041665999999999315 max: 0.5833319999999986 x̄: 0.11 x̃: 0
HURT stats (rel) min: 0.16% max: 100.00% x̄: 55.17% x̃: 50.00%
95% mean confidence interval for cycles value: -0.29 -0.21
95% mean confidence interval for cycles %-change: -1.37% 3.73%
Inconclusive result (%-change mean confidence interval includes 0).
total arith in shared programs: 4852.29 -> 4658.13 (-4.00%)
arith in affected programs: 4525.17 -> 4331 (-4.29%)
helped: 1112
HURT: 166
helped stats (abs) min: 0.041665999999999315 max: 6.25 x̄: 0.19 x̃: 0
helped stats (rel) min: 0.42% max: 33.33% x̄: 6.59% x̃: 5.36%
HURT stats (abs) min: 0.041665999999999315 max: 0.5833319999999986 x̄: 0.07 x̃: 0
HURT stats (rel) min: 0.16% max: 100.00% x̄: 25.05% x̃: 2.40%
95% mean confidence interval for arith value: -0.17 -0.14
95% mean confidence interval for arith %-change: -3.44% -1.51%
Arith are helped.
total quadwords in shared programs: 117141 -> 111394 (-4.91%)
quadwords in affected programs: 104390 -> 98643 (-5.51%)
helped: 1245
HURT: 76
helped stats (abs) min: 1.0 max: 69.0 x̄: 4.74 x̃: 4
helped stats (rel) min: 0.28% max: 35.00% x̄: 7.88% x̃: 6.45%
HURT stats (abs) min: 1.0 max: 8.0 x̄: 2.01 x̃: 1
HURT stats (rel) min: 0.20% max: 10.00% x̄: 3.52% x̃: 4.25%
95% mean confidence interval for quadwords value: -4.61 -4.09
95% mean confidence interval for quadwords %-change: -7.56% -6.88%
Quadwords are helped.
Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10961 >
2021-07-12 23:29:12 +00:00
Alyssa Rosenzweig
6bf8e960fa
pan/bi: Do helper termination analysis on clauses
...
Unlike the dependency analysis for the skip bits which is a function of
the data flow graph, the thread termination analysis is dependent on the
actual sequence of instructions. As such, it must be done after
scheduling to be correct in the presence of out-of-order scheduling.
Furthermore it's specified in terms of clauses, not instructions.
Reflecting this in our code gets a nice simplification. As a side effect
this puts extra td flags on subsequent clauses, which matches the DDK's
behaviour. (Maybe td is just a hint?)
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10961 >
2021-07-12 23:29:12 +00:00
Alyssa Rosenzweig
499397700c
pan/bi: Don't lower fpow
...
We can fuse the intermediate multiply with the FMA_RSCALE in the
exponent code and save an instruction. Whether this is better than
adding a NIR op remains to be seen.
total instructions in shared programs: 146614 -> 146190 (-0.29%)
instructions in affected programs: 40724 -> 40300 (-1.04%)
helped: 157
HURT: 0
helped stats (abs) min: 1.0 max: 9.0 x̄: 2.70 x̃: 2
helped stats (rel) min: 0.22% max: 10.34% x̄: 1.37% x̃: 1.20%
95% mean confidence interval for instructions value: -3.00 -2.40
95% mean confidence interval for instructions %-change: -1.58% -1.15%
Instructions are helped.
total tuples in shared programs: 128116 -> 127696 (-0.33%)
tuples in affected programs: 33421 -> 33001 (-1.26%)
helped: 150
HURT: 0
helped stats (abs) min: 1.0 max: 16.0 x̄: 2.80 x̃: 2
helped stats (rel) min: 0.28% max: 4.37% x̄: 1.36% x̃: 1.07%
95% mean confidence interval for tuples value: -3.24 -2.36
95% mean confidence interval for tuples %-change: -1.50% -1.21%
Tuples are helped.
total clauses in shared programs: 27531 -> 27483 (-0.17%)
clauses in affected programs: 719 -> 671 (-6.68%)
helped: 20
HURT: 0
helped stats (abs) min: 1.0 max: 8.0 x̄: 2.40 x̃: 1
helped stats (rel) min: 1.61% max: 12.90% x̄: 6.96% x̃: 5.33%
95% mean confidence interval for clauses value: -3.48 -1.32
95% mean confidence interval for clauses %-change: -9.10% -4.82%
Clauses are helped.
