2fba61434b
support log1p and expm1 jet
...
Currently, it is not possible to accurately evaluate the derivative of
d/dx log(1 + x) under all circumstances and significant deviations from
the actual derivative d/dx log1p(x) can occur. This changeset introduces
the necessary Jet overload and its inverse, expm1.
Change-Id: Ifcf88f6d684f61ba86bbe49f0d551b703f34ad0d
2021-11-16 00:14:39 +01:00
a668cabbcf
support norm jet
...
norm allows to compute the squared magnitude both of complex and real
numbers. While Jet does not support decaying of a std::complex to a
scalar performed by norm, the function is still useful when applied to
scalars alone for computing the square.
Change-Id: I27a3513f53f37fb3960411362e80d170d1ae6f74
2021-11-15 22:29:44 +00:00
a3a4b6d77f
support copysign jet
...
libc++ (default on macOS) uses the copysign function for std::complex
multiplication which causes compilation errors such as these if
std::complex is combined with jets:
/Applications/Xcode_12.4.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/../include/c++/v1/complex:607:19: error: no matching function for call to 'copysign'
__a = copysign(__libcpp_isinf_or_builtin(__a) ? _Tp(1) : _Tp(0), __a);
^~~~~~~~
foo.hpp:218:39: note: in instantiation of function template specialization 'std::__1::operator*<ceres::Jet<double, 3> >' requested here
std::complex<Scalar> result = lhs * rhs;
copysign is also useful in other situations where the sign is required
allowing it to compute without branching.
Change-Id: I84b2d23374f1bf3bae32833016bb686d97d74054
2021-11-15 23:27:33 +01:00
b75dac1694
fix abs jet test comment
...
The comment suggests a non-existing (incorrect) test. Update the comment
and extend the test.
Change-Id: I362f0b5a2e81d286b672476bdb4394ed20263eb1
2021-11-15 18:34:34 +00:00
034bf566f0
use copysign for abs jet
...
abs() implementation does not correctly handle the sign of the
infinitesimal part when the value is a NaN. Define the function in terms
of copysign which is the only portable way of manipulating the sign of
a NaN value.
Change-Id: I2f13fb1db62d35ca8b09e6bff1c05a736cb91513
2021-11-14 17:07:14 +01:00
06e02a1735
Delete unused files split.h/cc
...
Change-Id: I67a4542389a0c1162875975efd23cdfbe797c6bc
2021-09-18 06:10:26 -07:00
ac7268da52
Fix an 80cols issue in covariance_impl.cc
...
Change-Id: I2422282919e7100ac852be2176cd754941a8b472
2021-09-15 06:22:29 -07:00
17dccef91b
Add NumericDiffFirstOrderFunction
...
This has been a long requested feature so that users can minimize
functions using numeric differentiation.
As part of this, I have also redone rosenbrock.cc, which now has three
variants.
rosenbrock.cc now uses automatic differentiation.
rosenbrock_numeric_diff.cc uses numeric differentiation.
rosenbrock_analytic_diff.cc uses analytic derivatives.
This is analogus to how the helloworld example code is structured.
The tutorial for GradientProblemSolver has also been updated to reflect
this.
https://github.com/ceres-solver/ceres-solver/issues/691
Change-Id: Ib0fb9e35127fe4c8299d4793bea3558722c70dd7
2021-09-15 06:21:02 -07:00
03d64141ae
Fix a incorrect check in reorder_program.cc
...
Thanks to William Gandler for pointing this out.
Change-Id: I7a4e0eebd2e557f7572df7455d7fa77527c087b6
2021-09-14 17:36:07 -07:00
2a2b9bd6fa
Fix a bug in covariance_impl.cc
...
A matrix reference was being incorrectly sized. It was using
the block_size number of rows, where as there are only block_local_size
number of rows that are present in the underlying matrix.
This did not affect the correctness of the resulting code, because
right after construction we then access a subblock with block_local_size
rows.
Thanks to William Gandler for reporting this.
Change-Id: I35e1f5a30b79e31b0b9fac79f7e361821d6073df
2021-08-15 16:52:08 -07:00
27fade7b8c
Fix a bug in system_test.cc
...
Fix an error in the cost functor for Powell's function.
Reported by William Gandler.
