110bcb49193fdd4c2d86c0407bea761a4e2c0bb7
383 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Rhys Perry
|
110bcb4919 |
nir/algebraic: add various ffma optimizations
fossil-db (GFX10.3): Totals from 7532 (5.15% of 146267) affected shaders: VGPRs: 414696 -> 414304 (-0.09%); split: -0.18%, +0.08% CodeSize: 33393444 -> 33375908 (-0.05%); split: -0.13%, +0.08% MaxWaves: 149854 -> 150094 (+0.16%); split: +0.27%, -0.11% Instrs: 6279823 -> 6271364 (-0.13%); split: -0.18%, +0.05% Latency: 60308898 -> 60296025 (-0.02%); split: -0.13%, +0.11% InvThroughput: 13770542 -> 13745192 (-0.18%); split: -0.24%, +0.06% fossil-db (GFX10): Totals from 7532 (5.15% of 146267) affected shaders: VGPRs: 406664 -> 405564 (-0.27%); split: -0.39%, +0.12% CodeSize: 33544656 -> 33527568 (-0.05%); split: -0.13%, +0.08% MaxWaves: 158584 -> 158858 (+0.17%); split: +0.30%, -0.13% Instrs: 6316242 -> 6307913 (-0.13%); split: -0.18%, +0.05% Latency: 60243290 -> 60232844 (-0.02%); split: -0.13%, +0.11% InvThroughput: 13643345 -> 13620171 (-0.17%); split: -0.24%, +0.07% fossil-db (GFX9): Totals from 7543 (5.15% of 146401) affected shaders: SGPRs: 546384 -> 547472 (+0.20%); split: -0.08%, +0.28% VGPRs: 412636 -> 411896 (-0.18%); split: -0.27%, +0.09% CodeSize: 33216196 -> 33210564 (-0.02%); split: -0.12%, +0.11% MaxWaves: 38771 -> 38789 (+0.05%); split: +0.17%, -0.12% Instrs: 6419878 -> 6414891 (-0.08%); split: -0.18%, +0.11% Latency: 70972327 -> 70922754 (-0.07%); split: -0.15%, +0.08% InvThroughput: 33949039 -> 33909258 (-0.12%); split: -0.20%, +0.08% Signed-off-by: Rhys Perry <pendingchaos02@gmail.com> Reviewed-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8056> |
||
|
Eric Engestrom
|
f1eae2f8bb |
python: drop python2 support
Signed-off-by: Eric Engestrom <eric@engestrom.ch> Acked-by: Jose Fonseca <jfonseca@vmware.com> Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com> Reviewed-by: Dylan Baker <dylan@pnwbakers.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3674> |
||
|
Ian Romanick
|
84d2e53789 |
Revert "nir/algebraic: Convert some f2u to f2i"
Per https://gitlab.freedesktop.org/mesa/mesa/-/issues/5178#note_1019666,
the assumption fundamental to this optimization is false. Section
2.4.1 (Float to Integer) of Ivy Bridge PRMs describes the situation.
The wording of the section is somewhat confusing (because it doesn't
clearly delineate between signed and unsigned integers), but the last
two rows of the table make it clear that F->UD conversion clamps
negative float values to 0.
All other hardware mentioned in that thread seems to behave the same
way.
The real problem is that, with hardware that behaves in this ways,
converting f2u(2147483648.0) to f2i(2147483648.0) changes the bit pattern
that would be produced from 0x80000000 to 0x7fffffff.
This reverts commit
|
||
|
Rhys Perry
|
4e2b94331b |
nir/algebraic: improve irem by power-of-two optimization
Requires one less instruction. No fossil-db changes. Signed-off-by: Rhys Perry <pendingchaos02@gmail.com> Reviewed-by: Ian Romanick <ian.d.romanick@intel.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12039> |
||
|
Rhys Perry
|
b009467b81 |
nir/algebraic: add optimizations for imul(a, INT_MIN)
is_pos_power_of_two would catch this, but nir_op_imul has signed sources, so is_neg_power_of_two catches it instead, which creates a useless nir_op_ineg. fossil-db (Sienna Cichlid): Totals from 1014 (0.68% of 150170) affected shaders: CodeSize: 3592296 -> 3592288 (-0.00%); split: -0.00%, +0.00% Instrs: 671211 -> 670426 (-0.12%) Latency: 5268917 -> 5268479 (-0.01%); split: -0.01%, +0.00% InvThroughput: 2187349 -> 2187343 (-0.00%); split: -0.00%, +0.00% VClause: 8634 -> 8636 (+0.02%) Copies: 97585 -> 97604 (+0.02%) Signed-off-by: Rhys Perry <pendingchaos02@gmail.com> Reviewed-by: Ian Romanick <ian.d.romanick@intel.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12039> |
||
|
Rhys Perry
|
65cd5a0f22 |
nir/algebraic: don't optimize umod/imod/irem if lower_bitops=true
Match the udiv/idiv/imul by power-of-two optimizations. Signed-off-by: Rhys Perry <pendingchaos02@gmail.com> Reviewed-by: Ian Romanick <ian.d.romanick@intel.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12039> |
||
|
Rhys Perry
|
ec4b425f59 |
nir/algebraic: fix imod by negative power-of-two
If "a" is a multiple of "b", then the result would have been "b" instead
of 0.
No fossil-db changes.
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Fixes:
|
||
|
Dave Airlie
|
ad92c2b253 |
nir: add fisnormal lowering
just lower the 32-bit version for now. Thanks to alyssa for this suggested lowering. Reviewed-by: Jesse Natalie <jenatali@microsoft.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12207> |
||
|
Sagar Ghuge
|
06ab737686 |
nir: Add optimizations for iadd3
This patch also adds has_iadd3 bit to give more control if backend supports ternary add instruction or not. v2: - Add patterns in late optimization (Connor Abbott) Suggested-by: Alyssa/Jason Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com> Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com> Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11596> |
||
|
Jason Ekstrand
|
2e08bae9b3 |
nir,vc4: Suffix a bunch of unorm 4x8 opcodes _vc4
Reviewed-by: Alyssa Rosenzweig <alyssa@collabora.com> Reviewed-by: Iago Toral Quiroga <itoral@igalia.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11463> |
||
|
Rhys Perry
|
1cbcfb8b38 |
nir, nir/algebraic: add byte/word insertion instructions
Signed-off-by: Rhys Perry <pendingchaos02@gmail.com> Reviewed-by: Timur Kristóf <timur.kristof@gmail.com> Reviewed-by: Ian Romanick <ian.d.romanick@intel.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3151> |
||
|
Rhys Perry
|
edae3e5623 |
nir/algebraic: optimize extract of extract
Found in some sottr shaders (originally iand(ishr(a, 16), 0xffff)) Signed-off-by: Rhys Perry <pendingchaos02@gmail.com> Reviewed-by: Timur Kristóf <timur.kristof@gmail.com> Reviewed-by: Ian Romanick <ian.d.romanick@intel.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3151> |
||
|
Ian Romanick
|
d246c31ec1 |
nir/algebraic: Add algebraic opt for float comparisons with identical operands.