total cycles in shared programs: 12250.81 -> 12233.69 (-0.14%)
cycles in affected programs: 1251.50 -> 1234.38 (-1.37%)
helped: 141
HURT: 0
helped stats (abs) min: 0.041665999999999315 max: 0.6666670000000003 x̄: 0.12 x̃: 0
helped stats (rel) min: 0.29% max: 5.00% x̄: 1.48% x̃: 1.20%
95% mean confidence interval for cycles value: -0.14 -0.10
95% mean confidence interval for cycles %-change: -1.63% -1.32%
Cycles are helped.
total arith in shared programs: 4840.25 -> 4822.71 (-0.36%)
arith in affected programs: 1324.08 -> 1306.54 (-1.32%)
helped: 151
HURT: 0
helped stats (abs) min: 0.041665999999999315 max: 0.6666670000000003 x̄: 0.12 x̃: 0
helped stats (rel) min: 0.29% max: 5.00% x̄: 1.43% x̃: 1.13%
95% mean confidence interval for arith value: -0.13 -0.10
95% mean confidence interval for arith %-change: -1.59% -1.28%
Arith are helped.
total texture in shared programs: 1666.50 -> 1666.50 (0.00%)
texture in affected programs: 0 -> 0
helped: 0
HURT: 0
total vary in shared programs: 639.06 -> 639.06 (0.00%)
vary in affected programs: 0 -> 0
helped: 0
HURT: 0
total ldst in shared programs: 9682 -> 9682 (0.00%)
ldst in affected programs: 0 -> 0
helped: 0
HURT: 0
total quadwords in shared programs: 116758 -> 116378 (-0.33%)
quadwords in affected programs: 28054 -> 27674 (-1.35%)
helped: 148
HURT: 2
helped stats (abs) min: 1.0 max: 16.0 x̄: 2.58 x̃: 2
helped stats (rel) min: 0.29% max: 5.13% x̄: 1.54% x̃: 1.23%
HURT stats (abs) min: 1.0 max: 1.0 x̄: 1.00 x̃: 1
HURT stats (rel) min: 0.67% max: 0.85% x̄: 0.76% x̃: 0.76%
95% mean confidence interval for quadwords value: -2.94 -2.12
95% mean confidence interval for quadwords %-change: -1.69% -1.33%
Quadwords are helped.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733 >
2021-07-09 23:15:28 +00:00
Alyssa Rosenzweig
91f130fa1e
pan/bi: Factor out exp2/log2 code
...
Will be reused for fpow.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733 >
2021-07-09 23:15:28 +00:00
Alyssa Rosenzweig
0991e9592d
pan/bi: Comment the fexp2 implementation
...
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733 >
2021-07-09 23:15:28 +00:00
Alyssa Rosenzweig
297ac6e453
pan/bi: Simplify cube map descriptor generation
...
We don't need to do the bitwise manipulation ourselves, we can just use
a bitwise MUX instead.
total instructions in shared programs: 146840 -> 146614 (-0.15%)
instructions in affected programs: 15037 -> 14811 (-1.50%)
helped: 109
HURT: 0
helped stats (abs) min: 2.0 max: 4.0 x̄: 2.07 x̃: 2
helped stats (rel) min: 0.86% max: 4.00% x̄: 1.70% x̃: 1.77%
95% mean confidence interval for instructions value: -2.15 -2.00
95% mean confidence interval for instructions %-change: -1.81% -1.59%
Instructions are helped.
total tuples in shared programs: 128149 -> 128116 (-0.03%)
tuples in affected programs: 2896 -> 2863 (-1.14%)
helped: 16
HURT: 0
helped stats (abs) min: 1.0 max: 5.0 x̄: 2.06 x̃: 1
helped stats (rel) min: 0.65% max: 2.33% x̄: 1.16% x̃: 0.70%
95% mean confidence interval for tuples value: -3.01 -1.12
95% mean confidence interval for tuples %-change: -1.50% -0.83%
Tuples are helped.