Change-Id: I4ffbbca3b3731aa13c3608067baaebd06c1a044a
2021-08-15 16:44:15 -07:00
d4eb83ee58
Fix the Jacobian in trust_region_minimizer_test.cc
...
The analytical Jacobians for Powell's singular function were
incorrect. Thanks to William Gandler for reporting this.
Change-Id: I780ffe449daa6935a79b904835e13edcf1cdffa2
2021-08-15 16:37:47 -07:00
5f6071a1cc
Fix a bug in local_parameterization_test.cc
...
Fix an incorrect comparison between in QuaternionParameterizationTestHelper.
Reported by William Gandler.
Change-Id: If4d63a51f6a55013b4e6f0c299462c5af6e80c70
2021-08-15 16:10:12 -07:00
b2e732b1e2
Fix errors in comments from William Gandler.
...
Change-Id: I1c1f1f7bdc4a1748994c3b43e72270b88f2a2b96
2021-08-15 16:01:40 -07:00
aefd37b182
Refactor small_blas_gemm_benchmark
...
This allows benchmarking both dynamic and static problem sizes.
On my X86 2Ghz machine:
CPU Caches:
L1 Data 32 KiB (x48)
L1 Instruction 32 KiB (x48)
L2 Unified 1024 KiB (x48)
L3 Unified 39424 KiB (x2)
----------------------------------------------------------------------------------------------
Benchmark Time CPU Iterations
----------------------------------------------------------------------------------------------
BM_MatrixMatrixMultiplyEigen_Static_2x3x4 10.9 ns 10.9 ns 63960496
BM_MatrixMatrixMultiplyEigen_Static_3x3x3 11.9 ns 11.9 ns 58093785
BM_MatrixMatrixMultiplyEigen_Static_4x4x4 24.9 ns 24.9 ns 28540133
BM_MatrixMatrixMultiplyEigen_Static_8x8x8 258 ns 258 ns 2755921
BM_MatrixMatrixMultiplyEigen_Static_9x9x3 255 ns 255 ns 2778628
BM_MatrixMatrixMultiplyEigen_Static_9x3x3 37.7 ns 37.7 ns 18556359
BM_MatrixMatrixMultiplyEigen_Static_3x9x9 220 ns 220 ns 3195163
BM_MatrixMatrixMultiplyEigen_Dynamic_2x3x4 37.8 ns 37.7 ns 18491162
BM_MatrixMatrixMultiplyEigen_Dynamic_3x3x3 48.8 ns 48.8 ns 13724884
BM_MatrixMatrixMultiplyEigen_Dynamic_4x4x4 74.5 ns 74.5 ns 9313146
BM_MatrixMatrixMultiplyEigen_Dynamic_8x8x8 271 ns 271 ns 2595807
BM_MatrixMatrixMultiplyEigen_Dynamic_9x9x3 259 ns 259 ns 2688515
BM_MatrixMatrixMultiplyEigen_Dynamic_9x3x3 123 ns 123 ns 5792115
BM_MatrixMatrixMultiplyEigen_Dynamic_3x9x9 236 ns 236 ns 2963896
BM_MatrixMatrixMultiplyNaive_Static_2x3x4 12.2 ns 12.2 ns 56472772
BM_MatrixMatrixMultiplyNaive_Static_3x3x3 15.5 ns 15.5 ns 44346456
BM_MatrixMatrixMultiplyNaive_Static_4x4x4 41.5 ns 41.5 ns 17196984
BM_MatrixMatrixMultiplyNaive_Static_8x8x8 199 ns 199 ns 3561730
BM_MatrixMatrixMultiplyNaive_Static_9x9x3 148 ns 148 ns 4764814
BM_MatrixMatrixMultiplyNaive_Static_9x3x3 38.4 ns 38.4 ns 17259019
BM_MatrixMatrixMultiplyNaive_Static_3x9x9 115 ns 115 ns 6104752
BM_MatrixMatrixMultiplyNaive_Dynamic_2x3x4 9.66 ns 9.66 ns 74722971
BM_MatrixMatrixMultiplyNaive_Dynamic_3x3x3 13.0 ns 13.0 ns 53435308