The flt version could have been added in
|
||
|
Ian Romanick
|
64bcfc3a17 |
nir/algebraic: Rearrange some logic-joined comparisons and reduce
On Skylake and Broadwell, a single big compute shader in Dirt Rally has spills and fills *REALLY* helped. That same shader is hurt very slightly for spills and fills on Ice Lake. v2: Move the patterns earlier to be nearer other patterns that are similar. Mark the replacement fmin and fmax exact. Both suggested by Rhys. Reviewed-by: Rhys Perry <pendingchaos02@gmail.com> Tiger Lake total instructions in shared programs: 21073812 -> 21073041 (<.01%) instructions in affected programs: 77608 -> 76837 (-0.99%) helped: 522 HURT: 33 helped stats (abs) min: 1 max: 26 x̄: 1.58 x̃: 1 helped stats (rel) min: 0.22% max: 14.29% x̄: 1.29% x̃: 1.02% HURT stats (abs) min: 1 max: 8 x̄: 1.67 x̃: 1 HURT stats (rel) min: 0.25% max: 3.42% x̄: 1.06% x̃: 0.86% 95% mean confidence interval for instructions value: -1.57 -1.20 95% mean confidence interval for instructions %-change: -1.25% -1.05% Instructions are helped. total cycles in shared programs: 856224346 -> 856211096 (<.01%) cycles in affected programs: 2394231 -> 2380981 (-0.55%) helped: 603 HURT: 25 helped stats (abs) min: 1 max: 5218 x̄: 59.37 x̃: 28 helped stats (rel) min: 0.06% max: 5.61% x̄: 1.52% x̃: 1.37% HURT stats (abs) min: 2 max: 21394 x̄: 901.92 x̃: 10 HURT stats (rel) min: 0.02% max: 5.90% x̄: 0.95% x̃: 0.59% 95% mean confidence interval for cycles value: -93.61 51.41 95% mean confidence interval for cycles %-change: -1.50% -1.34% Inconclusive result (value mean confidence interval includes 0). LOST: 1 GAINED: 1 Ice Lake total instructions in shared programs: 20025692 -> 20024554 (<.01%) instructions in affected programs: 104981 -> 103843 (-1.08%) helped: 738 HURT: 0 helped stats (abs) min: 1 max: 30 x̄: 1.54 x̃: 1 helped stats (rel) min: 0.31% max: 10.53% x̄: 1.20% x̃: 1.06% 95% mean confidence interval for instructions value: -1.66 -1.43 95% mean confidence interval for instructions %-change: -1.26% -1.14% Instructions are helped. total cycles in shared programs: 979474407 -> 979422333 (<.01%) cycles in affected programs: 4136364 -> 4084290 (-1.26%) helped: 759 HURT: 59 helped stats (abs) min: 2 max: 11010 x̄: 72.78 x̃: 28 helped stats (rel) min: 0.03% max: 6.43% x̄: 1.23% x̃: 1.02% HURT stats (abs) min: 1 max: 698 x̄: 53.66 x̃: 8 HURT stats (rel) min: 0.02% max: 24.05% x̄: 1.64% x̃: 0.33% 95% mean confidence interval for cycles value: -97.08 -30.24 95% mean confidence interval for cycles %-change: -1.14% -0.91% Cycles are helped. total spills in shared programs: 10568 -> 10569 (<.01%) spills in affected programs: 102 -> 103 (0.98%) helped: 0 HURT: 1 total fills in shared programs: 11347 -> 11349 (0.02%) fills in affected programs: 277 -> 279 (0.72%) helped: 0 HURT: 1 LOST: 2 GAINED: 2 Skylake total instructions in shared programs: 18190419 -> 18188523 (-0.01%) instructions in affected programs: 102502 -> 100606 (-1.85%) helped: 791 HURT: 0 helped stats (abs) min: 1 max: 676 x̄: 2.40 x̃: 1 helped stats (rel) min: 0.34% max: 20.23% x̄: 1.41% x̃: 1.23% 95% mean confidence interval for instructions value: -4.07 -0.72 95% mean confidence interval for instructions %-change: -1.47% -1.34% Instructions are helped. total cycles in shared programs: 960737969 -> 960498951 (-0.02%) cycles in affected programs: 4435351 -> 4196333 (-5.39%) helped: 804 HURT: 67 helped stats (abs) min: 1 max: 198540 x̄: 300.54 x̃: 24 helped stats (rel) min: 0.03% max: 25.41% x̄: 1.21% x̃: 0.92% HURT stats (abs) min: 2 max: 680 x̄: 39.06 x̃: 6 HURT stats (rel) min: 0.05% max: 23.98% x̄: 1.12% x̃: 0.19% 95% mean confidence interval for cycles value: -722.03 173.20 95% mean confidence interval for cycles %-change: -1.15% -0.91% Inconclusive result (value mean confidence interval includes 0). total spills in shared programs: 9757 -> 9722 (-0.36%) spills in affected programs: 138 -> 103 (-25.36%) helped: 1 HURT: 0 total fills in shared programs: 9861 -> 9576 (-2.89%) fills in affected programs: 564 -> 279 (-50.53%) helped: 1 HURT: 0 LOST: 5 GAINED: 2 Broadwell total instructions in shared programs: 17853870 -> 17852414 (<.01%) instructions in affected programs: 101276 -> 99820 (-1.44%) helped: 777 HURT: 0 helped stats (abs) min: 1 max: 264 x̄: 1.87 x̃: 1 helped stats (rel) min: 0.34% max: 8.44% x̄: 1.37% x̃: 1.23% 95% mean confidence interval for instructions value: -2.54 -1.21 95% mean confidence interval for instructions %-change: -1.42% -1.32% Instructions are helped. total cycles in shared programs: 1029846029 -> 1029725458 (-0.01%) cycles in affected programs: 4435791 -> 4315220 (-2.72%) helped: 813 HURT: 43 helped stats (abs) min: 2 max: 68560 x̄: 149.95 x̃: 24 helped stats (rel) min: 0.02% max: 73.73% x̄: 1.43% x̃: 0.92% HURT stats (abs) min: 2 max: 726 x̄: 31.12 x̃: 13 HURT stats (rel) min: 0.01% max: 8.43% x̄: 0.62% x̃: 0.31% 95% mean confidence interval for cycles value: -299.58 17.87 95% mean confidence interval for cycles %-change: -1.63% -1.02% Inconclusive result (value mean confidence interval includes 0). total spills in shared programs: 20333 -> 20307 (-0.13%) spills in affected programs: 151 -> 125 (-17.22%) helped: 1 HURT: 0 total fills in shared programs: 25899 -> 25775 (-0.48%) fills in affected programs: 573 -> 449 (-21.64%) helped: 1 HURT: 0 LOST: 5 GAINED: 0 Sandy Bridge, Ivy Bridge, and Haswell had similar results. (Haswell shown) total instructions in shared programs: 16417658 -> 16416320 (<.01%) instructions in affected programs: 96495 -> 95157 (-1.39%) helped: 774 HURT: 0 helped stats (abs) min: 1 max: 18 x̄: 1.73 x̃: 1 helped stats (rel) min: 0.33% max: 9.80% x̄: 1.52% x̃: 1.20% 95% mean confidence interval for instructions value: -1.83 -1.63 95% mean confidence interval for instructions %-change: -1.59% -1.46% Instructions are helped. total cycles in shared programs: 1037104346 -> 1037080579 (<.01%) cycles in affected programs: 3787747 -> 3763980 (-0.63%) helped: 791 HURT: 53 helped stats (abs) min: 1 max: 5411 x̄: 65.87 x̃: 32 helped stats (rel) min: 0.02% max: 21.17% x̄: 1.44% x̃: 1.18% HURT stats (abs) min: 2 max: 14160 x̄: 534.72 x̃: 18 HURT stats (rel) min: 0.02% max: 15.37% x̄: 5.70% x̃: 0.54% 95% mean confidence interval for cycles value: -69.39 13.07 95% mean confidence interval for cycles %-change: -1.19% -0.80% Inconclusive result (value mean confidence interval includes 0). LOST: 12 GAINED: 2 GM45 and Iron Lake had similar results. (Iron Lake shown) total instructions in shared programs: 8132855 -> 8132703 (<.01%) instructions in affected programs: 8782 -> 8630 (-1.73%) helped: 38 HURT: 0 helped stats (abs) min: 4 max: 4 x̄: 4.00 x̃: 4 helped stats (rel) min: 1.66% max: 3.23% x̄: 1.77% x̃: 1.72% 95% mean confidence interval for instructions value: -4.00 -4.00 95% mean confidence interval for instructions %-change: -1.88% -1.65% Instructions are helped. total cycles in shared programs: 238300850 -> 238298568 (<.01%) cycles in affected programs: 257202 -> 254920 (-0.89%) helped: 62 HURT: 2 helped stats (abs) min: 4 max: 58 x̄: 36.90 x̃: 50 helped stats (rel) min: 0.15% max: 1.55% x̄: 0.87% x̃: 1.12% HURT stats (abs) min: 2 max: 4 x̄: 3.00 x̃: 3 HURT stats (rel) min: 0.12% max: 0.22% x̄: 0.17% x̃: 0.17% 95% mean confidence interval for cycles value: -41.34 -29.98 95% mean confidence interval for cycles %-change: -0.95% -0.73% Cycles are helped. Fossil-db results: All Intel platforms had similar results. (Ice Lake shown) Instructions in all programs: 145296888 -> 145296346 (-0.0%) SENDs in all programs: 6863696 -> 6863696 (+0.0%) Loops in all programs: 38334 -> 38334 (+0.0%) Cycles in all programs: 8800262303 -> 8800258950 (-0.0%) Spills in all programs: 216880 -> 216880 (+0.0%) Fills in all programs: 334248 -> 334248 (+0.0%) Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10012> |
||
|
Ian Romanick
|
adc2835646 |
nir/algebraic: Mark some more logic-joined comparison reductions as exact
If the values are known to be numbers, the the replacements are exact. This is only applied to the patterns with constants. Constants should always be numbers, and shaders with NaN constants should be handled in a different way. No shader-db or fossil-db changes on any Intel platform. The intention is to make these patterns more future proof. Reviewed-by: Rhys Perry <pendingchaos02@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10012> |
||
|
Ian Romanick
|
23bbf3932b |
nir/algebraic: Mark some more comparison reductions exact
Reviewed-by: Rhys Perry <pendingchaos02@gmail.com> All Haswell and later Intel platforms had similar results. (Tiger Lake shown) total instructions in shared programs: 21049056 -> 21048939 (<.01%) instructions in affected programs: 4716 -> 4599 (-2.48%) helped: 39 HURT: 0 helped stats (abs) min: 1 max: 6 x̄: 3.00 x̃: 3 helped stats (rel) min: 0.99% max: 5.43% x̄: 2.80% x̃: 2.51% 95% mean confidence interval for instructions value: -3.46 -2.54 95% mean confidence interval for instructions %-change: -3.22% -2.38% Instructions are helped. total cycles in shared programs: 855141411 -> 855141159 (<.01%) cycles in affected programs: 54491 -> 54239 (-0.46%) helped: 28 HURT: 5 helped stats (abs) min: 2 max: 34 x̄: 12.82 x̃: 12 helped stats (rel) min: 0.06% max: 2.73% x̄: 0.94% x̃: 0.75% HURT stats (abs) min: 2 max: 52 x̄: 21.40 x̃: 6 HURT stats (rel) min: 0.11% max: 2.46% x̄: 0.90% x̃: 0.56% 95% mean confidence interval for cycles value: -13.72 -1.55 95% mean confidence interval for cycles %-change: -1.01% -0.31% Cycles are helped. Tiger Lake Instructions in all programs: 160902191 -> 160899554 (-0.0%) SENDs in all programs: 6812435 -> 6812435 (+0.0%) Loops in all programs: 38225 -> 38225 (+0.0%) Cycles in all programs: 7428581420 -> 7428555881 (-0.0%) Spills in all programs: 192582 -> 192582 (+0.0%) Fills in all programs: 304539 -> 304539 (+0.0%) A lot of fragment shaders in Shadow of the Tomb Raider were helped, and a bunch of vertex shaders in Octopath Traveler were hurt. Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10012> |
||
|
Ian Romanick
|
7d85dc4f35 |
nir/algebraic: Equality comparison inversions require sources be numbers
v2: Update A630 expected image checksum for minetest.trace. Reviewed-by: Rhys Perry <pendingchaos02@gmail.com> Tiger Lake total instructions in shared programs: 21036690 -> 21049485 (0.06%) instructions in affected programs: 852085 -> 864880 (1.50%) helped: 240 HURT: 2514 helped stats (abs) min: 1 max: 46 x̄: 2.45 x̃: 2 helped stats (rel) min: 0.15% max: 4.30% x̄: 0.79% x̃: 0.55% HURT stats (abs) min: 1 max: 198 x̄: 5.32 x̃: 2 HURT stats (rel) min: 0.06% max: 10.71% x̄: 1.48% x̃: 1.04% 95% mean confidence interval for instructions value: 4.14 5.15 95% mean confidence interval for instructions %-change: 1.23% 1.34% Instructions are HURT. total cycles in shared programs: 856045255 -> 855816220 (-0.03%) cycles in affected programs: 16743786 -> 16514751 (-1.37%) helped: 790 HURT: 1973 helped stats (abs) min: 1 max: 10766 x̄: 627.97 x̃: 18 helped stats (rel) min: <.01% max: 32.59% x̄: 3.01% x̃: 0.64% HURT stats (abs) min: 1 max: 4078 x̄: 135.36 x̃: 18 HURT stats (rel) min: <.01% max: 54.56% x̄: 2.80% x̃: 0.82% 95% mean confidence interval for cycles value: -131.36 -34.42 95% mean confidence interval for cycles %-change: 0.88% 1.40% Inconclusive result (value mean confidence interval and %-change mean confidence interval disagree). total spills in shared programs: 9771 -> 9766 (-0.05%) spills