total cycles in shared programs: 12257.10 -> 12250.81 (-0.05%)
cycles in affected programs: 449.87 -> 443.58 (-1.40%)
helped: 92
HURT: 0
helped stats (abs) min: 0.0416660000000002 max: 0.20833400000000069 x̄: 0.07 x̃: 0
helped stats (rel) min: 0.93% max: 2.53% x̄: 1.40% x̃: 1.26%
95% mean confidence interval for cycles value: -0.08 -0.06
95% mean confidence interval for cycles %-change: -1.48% -1.32%
Cycles are helped.
total arith in shared programs: 4847.33 -> 4840.25 (-0.15%)
arith in affected programs: 490.37 -> 483.29 (-1.44%)
helped: 109
HURT: 0
helped stats (abs) min: 0.0416660000000002 max: 0.20833400000000069 x̄: 0.06 x̃: 0
helped stats (rel) min: 0.93% max: 5.56% x̄: 1.51% x̃: 1.26%
95% mean confidence interval for arith value: -0.07 -0.06
95% mean confidence interval for arith %-change: -1.64% -1.39%
Arith are helped.
total quadwords in shared programs: 116775 -> 116758 (-0.01%)
quadwords in affected programs: 1331 -> 1314 (-1.28%)
helped: 7
HURT: 0
helped stats (abs) min: 1.0 max: 4.0 x̄: 2.43 x̃: 3
helped stats (rel) min: 0.91% max: 2.38% x̄: 1.65% x̃: 1.39%
95% mean confidence interval for quadwords value: -3.48 -1.38
95% mean confidence interval for quadwords %-change: -2.27% -1.04%
Quadwords are helped.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733 >
2021-07-09 23:15:28 +00:00
Alyssa Rosenzweig
99b2dddebf
pan/bi: Add a constant subexpression elimination pass
...
ALU only. Intended to clean up the lowerings used with complex
texturings. Ex: if a shader reads two cube maps at the same coordinates,
this deduplicates the cube map transformation.
This needs to happen in the backend since we do the cube map
transformation with the backend builder, rather than special NIR ops.
This is a tradeoff.
Pass based on ir3's, which in turn is inspired by NIR's.
total instructions in shared programs: 148799 -> 147348 (-0.98%)
instructions in affected programs: 20509 -> 19058 (-7.07%)
helped: 145
HURT: 0
helped stats (abs) min: 4.0 max: 30.0 x̄: 10.01 x̃: 8
helped stats (rel) min: 1.92% max: 54.55% x̄: 10.87% x̃: 7.41%
95% mean confidence interval for instructions value: -10.73 -9.28
95% mean confidence interval for instructions %-change: -12.81% -8.94%
Instructions are helped.
total tuples in shared programs: 129992 -> 128908 (-0.83%)
tuples in affected programs: 17624 -> 16540 (-6.15%)
helped: 145
HURT: 0
helped stats (abs) min: 2.0 max: 25.0 x̄: 7.48 x̃: 7
helped stats (rel) min: 0.74% max: 42.86% x̄: 9.16% x̃: 7.22%
95% mean confidence interval for tuples value: -7.96 -6.99
95% mean confidence interval for tuples %-change: -10.52% -7.79%
Tuples are helped.
total clauses in shared programs: 27632 -> 27582 (-0.18%)
clauses in affected programs: 1077 -> 1027 (-4.64%)
helped: 44
HURT: 0
helped stats (abs) min: 1.0 max: 3.0 x̄: 1.14 x̃: 1
helped stats (rel) min: 2.50% max: 16.67% x̄: 4.99% x̃: 4.45%
95% mean confidence interval for clauses value: -1.26 -1.01
95% mean confidence interval for clauses %-change: -5.70% -4.27%
Clauses are helped.
total cycles in shared programs: 12323 -> 12285.63 (-0.30%)
cycles in affected programs: 618.25 -> 580.88 (-6.05%)
helped: 120
HURT: 0
helped stats (abs) min: 0.08333299999999966 max: 0.5416680000000014 x̄: 0.31 x̃: 0
helped stats (rel) min: 0.77% max: 66.67% x̄: 7.60% x̃: 7.37%
95% mean confidence interval for cycles value: -0.33 -0.29
95% mean confidence interval for cycles %-change: -8.73% -6.47%
Cycles are helped.