BM_MatrixMatrixMultiplyNaive_Dynamic_4x4x4 47.8 ns 47.8 ns 14358184
BM_MatrixMatrixMultiplyNaive_Dynamic_8x8x8 200 ns 200 ns 3572809
BM_MatrixMatrixMultiplyNaive_Dynamic_9x9x3 104 ns 104 ns 6793797
BM_MatrixMatrixMultiplyNaive_Dynamic_9x3x3 34.0 ns 34.0 ns 20790695
BM_MatrixMatrixMultiplyNaive_Dynamic_3x9x9 130 ns 130 ns 5170402
BM_MatrixTransposeMatrixMultiplyEigen_Static_2x3x4 10.3 ns 10.3 ns 69105234
BM_MatrixTransposeMatrixMultiplyEigen_Static_3x3x3 28.9 ns 28.9 ns 24478934
BM_MatrixTransposeMatrixMultiplyEigen_Static_4x4x4 23.7 ns 23.7 ns 29351926
BM_MatrixTransposeMatrixMultiplyEigen_Static_8x8x8 233 ns 233 ns 2929398
BM_MatrixTransposeMatrixMultiplyEigen_Static_9x9x3 211 ns 211 ns 3287409
BM_MatrixTransposeMatrixMultiplyEigen_Static_9x3x3 26.5 ns 26.5 ns 26515136
BM_MatrixTransposeMatrixMultiplyEigen_Static_3x9x9 196 ns 196 ns 3594314
BM_MatrixTransposeMatrixMultiplyEigen_Dynamic_2x3x4 9.05 ns 9.05 ns 77621001
BM_MatrixTransposeMatrixMultiplyEigen_Dynamic_3x3x3 11.1 ns 11.1 ns 62227812
BM_MatrixTransposeMatrixMultiplyEigen_Dynamic_4x4x4 25.5 ns 25.5 ns 27356089
BM_MatrixTransposeMatrixMultiplyEigen_Dynamic_8x8x8 248 ns 248 ns 2834983
BM_MatrixTransposeMatrixMultiplyEigen_Dynamic_9x9x3 229 ns 229 ns 3082369
BM_MatrixTransposeMatrixMultiplyEigen_Dynamic_9x3x3 28.4 ns 28.4 ns 24318629
BM_MatrixTransposeMatrixMultiplyEigen_Dynamic_3x9x9 229 ns 229 ns 3091288
BM_MatrixTransposeMatrixMultiplyNaive_Static_2x3x4 11.0 ns 11.0 ns 63773538
BM_MatrixTransposeMatrixMultiplyNaive_Static_3x3x3 19.1 ns 19.1 ns 37003139
BM_MatrixTransposeMatrixMultiplyNaive_Static_4x4x4 49.1 ns 49.1 ns 14142301
BM_MatrixTransposeMatrixMultiplyNaive_Static_8x8x8 244 ns 244 ns 2874755
BM_MatrixTransposeMatrixMultiplyNaive_Static_9x9x3 140 ns 140 ns 4992156
BM_MatrixTransposeMatrixMultiplyNaive_Static_9x3x3 46.2 ns 46.2 ns 15068317
BM_MatrixTransposeMatrixMultiplyNaive_Static_3x9x9 112 ns 112 ns 6213574
BM_MatrixTransposeMatrixMultiplyNaive_Dynamic_2x3x4 9.74 ns 9.74 ns 72155001
BM_MatrixTransposeMatrixMultiplyNaive_Dynamic_3x3x3 11.5 ns 11.5 ns 60070577
BM_MatrixTransposeMatrixMultiplyNaive_Dynamic_4x4x4 52.5 ns 52.5 ns 13473642
BM_MatrixTransposeMatrixMultiplyNaive_Dynamic_8x8x8 224 ns 224 ns 3124264
BM_MatrixTransposeMatrixMultiplyNaive_Dynamic_9x9x3 98.0 ns 98.0 ns 7199292
BM_MatrixTransposeMatrixMultiplyNaive_Dynamic_9x3x3 34.7 ns 34.6 ns 20203685
BM_MatrixTransposeMatrixMultiplyNaive_Dynamic_3x9x9 105 ns 105 ns 6653151
Change-Id: Iee403b4d27801d1614ecc1f78a1f8c0011514bd7
2021-08-09 21:24:17 -07:00
c3129c3d4a
Fix tests not executing
...
- The operator() wasn't called and tests where not executed.
- Fill test matrices with non-constant data.