in affected programs: 47 -> 42 (-10.64%) helped: 1 HURT: 0 total fills in shared programs: 9451 -> 9430 (-0.22%) fills in affected programs: 91 -> 70 (-23.08%) helped: 1 HURT: 0 LOST: 16 GAINED: 51 All Intel GPUs from Sandybridge through Ice Lake had similar results. (Ice Lake shown) total instructions in shared programs: 20024781 -> 20025568 (<.01%) instructions in affected programs: 103309 -> 104096 (0.76%) helped: 12 HURT: 389 helped stats (abs) min: 1 max: 2 x̄: 1.17 x̃: 1 helped stats (rel) min: 0.20% max: 2.70% x̄: 1.36% x̃: 1.37% HURT stats (abs) min: 1 max: 8 x̄: 2.06 x̃: 1 HURT stats (rel) min: 0.05% max: 7.14% x̄: 1.25% x̃: 0.95% 95% mean confidence interval for instructions value: 1.78 2.15 95% mean confidence interval for instructions %-change: 1.06% 1.28% Instructions are HURT. total cycles in shared programs: 979419070 -> 979439180 (<.01%) cycles in affected programs: 4968711 -> 4988821 (0.40%) helped: 60 HURT: 381 helped stats (abs) min: 1 max: 1296 x̄: 96.92 x̃: 26 helped stats (rel) min: <.01% max: 27.10% x̄: 1.64% x̃: 0.65% HURT stats (abs) min: 1 max: 7320 x̄: 68.04 x̃: 30 HURT stats (rel) min: <.01% max: 19.77% x̄: 1.32% x̃: 0.87% 95% mean confidence interval for cycles value: 10.25 80.95 95% mean confidence interval for cycles %-change: 0.69% 1.15% Cycles are HURT. LOST: 1 GAINED: 2 GM45 and Iron Lake had similar results. (Iron Lake shown) total instructions in shared programs: 8128474 -> 8132527 (0.05%) instructions in affected programs: 642323 -> 646376 (0.63%) helped: 12 HURT: 1972 helped stats (abs) min: 1 max: 4 x̄: 3.00 x̃: 4 helped stats (rel) min: 0.72% max: 1.72% x̄: 1.09% x̃: 0.83% HURT stats (abs) min: 1 max: 16 x̄: 2.07 x̃: 3 HURT stats (rel) min: 0.12% max: 7.14% x̄: 0.77% x̃: 0.70% 95% mean confidence interval for instructions value: 1.99 2.10 95% mean confidence interval for instructions %-change: 0.74% 0.79% Instructions are HURT. total cycles in shared programs: 238280994 -> 238294376 (<.01%) cycles in affected programs: 8841250 -> 8854632 (0.15%) helped: 84 HURT: 1192 helped stats (abs) min: 4 max: 64 x̄: 12.50 x̃: 8 helped stats (rel) min: 0.02% max: 1.61% x̄: 0.28% x̃: 0.17% HURT stats (abs) min: 2 max: 198 x̄: 12.11 x̃: 12 HURT stats (rel) min: 0.02% max: 8.03% x̄: 0.28% x̃: 0.14% 95% mean confidence interval for cycles value: 9.65 11.32 95% mean confidence interval for cycles %-change: 0.22% 0.27% Cycles are HURT. No fossil-db changes on any Intel platform. Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10012> |
||
|
Ian Romanick
|
4246c2869c |
nir/algebraic: Invert comparisons less often
This fixes the piglit test range_analysis_fsat_of_nan.shader_test. That
test contains some code like
o = saturate(X) > 0 ? vec4(1.0, 0.0, 0.0, 1.0)
: vec4(0.0, 1.0, 0.0, 1.0);
A clever optimizer will convert this to
o = vec4(float(saturate(X) > 0),
float(!(saturate(X) > 0)),
0, 1);
Due to the ordering of optimizations in the compiler, the `saturate`
operations are removed. This is safe even in the presense of NaN.
o = vec4(float(X > 0), float(!(X > 0)), 0, 1);
Since the calculations are not marked precise, an overzealous
optimizer may reduce this to
o = vec4(float(X > 0), float(X <= 0), 0, 1);
This will result in black being output. The GLSL spec gives quite a bit
of leeway with respect to NaN, but that seems too far. The shader
author asked for a result of red or green. A result of black is still
"undefined behavior," but it's also a little mean.
This also enables CSE to do its job better.
v2: Update A530 expected image checksum for minetest.trace.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/4531
Fixes:
|
||
|
Ian Romanick
|
49177b9e2f |
nir/algebraic: Tautology replacements require sources be numbers
It seems worth the small amount of damage to give an extra cushion of not having to debug problems later. Reviewed-by: Rhys Perry <pendingchaos02@gmail.com> All Intel platforms had similar results. (Tiger Lake shown) total instructions in shared programs: 21043197 -> 21043359 (<.01%) instructions in affected programs: 4409 -> 4571 (3.67%) helped: 0 HURT: 25 HURT stats (abs) min: 1 max: 16 x̄: 6.48 x̃: 5 HURT stats (rel) min: 0.39% max: 15.38% x̄: 4.59% x̃: 4.40% 95% mean confidence interval for instructions value: 4.37 8.59 95% mean confidence interval for instructions %-change: 2.93% 6.26% Instructions are HURT. total cycles in shared programs: 856175986 -> 856176921 (<.01%) cycles in affected programs: 58908 -> 59843 (1.59%) helped: 0 HURT: 25 HURT stats (abs) min: 7 max: 70 x̄: 37.40 x̃: 38 HURT stats (rel) min: 0.27% max: 5.63% x̄: 1.87% x̃: 1.39% 95% mean confidence interval for cycles value: 31.11 43.69 95% mean confidence interval for cycles %-change: 1.35% 2.39% Cycles are HURT. No fossil-db changes on any Intel platform. Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10012> |
||
|
Ian Romanick
|
d69ba58644 |
nir/algebraic: Remove some optimizations of comparisons with fsat
When most of these patterns were created, we believed, incorrectly, that fsat(NaN) was NaN. We have since realized that fsat(NaN) is zero. Originally, this changed the patterns to use is_a_number. This didn't help any shaders, so it's easier to just drop the optimizations. This commit crossed paths with |
||
|
Jesse Natalie
|
d7ca0319d7 |
nir: Add relaxed 24bit opcodes
These are equivalent to the 32bit opcodes if there are no more efficient 24bit opcodes available, but inputs are guaranteed to already be 24bit, so the 24bit opcodes can be used instead if they exist and are efficient. Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> Reviewed-by: Karol Herbst <kherbst@redhat.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10549> |
||
|
Gert Wollny
|
a199697642 |