total arith in shared programs: 4916.75 -> 4866.88 (-1.01%)
arith in affected programs: 677.79 -> 627.92 (-7.36%)
helped: 145
HURT: 0
helped stats (abs) min: 0.08333299999999966 max: 1.0833329999999997 x̄: 0.34 x̃: 0
helped stats (rel) min: 0.77% max: 66.67% x̄: 12.81% x̃: 7.87%
95% mean confidence interval for arith value: -0.37 -0.32
95% mean confidence interval for arith %-change: -15.33% -10.29%
Arith are helped.
total quadwords in shared programs: 118117 -> 117262 (-0.72%)
quadwords in affected programs: 15283 -> 14428 (-5.59%)
helped: 143
HURT: 0
helped stats (abs) min: 1.0 max: 23.0 x̄: 5.98 x̃: 5
helped stats (rel) min: 0.44% max: 25.71% x̄: 7.56% x̃: 5.56%
95% mean confidence interval for quadwords value: -6.46 -5.50
95% mean confidence interval for quadwords %-change: -8.59% -6.53%
Quadwords are helped.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733 >
2021-07-09 23:15:28 +00:00
Alyssa Rosenzweig
3fef3b6afc
pan/bi: Analyze helper invocations
...
Set the .skip bit on texture instructions and the terminate discarded
threads bit on the clause header based on data flow analysis of helper
invocations. This code is adapted from Midgard, which requires the same
analysis with a few details changed.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733 >
2021-07-09 23:15:28 +00:00
Alyssa Rosenzweig
134c74e301
pan/bi: Track LOD mode even for TEXC
...
Redundant with the texture operation descriptor, but we don't want to
parse that in the rest of the compiler. Handling it as a pseudo-modifier
lets us share a code path with TEXS.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733 >
2021-07-09 23:15:28 +00:00
Alyssa Rosenzweig
b7ca125278
pan/bi: Report cycle counts
...
Based on analysis of results from the Mali Offline Compiler. I am
uncertain how well these translate to real life, and they are
normalized counts only...
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733 >
2021-07-09 23:15:28 +00:00
Alyssa Rosenzweig
113157dafe
pan/bi: Try to hit full occupancy on v7
...
Bifrost v7 trades off register pressure and occupancy. If we restrict to
[R0, R15] U [R48, R63], we get full occupancy, but if we use the full
register file, we only get half occupancy. Try to allocate just 32
registers, and only use the full 64 registers if that would spill.
Clever heuristics could make this both more effective (live range
splitting, shuffling, spilling if deemed acceptable) and cheaper at
compile-time (tracking maximum liveness to determine if it's possible to
hit at all). For now, this should suffice.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11733 >
2021-07-09 23:15:28 +00:00
Icecream95
c246af0dd8
panfrost: Only upload UBOs when needed
...
If all of the used values from a UBO are pushed, it doesn't need to be
uploaded.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11700 >
2021-07-03 13:23:29 +00:00
Jason Ekstrand
0afbfee8da
nir,panfrost: Suffix fsat_signed and fclamp_pos with _mali
...
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11463 >
2021-06-21 09:03:34 -05:00
Alyssa Rosenzweig
295e65bb50
pan/bi: Add back custom algebraic opts
...
Right now just do a trivial one to test the infrastructure. In the next
commit we'll use this for a more interesting optimization that's a bit
painful in BIR but trivial with nir_search.
total instructions in shared programs: 149566 -> 149562 (<.01%)
instructions in affected programs: 502 -> 498 (-0.80%)
helped: 3
HURT: 0
helped stats (abs) min: 1 max: 2 x̄: 1.33 x̃: 1
helped stats (rel) min: 0.38% max: 1.30% x̄: 0.97% x̃: 1.21%
total tuples in shared programs: 130957 -> 130487 (-0.36%)
tuples in affected programs: 54752 -> 54282 (-0.86%)
helped: 303
HURT: 2
helped stats (abs) min: 1 max: 29 x̄: 1.56 x̃: 1
helped stats (rel) min: 0.13% max: 7.14% x̄: 1.08% x̃: 0.92%
HURT stats (abs) min: 1 max: 2 x̄: 1.50 x̃: 1
HURT stats (rel) min: 1.89% max: 2.99% x̄: 2.44% x̃: 2.44%
95% mean confidence interval for tuples value: -1.79 -1.30
95% mean confidence interval for tuples %-change: -1.17% -0.95%
Tuples are helped.