Change-Id: Iacc9e8cb47d8539aded05bb821d3ce2f08a740d6
2021-08-03 14:04:01 -07:00
4fbe218f24
Refactor small_blas_test
...
- Add testMatrix{Transpose}MatrixMultiply size
templated functions and test multiple problem
sizes.
- Removes duplicated code for matrix{transpose}-
matrix{naive} test cases
Change-Id: I9a5d4b0362732ea7bef58572edcaada778fb4aa0
2021-07-30 13:18:18 -07:00
0c85c40923
Revert "Reduce copies involved in Jet operations"
...
This reverts commit c036c78196
2021-07-01 22:55:31 -07:00
7b2c223be7
Add fmax/fmin overloads for scalars
...
Comparison operators are defined for Jet-scalar and scalar-Jet.
Add corresponding overloads for fmax and fmix so users
don't have to convert to Jet only to access these functions,
plus it can potentially save an unnecessary conversion.
Change-Id: I20cc2d5874935e019d4ecb2d6a8c2ae631c8a4d3
2021-06-23 13:40:58 -07:00
c036c78196
Reduce copies involved in Jet operations
...
- Avoid copies in compound assignment operators (+=, -=, *=, /=)
- Add r-value overloads to +,-,*./ operators
Addition<3>/1000 +0.0382 +0.0382 2798 2905 2798 2905
Addition<10>/1000 -0.0009 -0.0009 4831 4827 4831 4827
Addition<15>/1000 -0.0011 -0.0010 8005 7996 8005 7996
Addition<25>/1000 +0.0014 +0.0014 13951 13970 13951 13970
Addition<32>/1000 +0.0010 +0.0009 13638 13651 13638 13651
Addition<200>/160 -0.1732 -0.1732 40301 33321 40300 33320
AdditionScalar<3>/1000 +0.0016 +0.0016 2493 2497 2493 2497
AdditionScalar<10>/1000 +0.0060 +0.0060 2701 2718 2701 2717
AdditionScalar<15>/1000 -0.0024 -0.0024 2598 2592 2598 2592
AdditionScalar<25>/1000 -0.0038 -0.0038 3730 3716 3730 3716
AdditionScalar<32>/1000 +0.0062 +0.0062 4137 4162 4137 4162
AdditionScalar<200>/160 -0.5626 -0.5626 19355 8466 19355 8466
Subtraction<3>/1000 -0.0011 -0.0012 3162 3158 3162 3158
Subtraction<10>/1000 -0.0027 -0.0026 4982 4969 4982 4969
Subtraction<15>/1000 -0.0112 -0.0112 7820 7733 7820 7733
Subtraction<25>/1000 +0.0181 +0.0181 13088 13325 13088 13325
Subtraction<32>/1000 -0.0007 -0.0007 14514 14505 14514 14504
Subtraction<200>/160 -0.1668 -0.1668 44586 37148 44586 37147
SubtractionScalar<3>/1000 -0.0028 -0.0028 2540 2533 2540 2533
SubtractionScalar<10>/1000 -0.0024 -0.0024 2682 2676 2682 2676
SubtractionScalar<15>/1000 -0.0023 -0.0023 2603 2597 2603 2597
SubtractionScalar<25>/1000 -0.0018 -0.0018 3570 3563 3570 3563
SubtractionScalar<32>/1000 -0.0012 -0.0012 4223 4218 4223 4218
SubtractionScalar<200>/160 -0.5070 -0.5070 20375 10044 20375 10044
Multiplication<3>/1000 -0.0543 -0.0544 6171 5836 6172 5836
Multiplication<10>/1000 -0.0617 -0.0617 11860 11128 11860 11128
Multiplication<15>/1000 -0.0106 -0.0106 17456 17271 17456 17271
Multiplication<25>/1000 -0.0114 -0.0114 27204 26893 27204 26893