nir/opt_algebraic: optimizations for add umax/umin with zero
For unsigned comparisons with zero these ops can be eliminated. v2: Add comparison optimizations with -1 (Rhys Perry) Signed-off-by: Gert Wollny <gert.wollny@collabora.com> Reviewed-by: Eric Anholt <eric@anholt.net> (v1) Reviewed-by: Rhys Perry <pendingchaos02@gmail.com> Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10583> |
||
|
Jesse Natalie
|
3c8bcdc863 |
nir: Add a new opcode for [un]packing doubles
HLSL doesn't support bitcasting a 64bit integer to a double. DXIL doesn't have generic pack/unpack instructions, so we lower those to integer bitwise ops. As a result, NIR generic double pack/unpack would require our backend to emit a bitcast to get a double, but we want to match HLSL semantics and emit MakeDouble/SplitDouble. Adding a dedicated opcode for double pack/unpack allows us to add a pass to emit that instead, which lets our backend emit the right instruction to pack and unpack doubles. Acked-by: Jason Ekstrand <jason@jlekstrand.net> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/10063> |
||
|
Gert Wollny
|
0f5b3c37c5 |
nir: Add opcodes for fused comp + csel and optimizations
Some backends, like r600 support a fused version of int and float compare against zero and and csel. Adding these opcodes here makes it possible to optimize this in nir. v2: Add rules for float compare + csel Signed-off-by: Gert Wollny <gert.wollny@collabora.com> Reviewed-by: Kristian H. Kristensen <hoegsberg@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9452> |
||
|
Gert Wollny
|
a5747f8ab3 |
nir: add opcodes for *find_msb_rev and lowering
Some hardware supports a version of find_msb where the bits are counted starting at the high bit, and this needs some lowering to obtain the value that is expected by *find_msb Signed-off-by: Gert Wollny <gert.wollny@collabora.com> Reviewed-by: Kristian H. Kristensen <hoegsberg@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9452> |
||
|
Timur Kristóf
|
132171dc4e |
nir: Add a few more algebraic optimizations to help address calculation.
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com> Reviewed-by: Rhys Perry <pendingchaos02@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9201> |
||
|
Ian Romanick
|
2c4fd24c01 |
nir/algebraic: Apply addition property of equality to the other ordering too
Inequality comparison operations are not commutative, so `foo < bar` and `bar < foo` both have to be explicitly listed. Reviewed-by: Rhys Perry <pendingchaos02@gmail.com> All Intel GPUs had similar results. (Ice Lake shown) total instructions in shared programs: 20027051 -> 20026899 (<.01%) instructions in affected programs: 37181 -> 37029 (-0.41%) helped: 85 HURT: 0 helped stats (abs) min: 1 max: 20 x̄: 1.79 x̃: 1 helped stats (rel) min: 0.05% max: 6.78% x̄: 0.92% x̃: 0.68% 95% mean confidence interval for instructions value: -2.42 -1.15 95% mean confidence interval for instructions %-change: -1.23% -0.61% Instructions are helped. total cycles in shared programs: 979762793 -> 979753527 (<.01%) cycles in affected programs: 2653905 -> 2644639 (-0.35%) helped: 104 HURT: 50 helped stats (abs) min: 1 max: 1048 x̄: 119.99 x̃: 11 helped stats (rel) min: <.01% max: 9.88% x̄: 0.77% x̃: 0.20% HURT stats (abs) min: 1 max: 734 x̄: 64.26 x̃: 8 HURT stats (rel) min: <.01% max: 3.06% x̄: 0.36% x̃: 0.10% 95% mean confidence interval for cycles value: -98.65 -21.68 95% mean confidence interval for cycles %-change: -0.66% -0.15% Cycles are helped. Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9374> |
||
|
Ian Romanick
|
33031bdab6 |
nir/algebraic: Apply addition property of equality more conservatively
This allows a lot more CSE. Depending on where the addition and the comparison are scheduled, it may also reduce register pressure by reducing the live range of the addends. Across all the platforms, the shaders affected for spills or fills were all fragment shaders from Dirt Rally. Reviewed-by: Rhys Perry <pendingchaos02@gmail.com> Tiger Lake and Ice Lake had similar results. (Tiger Lake shown) total instructions in shared programs: 21043103 -> 21038804 (-0.02%) instructions in affected programs: 892878 -> 888579 (-0.48%) helped: 1549 HURT: 724 helped stats (abs) min: 1 max: 225 x̄: 4.14 x̃: 2 helped stats (rel) min: 0.05% max: 11.18% x̄: 1.04% x̃: 0.78% HURT stats (abs) min: 1 max: 71 x̄: 2.93 x̃: 1 HURT stats (rel) min: 0.07% max: 6.90% x̄: 0.80% x̃: 0.56% 95% mean confidence interval for instructions value: -2.33 -1.45 95% mean confidence interval for instructions %-change: -0.50% -0.40% Instructions are helped. total cycles in shared programs: 855054155 -> 855757566 (0.08%) cycles in affected programs: 58275918 -> 58979329 (1.21%) helped: 1213 HURT: 1680 helped stats (abs) min: 1 max: 107405 x̄: 1684.00 x̃: 10 helped stats (rel) min: <.01% max: 38.09% x̄: 1.51% x̃: 0.25% HURT stats (abs) min: 1 max: 126632 x̄: 1634.59 x̃: 12 HURT stats (rel) min: <.01% max: 85.91% x̄: 2.75% x̃: 0.49% 95% mean confidence interval for cycles value: -98.06 584.35 95% mean confidence interval for cycles %-change: 0.71% 1.22% Inconclusive result (value mean confidence interval includes 0). total spills in shared programs: 9843 -> 9771 (-0.73%) spills in affected programs: 72 -> 0 helped: 5 HURT: 0 total fills in shared programs: 9600 -> 9451 (-1.55%) fills in affected programs: 149 -> 0 helped: 5 HURT: 0 LOST: 14 GAINED: 9 Skylake total instructions in shared programs: 18185074 -> 18183866 (<.01%) instructions in affected programs: 575180 -> 573972 (-0.21%) helped: 1286 HURT: 468 helped stats (abs) min: 1 max: 15 x̄: 1.55 x̃: 1 helped stats (rel) min: 0.03% max: 4.08% x̄: 0.67% x̃: 0.65% HURT stats (abs) min: 1 max: 8 x̄: 1.69 x̃: 1 HURT stats (rel) min: 