total clauses in shared programs: 27877 -> 27827 (-0.18%)
clauses in affected programs: 1556 -> 1506 (-3.21%)
helped: 45
HURT: 0
helped stats (abs) min: 1 max: 2 x̄: 1.11 x̃: 1
helped stats (rel) min: 1.43% max: 9.52% x̄: 3.88% x̃: 3.57%
95% mean confidence interval for clauses value: -1.21 -1.02
95% mean confidence interval for clauses %-change: -4.38% -3.39%
Clauses are helped.
total quadwords in shared programs: 119058 -> 118563 (-0.42%)
quadwords in affected programs: 33777 -> 33282 (-1.47%)
helped: 250
HURT: 2
helped stats (abs) min: 1 max: 29 x̄: 1.99 x̃: 1
helped stats (rel) min: 0.23% max: 11.11% x̄: 1.67% x̃: 1.40%
HURT stats (abs) min: 1 max: 1 x̄: 1.00 x̃: 1
HURT stats (rel) min: 1.64% max: 2.00% x̄: 1.82% x̃: 1.82%
95% mean confidence interval for quadwords value: -2.27 -1.66
95% mean confidence interval for quadwords %-change: -1.80% -1.49%
Quadwords are helped.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11327 >
2021-06-15 20:27:22 +00:00
Alyssa Rosenzweig
41070fedca
pan/bi: Propagate fabs/neg/sat
...
Initial support for modifier propagation. Bifrost makes this
unreasonably hard.
total instructions in shared programs: 151604 -> 150761 (-0.56%)
instructions in affected programs: 48773 -> 47930 (-1.73%)
helped: 212
HURT: 0
helped stats (abs) min: 1 max: 28 x̄: 3.98 x̃: 1
helped stats (rel) min: 0.29% max: 12.70% x̄: 1.75% x̃: 1.26%
95% mean confidence interval for instructions value: -4.71 -3.25
95% mean confidence interval for instructions %-change: -1.97% -1.53%
Instructions are helped.
total tuples in shared programs: 131876 -> 131560 (-0.24%)
tuples in affected programs: 25393 -> 25077 (-1.24%)
helped: 104
HURT: 3
helped stats (abs) min: 1 max: 28 x̄: 3.08 x̃: 2
helped stats (rel) min: 0.34% max: 8.57% x̄: 1.55% x̃: 1.04%
HURT stats (abs) min: 1 max: 2 x̄: 1.33 x̃: 1
HURT stats (rel) min: 0.51% max: 2.86% x̄: 1.30% x̃: 0.53%
95% mean confidence interval for tuples value: -3.63 -2.28
95% mean confidence interval for tuples %-change: -1.73% -1.21%
Tuples are helped.
total clauses in shared programs: 28122 -> 28032 (-0.32%)
clauses in affected programs: 2720 -> 2630 (-3.31%)
helped: 58
HURT: 1
helped stats (abs) min: 1 max: 6 x̄: 1.57 x̃: 1
helped stats (rel) min: 0.88% max: 14.29% x̄: 4.06% x̃: 3.67%
HURT stats (abs) min: 1 max: 1 x̄: 1.00 x̃: 1
HURT stats (rel) min: 7.69% max: 7.69% x̄: 7.69% x̃: 7.69%
95% mean confidence interval for clauses value: -1.85 -1.20
95% mean confidence interval for clauses %-change: -4.60% -3.13%
Clauses are helped.
total quadwords in shared programs: 119778 -> 119509 (-0.22%)
quadwords in affected programs: 20698 -> 20429 (-1.30%)
helped: 95
HURT: 1
helped stats (abs) min: 1 max: 28 x̄: 2.85 x̃: 2
helped stats (rel) min: 0.38% max: 7.14% x̄: 1.50% x̃: 1.13%
HURT stats (abs) min: 2 max: 2 x̄: 2.00 x̃: 2
HURT stats (rel) min: 3.23% max: 3.23% x̄: 3.23% x̃: 3.23%
95% mean confidence interval for quadwords value: -3.49 -2.11
95% mean confidence interval for quadwords %-change: -1.71% -1.20%
Quadwords are helped.
Signed-off-by: Alyssa Rosenzweig <alyssa@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11327 >
2021-06-15 20:27:22 +00:00