Multiplication<32>/1000 -0.0125 -0.0125 32704 32293 32703 32293
Multiplication<200>/160 -0.3849 -0.3849 60557 37246 60557 37246
MultiplicationLeftScalar<3>/1000 +0.0005 +0.0005 3086 3088 3086 3088
MultiplicationLeftScalar<10>/1000 -0.0008 -0.0008 4486 4483 4486 4483
MultiplicationLeftScalar<15>/1000 -0.0010 -0.0010 5840 5834 5840 5834
MultiplicationLeftScalar<25>/1000 -0.0047 -0.0047 11350 11297 11350 11298
MultiplicationLeftScalar<32>/1000 -0.0014 -0.0014 13892 13873 13892 13873
MultiplicationLeftScalar<200>/160 -0.2646 -0.2646 28124 20682 28124 20682
MultiplicationRightScalar<3>/1000 +0.0024 +0.0025 3090 3098 3090 3098
MultiplicationRightScalar<10>/1000 +0.0008 +0.0009 4596 4600 4596 4600
MultiplicationRightScalar<15>/1000 -0.0033 -0.0033 5832 5812 5832 5812
MultiplicationRightScalar<25>/1000 -0.0005 -0.0005 11353 11347 11353 11347
MultiplicationRightScalar<32>/1000 +0.0043 +0.0043 14041 14102 14041 14102
MultiplicationRightScalar<200>/160 -0.2600 -0.2600 28104 20796 28104 20796
Division<3>/1000 -0.0006 -0.0006 8977 8972 8977 8972
Division<10>/1000 -0.0095 -0.0095 14613 14474 14613 14473
Division<15>/1000 -0.0081 -0.0081 23109 22922 23108 22922
Division<25>/1000 -0.0081 -0.0081 37292 36992 37292 36992
Division<32>/1000 -0.0234 -0.0234 41693 40718 41692 40717
Division<200>/160 -0.1306 -0.1306 44486 38677 44486 38677
DivisionLeftScalar<3>/1000 -0.0080 -0.0080 6683 6630 6683 6630
DivisionLeftScalar<10>/1000 -0.0018 -0.0018 9255 9239 9255 9238
DivisionLeftScalar<15>/1000 -0.0011 -0.0011 12180 12166 12180 12166
DivisionLeftScalar<25>/1000 +0.0060 +0.0060 22381 22515 22381 22516
DivisionLeftScalar<32>/1000 -0.0016 -0.0016 20159 20126 20159 20126
DivisionLeftScalar<200>/160 -0.2528 -0.2528 28773 21499 28772 21499
DivisionRightScalar<3>/1000 -0.0006 -0.0006 3784 3782 3784 3782
DivisionRightScalar<10>/1000 -0.0014 -0.0014 5699 5691 5699 5691
DivisionRightScalar<15>/1000 -0.0009 -0.0009 7490 7484 7490 7484
DivisionRightScalar<25>/1000 +0.0031 +0.0031 13751 13793 13751 13793
DivisionRightScalar<32>/1000 -0.0006 -0.0006 16712 16702 16712 16702
DivisionRightScalar<200>/160 -0.2532 -0.2532 28412 21217 28411 21217
MultiplyAndAdd<3>/1000 -0.0191 -0.0191 4382 4299 4382 4299
MultiplyAndAdd<10>/1000 -0.0060 -0.0060 10238 10176 10238 10176
MultiplyAndAdd<15>/1000 -0.0053 -0.0053 11696 11634 11696 11634
MultiplyAndAdd<25>/1000 -0.0171 -0.0171 19185 18858 19185 18858
MultiplyAndAdd<32>/1000 +0.0070 +0.0070 23783 23951 23783 23951
MultiplyAndAdd<200>/160 -0.1252 -0.1252 61470 53776 61469 53776
Change-Id: Ib366d5753ed4f398667766384e05389d45e0d444
2021-06-18 15:04:17 -07:00
51945e0619
Introduce benchmark for Jet operations
...