0.13% max: 7.69% x̄: 0.87% x̃: 0.45% 95% mean confidence interval for instructions value: -0.77 -0.60 95% mean confidence interval for instructions %-change: -0.30% -0.22% Instructions are helped. total cycles in shared programs: 960518105 -> 960608234 (<.01%) cycles in affected programs: 42536073 -> 42626202 (0.21%) helped: 1210 HURT: 1714 helped stats (abs) min: 1 max: 7015 x̄: 123.41 x̃: 10 helped stats (rel) min: <.01% max: 33.76% x̄: 1.32% x̃: 0.26% HURT stats (abs) min: 1 max: 14474 x̄: 139.71 x̃: 14 HURT stats (rel) min: <.01% max: 58.94% x̄: 2.00% x̃: 0.44% 95% mean confidence interval for cycles value: 4.02 57.63 95% mean confidence interval for cycles %-change: 0.43% 0.82% Cycles are HURT. LOST: 16 GAINED: 42 Broadwell total instructions in shared programs: 17856880 -> 17852158 (-0.03%) instructions in affected programs: 564836 -> 560114 (-0.84%) helped: 1243 HURT: 418 helped stats (abs) min: 1 max: 115 x̄: 4.36 x̃: 1 helped stats (rel) min: 0.03% max: 9.67% x̄: 0.90% x̃: 0.67% HURT stats (abs) min: 1 max: 8 x̄: 1.67 x̃: 1 HURT stats (rel) min: 0.14% max: 7.69% x̄: 0.89% x̃: 0.46% 95% mean confidence interval for instructions value: -3.45 -2.23 95% mean confidence interval for instructions %-change: -0.51% -0.38% Instructions are helped. total cycles in shared programs: 1031140321 -> 1029856892 (-0.12%) cycles in affected programs: 66986946 -> 65703517 (-1.92%) helped: 1084 HURT: 1653 helped stats (abs) min: 1 max: 415168 x̄: 1835.32 x̃: 10 helped stats (rel) min: <.01% max: 57.16% x̄: 1.19% x̃: 0.28% HURT stats (abs) min: 1 max: 43930 x̄: 427.14 x̃: 12 HURT stats (rel) min: <.01% max: 57.53% x̄: 1.32% x̃: 0.39% 95% mean confidence interval for cycles value: -915.76 -22.07 95% mean confidence interval for cycles %-change: 0.17% 0.47% Inconclusive result (value mean confidence interval and %-change mean confidence interval disagree). total spills in shared programs: 20891 -> 20335 (-2.66%) spills in affected programs: 1567 -> 1011 (-35.48%) helped: 70 HURT: 0 total fills in shared programs: 27307 -> 25905 (-5.13%) fills in affected programs: 5381 -> 3979 (-26.05%) helped: 71 HURT: 0 LOST: 17 GAINED: 20 Haswell total instructions in shared programs: 16411850 -> 16409414 (-0.01%) instructions in affected programs: 602666 -> 600230 (-0.40%) helped: 1152 HURT: 781 helped stats (abs) min: 1 max: 103 x̄: 3.59 x̃: 1 helped stats (rel) min: 0.03% max: 8.61% x̄: 0.85% x̃: 0.65% HURT stats (abs) min: 1 max: 41 x̄: 2.18 x̃: 1 HURT stats (rel) min: 0.12% max: 7.69% x̄: 0.88% x̃: 0.69% 95% mean confidence interval for instructions value: -1.74 -0.78 95% mean confidence interval for instructions %-change: -0.21% -0.10% Instructions are helped. total cycles in shared programs: 1035338781 -> 1036977801 (0.16%) cycles in affected programs: 68961096 -> 70600116 (2.38%) helped: 1246 HURT: 2206 helped stats (abs) min: 1 max: 392022 x̄: 1040.28 x̃: 14 helped stats (rel) min: <.01% max: 56.44% x̄: 2.32% x̃: 0.38% HURT stats (abs) min: 1 max: 68630 x̄: 1330.56 x̃: 18 HURT stats (rel) min: <.01% max: 69.97% x̄: 3.31% x̃: 0.61% 95% mean confidence interval for cycles value: 90.43 859.17 95% mean confidence interval for cycles %-change: 1.02% 1.54% Cycles are HURT. total spills in shared programs: 17805 -> 17457 (-1.95%) spills in affected programs: 1202 -> 854 (-28.95%) helped: 34 HURT: 31 total fills in shared programs: 20939 -> 20387 (-2.64%) fills in affected programs: 2702 -> 2150 (-20.43%) helped: 34 HURT: 31 LOST: 24 GAINED: 45 Ivy Bridge and earlier Intel GPUs had similar results. (Ivy Bridge shown) total instructions in shared programs: 15515912 -> 15516757 (<.01%) instructions in affected programs: 396569 -> 397414 (0.21%) helped: 578 HURT: 858 helped stats (abs) min: 1 max: 9 x̄: 1.32 x̃: 1 helped stats (rel) min: 0.04% max: 3.70% x̄: 0.65% x̃: 0.65% HURT stats (abs) min: 1 max: 11 x̄: 1.87 x̃: 1 HURT stats (rel) min: 0.08% max: 12.90% x̄: 0.95% x̃: 0.53% 95% mean confidence interval for instructions value: 0.47 0.70 95% mean confidence interval for instructions %-change: 0.24% 0.37% Instructions are HURT. total cycles in shared programs: 584395455 -> 584466352 (0.01%) cycles in affected programs: 20346570 -> 20417467 (0.35%) helped: 1192 HURT: 1896 helped stats (abs) min: 1 max: 4108 x̄: 123.27 x̃: 14 helped stats (rel) min: <.01% max: 37.20% x̄: 2.27% x̃: 0.46% HURT stats (abs) min: 1 max: 3698 x̄: 114.89 x̃: 19 HURT stats (rel) min: <.01% max: 70.28% x̄: 3.02% x̃: 0.71% 95% mean confidence interval for cycles value: 10.75 35.16 95% mean confidence interval for cycles %-change: 0.73% 1.23% Cycles are HURT. LOST: 20 GAINED: 12 Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9374> |
||
|
Timothy Arceri
|
9f474bd4b4 |
nir: handle negatives in ffma reassociation optimisation
shader-db results Iris (BDW): total instructions in shared programs: 16632076 -> 16631057 (<.01%) instructions in affected programs: 48010 -> 46991 (-2.12%) helped: 47 HURT: 6 total cycles in shared programs: 915266726 -> 915263622 (<.01%) cycles in affected programs: 1182283 -> 1179179 (-0.26%) helped: 18 HURT: 27 total loops in shared programs: 4929 -> 4929 (0.00%) loops in affected programs: 0 -> 0 helped: 0 HURT: 0 total spills in shared programs: 18834 -> 18801 (-0.18%) spills in affected programs: 525 -> 492 (-6.29%) helped: 3 HURT: 0 total fills in shared programs: 23008 -> 22981 (-0.12%) fills in affected programs: 435 -> 408 (-6.21%) helped: 3 HURT: 0 Reviewed-by: Ian Romanick <ian.d.romanick@intel.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8608> |
||
|
Ian Romanick
|
3250e04d25 |
nir/algebraic: Add some max/min optimizations with 3 variables
Specifically, ARB assembly shaders with code like
SLT r0, r0, c[0].xxxx;
...