Run on (20 X 4300 MHz CPU s)
CPU Caches:
L1 Data 32 KiB (x10)
L1 Instruction 32 KiB (x10)
L2 Unified 1024 KiB (x10)
L3 Unified 14080 KiB (x1)
Load Average: 2.37, 3.00, 3.08
-----------------------------------------------------------------------------
Benchmark Time CPU Iterations
-----------------------------------------------------------------------------
Addition<3>/1000 2872 ns 2872 ns 240449
Addition<10>/1000 5304 ns 5304 ns 100000
Addition<15>/1000 8211 ns 8210 ns 78742
Addition<25>/1000 14214 ns 14213 ns 46762
Addition<32>/1000 13746 ns 13746 ns 50892
Addition<200>/160 41228 ns 41228 ns 17183
AdditionScalar<3>/1000 2514 ns 2514 ns 273996
AdditionScalar<10>/1000 2733 ns 2733 ns 255508
AdditionScalar<15>/1000 2622 ns 2622 ns 264291
AdditionScalar<25>/1000 3753 ns 3753 ns 183508
AdditionScalar<32>/1000 4254 ns 4254 ns 167016
AdditionScalar<200>/160 18314 ns 18314 ns 38116
Subtraction<3>/1000 3241 ns 3241 ns 206370
Subtraction<10>/1000 5023 ns 5023 ns 139271
Subtraction<15>/1000 8387 ns 8386 ns 89927
Subtraction<25>/1000 14951 ns 14950 ns 48756
Subtraction<32>/1000 14587 ns 14587 ns 47056
Subtraction<200>/160 47175 ns 47175 ns 15574
SubtractionScalar<3>/1000 2572 ns 2572 ns 264468
SubtractionScalar<10>/1000 2713 ns 2713 ns 257920
SubtractionScalar<15>/1000 2621 ns 2621 ns 265289
SubtractionScalar<25>/1000 3593 ns 3593 ns 192266
SubtractionScalar<32>/1000 4255 ns 4255 ns 163738
SubtractionScalar<200>/160 19906 ns 19906 ns 35295
Multiplication<3>/1000 6058 ns 6058 ns 114067
Multiplication<10>/1000 11999 ns 11999 ns 58492
Multiplication<15>/1000 17906 ns 17905 ns 39565
Multiplication<25>/1000 27361 ns 27360 ns 25335
Multiplication<32>/1000 33074 ns 33074 ns 20875
Multiplication<200>/160 61364 ns 61362 ns 11542
MultiplicationLeftScalar<3>/1000 3104 ns 3104 ns 223720
MultiplicationLeftScalar<10>/1000 4549 ns 4549 ns 154366
MultiplicationLeftScalar<15>/1000 5921 ns 5921 ns 119294
MultiplicationLeftScalar<25>/1000 11429 ns 11428 ns 61685
MultiplicationLeftScalar<32>/1000 14094 ns 14094 ns 49941
MultiplicationLeftScalar<200>/160 28186 ns 28185 ns 24484
MultiplicationRightScalar<3>/1000 3110 ns 3110 ns 223333
MultiplicationRightScalar<10>/1000 4655 ns 4655 ns 150534
MultiplicationRightScalar<15>/1000 5890 ns 5890 ns 119746
MultiplicationRightScalar<25>/1000 11464 ns 11464 ns 61483
MultiplicationRightScalar<32>/1000 14243 ns 14242 ns 49492
MultiplicationRightScalar<200>/160 28282 ns 28281 ns 24604
Division<3>/1000 9128 ns 9128 ns 77846
Division<10>/1000 14811 ns 14811 ns 47682
Division<15>/1000 23293 ns 23292 ns 30091
Division<25>/1000 37313 ns 37313 ns 18608
Division<32>/1000 41229 ns 41229 ns 16982
Division<200>/160 44802 ns 44802 ns 15573
DivisionLeftScalar<3>/1000 6720 ns 6720 ns 104747
DivisionLeftScalar<10>/1000 9403 ns 9402 ns 75216
DivisionLeftScalar<15>/1000 12313 ns 12313 ns 57366
DivisionLeftScalar<25>/1000 22739 ns 22739 ns 30421
DivisionLeftScalar<32>/1000 20321 ns 20321 ns 34191
DivisionLeftScalar<200>/160 29018 ns 29017 ns 23908
DivisionRightScalar<3>/1000 3815 ns 3815 ns 182333
DivisionRightScalar<10>/1000 5750 ns 5750 ns 121691
DivisionRightScalar<15>/1000 7574 ns 7574 ns 92994
DivisionRightScalar<25>/1000 13953 ns 13953 ns 49250
DivisionRightScalar<32>/1000 16892 ns 16892 ns 41668
DivisionRightScalar<200>/160 28663 ns 28662 ns 24226
MultiplyAndAdd<3>/1000 4399 ns 4399 ns 158635
MultiplyAndAdd<10>/1000 10453 ns 10453 ns 68112
MultiplyAndAdd<15>/1000 11830 ns 11830 ns 59598
MultiplyAndAdd<25>/1000 19624 ns 19624 ns 36240
MultiplyAndAdd<32>/1000 25539 ns 25538 ns 29066
MultiplyAndAdd<200>/160 65362 ns 65358 ns 11086
Change-Id: Ie62492b3fd19ff9d3394f90bd00f0aa01522fc2a
2021-06-16 16:17:52 -07:00
ec4f2995bb
Do not check MaxNumThreadsAvailable if the thread number is set to 1.