KIL r0.xyzx;
can result in this pattern. The other cases (e.g., 'KIL r0.xxxx' and
'KIL r0.xyxx') are handled by existing patterns.
Reviewed-by: Matt Turner <mattst88@gmail.com>
All Intel platforms had similar results. (Tiger Lake shown)
total instructions in shared programs: 21050098 -> 21050065 (<.01%)
instructions in affected programs: 2062 -> 2029 (-1.60%)
helped: 31
HURT: 1
helped stats (abs) min: 1 max: 3 x̄: 1.10 x̃: 1
helped stats (rel) min: 1.14% max: 4.35% x̄: 1.89% x̃: 1.69%
HURT stats (abs) min: 1 max: 1 x̄: 1.00 x̃: 1
HURT stats (rel) min: 0.65% max: 0.65% x̄: 0.65% x̃: 0.65%
95% mean confidence interval for instructions value: -1.23 -0.84
95% mean confidence interval for instructions %-change: -2.12% -1.50%
Instructions are helped.
total cycles in shared programs: 855105466 -> 855105055 (<.01%)
cycles in affected programs: 50136 -> 49725 (-0.82%)
helped: 33
HURT: 0
helped stats (abs) min: 3 max: 22 x̄: 12.45 x̃: 12
helped stats (rel) min: 0.13% max: 1.57% x̄: 0.86% x̃: 0.92%
95% mean confidence interval for cycles value: -13.78 -11.13
95% mean confidence interval for cycles %-change: -0.97% -0.76%
Cycles are helped.
No fossil-db changes on any Intel platform.
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9122>
|
||
|
Ian Romanick
|
d9b5bce85a |
nir/algebraic: Remove some redundant b2f logic-op reduction patterns
There are patterns that will re-write the fmin or fmax part into a form
that other patterns will gradually convert to the same ior or iand. For
example,
fmax(b2f(a), b2f(b)) != 0
b2f(a || b) != 0
a || b
No shader-db or fossil-db changes on any Intel platform.
Reviewed-by: Matt Turner <mattst88@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9122>
|
||
|
Ian Romanick
|
7e127c1fca |
nir/algebraic: Fix some min/max of b2f replacements
fmin(-A, -B) is -fmax(A, B), and fmax(-A, -B) is -fmin(A, B). Therefore the logic joining A and B should toggle between ior and iand for the negated versions. At the very least, a shader from Euro Truck Simulator 2 in shader-db is affected by this. The KIL instruction in the (ARB assembly) shader ends up with the wrong logic. This is _probably_ the source of https://gitlab.freedesktop.org/mesa/mesa/-/issues/1346. That said, the issue mentions that Mesa 18.0.5 works, but commit |
||
|
Jason Ekstrand
|
2491d5a662 |
nir/algebraic: Covert up-cast of down-cast to extract on Intel
This starts generating extract for bit sizes other than 32 but our back-end handles that just fine. Reviewed-by: Ian Romanick <ian.d.romanick@.intel.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8872> |
||
|
Jason Ekstrand
|
f9b3be09e1 |
nir/algebraic: Clean up up-cast of down-cast when we can
There are a bunch of cases where we can pretty quickly determine that the high bits don't matter. In these cases, delete the casts. Reviewed-by: Ian Romanick <ian.d.romanick@intel.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8872> |
||
|
Ian Romanick
|
ed138f2861 |
nir/algebraic: Partially revert 3f782cdd25
I'm not sure what the logic was, but there is no opportunity for anything to flush to zero here. 'a' is a Boolean value, and b2f produces 1.0 or 0.0. This was originally part of https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/3765/. Reviewed-by: Matt Turner <mattst88@gmail.com> Cc: Andres Gomez <agomez@igalia.com> Cc: Samuel Iglesias Gonsálvez <siglesias@igalia.com> Cc: Connor Abbott <cwabbott0@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8910> |
||
|
Ian Romanick
|
5923742356 |
nir/algebraic: add patterns for a >> #b << #b and a << #b >> #b
Commit
|
||
|
Ian Romanick
|
6b0443a900 |
nir/algebraic: Fix a >> #b << #b for sizes other than 32-bit
The base mask previously used was 0xffffffff. This is not correct (but should still work) for 16-bit and 8-bit values, but it means the high 32-bits of 64-bit values will get chopped off. Instead of just restricting the pattern to 32-bits (as was done before |
||
|
Samuel Pitoiset
|
4c3ad4d065 |
nir/algebraic: mark more optimization with fsat(NaN) as inexact
These optimizations are duplicated from the main optimization table
to the late one... And I missed some in the original fix.
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3368
Fixes:
|
||
|
Rhys Perry
|
b729cd58d7 |
nir/algebraic: eliminate exact a*0.0 if float execution mode allow it
fossil-db (GFX10): Totals from 611 (0.44% of 139391) affected shaders: SGPRs: 40528 -> 40288 (-0.59%) VGPRs: 16136 -> 16152 (+0.10%); split: -0.15%, +0.25% CodeSize: 970192 -> 951036 (-1.97%) MaxWaves: 10561 -> 10557 (-0.04%); split: +0.08%, -0.11% Instrs: 174874 -> 172879 (-1.14%); split: -1.18%, +0.04% fossil-db (GFX10.3): Totals from 611 (0.44% of 139391) affected shaders: SGPRs: 40680 -> 40488 (-0.47%) VGPRs: 18368 -> 18276 (-0.50%); split: -0.57%, +0.07% CodeSize: 1050712 -> 1033624 (-1.63%); split: -1.64%, +0.02% MaxWaves: 8658 -> 8674 (+0.18%) Instrs: 205364 -> 201220 (-2.02%) Signed-off-by: Rhys Perry <pendingchaos02@gmail.com> Reviewed-by: Daniel Schürmann <daniel@schuermann.dev> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5523> |
||
|
Rhys Perry
|
614ab26afd |
nir/algebraic: optimize out exact a+0.0 if it's used only as a float
fossil-db (GFX10): Totals from 133 (0.10% of 139391) affected shaders: SGPRs: 7864 -> 7856 (-0.10%); split: -0.20%, +0.10% VGPRs: 4884 -> 4836 (-0.98%) CodeSize: 288932 -> 287084 (-0.64%) MaxWaves: 1973 -> 1979 (+0.30%) Instrs: 53899 -> 53550 (-0.65%) fossil-db (GFX10.3): Totals from 133 (0.10% of 139391) affected shaders: SGPRs: 7832 -> 7835 (+0.04%); split: -0.06%, +0.10% VGPRs: 5144 -> 5088 (-1.09%) CodeSize: 318912 -> 316696 (-0.69%) MaxWaves: 1735 -> 1746 (+0.63%) Instrs: 65367 -> 64853 (-0.79%) Signed-off-by: Rhys Perry <pendingchaos02@gmail.com> Reviewed-by: Daniel Schürmann <daniel@schuermann.dev> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5523> |
||
|
Rhys Perry
|
2849f0b5aa |
nir/algebraic: optimize out exact a*1.0 if it's used only as a float