...
- this saves the sysfs cost in MaxNumThreadsAvailable.
Change-Id: I6b33ceae363da19346341be95abbd89d7b7b0f2f
2021-03-24 22:16:45 +00:00
98f639f542
Add a macro CERES_GET_FLAG.
...
This is needed to add a layer of indirection which will allow
us to use gflags in the public version and absl inside Google.
Change-Id: I32f3da23200a01c9b658bbf8aaa66cb8fddc2cc3
2021-03-18 11:07:57 -07:00
766f2cab55
Reduce log spam in covariance_impl.cc.
...
Make the rank deficiency warning in the covariance module a VLOG(1) instead of
LOG(ERROR) since this is a commonly occuring case.
Change-Id: Id46eba6dcf9afbaf0554c2d6f230b2af7a1fbdea
2021-03-17 13:55:00 +00:00
2b32b32124
Revert "Group specializations into groups of four"
...
This reverts commit 4ba244cdbc
2021-02-17 18:38:43 +00:00
313caf1aec
Allow Unity build.
...
Two parts:
- missing include guard in solver_utils.h
- exclude schur_specializations from unity build. This
can only be done with cmake >= 3.16, so only allow
unity build for these versions.
Cuts build time by 5 minutes with default settings.
libceres.a shrinks from 7549968 bytes to 7009696 bytes.
Change-Id: I833ac8266623da0e725cc2f07a6415b4fa72bc18
2021-02-16 08:30:36 +09:00
4ba244cdbc
Group specializations into groups of four
...
Genrerate the specialization files by grouping the specializations into
five groups, cutting build time by three minutes (roughly 20%) in my
tests.
Change-Id: Ifd6a4523a47346c5e84c1cda15b7abba60d8e953
2021-02-12 19:56:53 +09:00
d77a8100a3
Make miniglog's InitGoogleLogging argument const.
...
This is in line with glog. Also, the argument is unused. Third try.
Change-Id: I23331ae26089956454d66ab9e93fc04d4b24ed2e
2021-02-08 00:56:02 +00:00
d15b1bcd3e
Increase tolerance in small_blas_test.cc
...
Change-Id: I7b8460bec65e56029d01c293e83743e99a57d10c
2021-01-12 09:15:05 -08:00
64029909b3
Fix -Wno-maybe-uninitialized error
...
One of my compilers couldn't deduce that Chunk was assigned before being
used based on the context. Make it more obvious...
Change-Id: I1e331c6c02a2783eff8db03c6359823ac3c46686
2021-01-04 14:44:49 -08:00
21294123d7
Fix nonnull arg compared to NULL error.
...
external/com_google_ceres_solver/internal/ceres/rotation_test.cc: In member function 'bool ceres::internal::IsNormalizedQuaternionMatcher::gmock_Impl<arg_type>::MatchAndExplain(typename testing::internal::ConstRef<arg_type>::type, testing::MatchResultListener*) const [with arg_type = const double (&)[4]]':
external/com_google_ceres_solver/internal/ceres/rotation_test.cc:74:3: error: nonnull argument 'arg' compared to NULL [-Werror=nonnull-compare]
if (arg == NULL) {
^~
Change-Id: I21abd87e97e44188580f4926b3700dc321fecb0e
2021-01-04 14:44:11 -08:00
1dd4174109
Fix -Wno-format-nonliteral
...
GCC doesn't like printf formats which aren't char strings.
Change-Id: I41464ba336076b892d893a5729611488c159f01d
2021-01-04 14:40:44 -08:00
6c106bf510
Fix -Wmissing-field-initializers error
...
Clang was complaining that fields were uninitialized.
Change-Id: Iff76fb2ebf823eb072465c0b0381a4a5cdee0fa9
2020-12-23 22:50:28 -08:00
e0e14a5cdb
Fix errors found by -Werror
...
"'const' type qualifier on return type has no effect" is one of the
errors, and an unused parameter for another.