fossil-db (GFX10): Totals from 10180 (7.30% of 139391) affected shaders: SGPRs: 549392 -> 549448 (+0.01%); split: -0.00%, +0.01% VGPRs: 243228 -> 243008 (-0.09%); split: -0.11%, +0.02% CodeSize: 12939080 -> 12603996 (-2.59%); split: -2.59%, +0.00% MaxWaves: 186948 -> 186976 (+0.01%) Instrs: 2497266 -> 2414648 (-3.31%) fossil-db (GFX10.3): Totals from 10180 (7.30% of 139391) affected shaders: SGPRs: 549672 -> 549280 (-0.07%); split: -0.23%, +0.16% VGPRs: 289296 -> 283672 (-1.94%); split: -2.83%, +0.88% CodeSize: 13920180 -> 13255560 (-4.77%); split: -4.77%, +0.00% MaxWaves: 151789 -> 153165 (+0.91%) Instrs: 2756978 -> 2671517 (-3.10%); split: -3.10%, +0.00% Signed-off-by: Rhys Perry <pendingchaos02@gmail.com> Reviewed-by: Daniel Schürmann <daniel@schuermann.dev> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5523> |
||
|
Daniel Schürmann
|
bd8e84eb8d |
nir: replace .lower_sub with .has_fsub and .has_isub
This allows a more fine-grained control about whether a backend supports one of these instructions. Reviewed-by: Eric Anholt <eric@anholt.net> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6597> |
||
|
Daniel Schürmann
|
b3ce55b445 |
nir,vc4: Lower fneg to fmul(x, -1.0)
This patch also replaces lower_negate with lower_ineg / lower_fneg. The fneg semantics have been clarified as of Version 1.5, Revision 1 of the SPIR-V specification, which means that the previous lowering to fsub is not a viable solution anymore, and is replaced with lowering to fmul(x, -1.0). Reviewed-by: Erik Faye-Lund <erik.faye-lund@collabora.com> Reviewed-by: Eric Anholt <eric@anholt.net> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6597> |
||
|
Ian Romanick
|
539c25c2da |
nir/algebraic: Move the flrp -> bcsel rule earlier
If multiple rules could match, the rule that appears first in the file
is used.
Only Tiger Lake and Ice Lake are affected. Other platforms either have
a LRP instruction or can't run any shaders from shader-db that would
benefit.
v2: Fix issues created when this commit was rebased on top of
|
||
|
Ian Romanick
|
ec16f935fe |
nir/algebraic: Mark comparisons generated from lowered fsign precise
This prevents other transformations from converting them to 'a != 0'.
For example, both of these transformations can do this:
(('~flt', 0.0, ('fabs', a)), ('fne', a, 0.0)),
(('~flt', ('fneg', ('fabs', a)), 0.0), ('fne', a, 0.0)),
Both fsign(fabs(NaN)) and fsign(fneg(fabs(NaN))) should produce zero,
but, since 'NaN != 0.0' is true, cascading these transformations could
cause them to generate 1.0 or -1.0 respecively.
No shader-db or fossil-db changes on any Intel platform.
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358>
|
||
|
Ian Romanick
|
9771af5dde |
nir/algebraic: Fix broken NaN and -0.0 behavior
No shader-db or fossil-db changes on any Intel platform.
v2: Add a coding line to fix SCons build problems caused by the ±
character.
Fixes:
|
||
|
Ian Romanick
|
55621c6d1c |
nir/algebraic: Add some compare-with-zero optimizations that are exact
This prevents some fossil-db regressions in "spir-v: Mark floating point comparisons exact". v2: Note that the patterns and replacements produce the same value when isnan(b). Suggested by Caio. v3: Use C99 isfinite() instead of (obsolete) BSD finite(). Fixes various Windows builds. No fossil-db changes on any Inetl platform, Vega, or Polaris10. All Intel platforms had similar results. (Tiger Lake shown) total instructions in shared programs: 20908670 -> 20908672 (<.01%) instructions in affected programs: 69 -> 71 (2.90%) helped: 0 HURT: 1 total cycles in shared programs: 473515288 -> 473513940 (<.01%) cycles in affected programs: 4942 -> 3594 (-27.28%) helped: 2 HURT: 0 Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358> |
||
|
Ian Romanick
|
9167324a86 |
nir/algebraic: Mark some logic-joined comparison reductions as exact
This also prevents some fossil-db regressions in "spir-v: Mark floating point comparisons exact". v2: Mark the fmin / fmax in the replacement exact to prevent other optimizations from ruining the NaN-clensing property of the fmin / fmax. Suggested by Rhys. Don't assume that constants are not NaN because some components of a vector might be NaN while others are numbers. Noticed by Rhys. This causes ~8 more shaders in Age of Wonders III (dxvk) to regress on cycles (not instructions) by less than 1% when "spir-v: Mark floating point comparisons exact" is applied. This difference is too small to care. All Intel platforms had similar results. (Tiger Lake shown) total instructions in shared programs: 20908668 -> 20908670 (<.01%) instructions in affected programs: 9196 -> 9198 (0.02%) helped: 10 HURT: 5 helped stats (abs) min: 1 max: 2 x̄: 1.40 x̃: 1 helped stats (rel) min: 0.02% max: 5.41% x̄: 2.20% x̃: 2.16% HURT stats (abs) min: 2 max: 6 x̄: 3.20 x̃: 3 HURT stats (rel) min: 2.44% max: 16.67% x̄: 9.39% x̃: 12.50% 95% mean confidence interval for instructions value: -1.22 1.49 95% mean confidence interval for instructions %-change: -2.08% 5.41% Inconclusive result (value mean confidence interval includes 0). total cycles in shared programs: 473515330 -> 473515288 (<.01%) cycles in affected programs: 67146 -> 67104 (-0.06%) helped: 10 HURT: 7 helped stats (abs) min: 1 max: 36 x̄: 15.90 x̃: 17 helped stats (rel) min: 0.01% max: 1.29% x̄: 0.66% x̃: 0.89% HURT stats (abs) min: 1 max: 48 x̄: 16.71 x̃: 4 HURT stats (rel) min: 0.08% max: 1.94% x̄: 0.87% x̃: 0.19% 95% mean confidence interval for cycles value: -13.88 8.94 95% mean confidence interval for cycles %-change: -0.56% 0.49% Inconclusive result (value mean confidence interval includes 0). Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358> |
||
|
Ian Romanick
|
fe3c518277 |
nir/algebraic: Don't add reordered version of patterns for commutative instructions
The reordered are automatically considered by nir_algebraic rules for commutative instructions. No shader-db or fossil-db changes on any Intel platform. Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/6358> |
||
|
Ian Romanick
|
314a40c902 |
Revert "nir: Replace an odd comparison involving fmin of -b2f"
I originally noticed that |