Change-Id: Ib4bc245d992c1fc1bea0b25d5396871ca1df8411
2020-12-23 22:37:54 -08:00
d45ec47b58
Fix a typo in schur_eliminator.h
...
Change-Id: Ib2c6ea9598d49ee77e2f0429686045cf58cd8574
2020-11-18 08:59:06 -08:00
bb127272f9
Fix typos.
...
Contributed by Ishamis@, IanBoyanZhang@, gkrobner@ & mithunjacob@.
Change-Id: Iab3c19a07a6f3db2486e3557dcb55bfe5de2aee5
2020-10-19 09:32:47 -07:00
3f6d273676
Unify symbol visibility configuration for all compilers
...
This makes it possible to build unit tests with shared libraries on MSVC.
Change-Id: I1db66a80b2c78c4f3d354e35235244d17bac9809
2020-10-15 16:56:07 -04:00
600e8c529e
fix minor typos
...
all timing values in the summary are initialized to -1, so the one
+1 is likely an oversight.
Change-Id: Ie355f3b7da08a56d49d19ca9a5bc48fe5581dee3
2020-10-12 23:01:02 +02:00
8bfdb02fb1
Rewrite uses of VLOG_IF and LOG_IF.
...
VLOG_IF's evaluation order is ambiguous - does it mean
`if (cond) VLOG(lvl)` or `if (VLOG_IS_ON(lvl) && cond) LOG(INFO)`?
In particular, the way it works now is inconsistent with the way the
rest of the LOG macros evaluate their arguments.
Fixing this would be hard, and the macro's behavior would still surprise
some people. Replacing it with an if statement is simple, clear, and unambiguous.
Change-Id: I97a92d17a932c0a5344a1bf98d676308793ba877
2020-10-12 18:34:24 +00:00
542613c13d
minor formatting fix for trust_region_minimizer.cc
...
Change-Id: I18ba27825fc23dd0e9e3e15dc13fc0833db01b5b
2020-10-06 22:49:32 +02:00
6d9e9843d8
Remove inclusion of ceres/eigen.h
...
The initial reason for this is because of a previous reformatting CL
triggered a macro redefinition warning in the schur eliminator. But
actually it was worse because the reordering had caused the macro
definition to be ignored and caused a performance regression.
This simplifies the generated files, fixes some formatting errors
and recovers the performance.
Change-Id: I9dbeffc38743b3f24b25843feec2e26a73188413
2020-09-28 19:20:24 +00:00
eafeca5dcb
Fix a logging bug in TrustRegionMinimizer.
...
Upon encountering an unsuccessful step (one where the cost goes up)
the the trust region minimizer failed to populate the gradient norm
in the IterationSummary. This would cause the gradient norm to be
logged as zero which is incorrect. Instead it should be the gradient
norm at the current point.
This CL fixes this issue.
Before:
iter cost cost_change |gradient| |step| tr_ratio tr_radius ls_iter iter_time total_time
0 1.115206e+07 0.00e+00 1.90e+07 0.00e+00 0.00e+00 1.00e+04 0 2.72e-01 1.33e+00
1 3.687552e+06 7.46e+06 1.84e+08 2.86e+03 6.91e-01 1.06e+04 1 1.32e+00 2.65e+00
2 3.670266e+10 -3.67e+10 0.00e+00 3.27e+03 -1.07e+04 5.30e+03 1 7.52e-01 3.40e+00
3 4.335397e+07 -3.97e+07 0.00e+00 2.74e+03 -1.16e+01 1.32e+03 1 7.28e-01 4.13e+00
4 1.345488e+06 2.34e+06 4.12e+07 1.55e+03 6.87e-01 1.40e+03 1 9.31e-01 5.06e+00
5 5.376653e+05 8.08e+05 9.99e+06 6.64e+02 7.46e-01 1.59e+03 1 9.64e-01 6.03e+00
After:
iter cost cost_change |gradient| |step| tr_ratio tr_radius ls_iter iter_time total_time
0 1.115206e+07 0.00e+00 1.90e+07 0.00e+00 0.00e+00 1.00e+04 0 2.37e-01 1.13e+00
1 3.687552e+06 7.46e+06 1.84e+08 2.86e+03 6.91e-01 1.06e+04 1 1.08e+00 2.21e+00
2 3.670266e+10 -3.67e+10 1.84e+08 3.27e+03 -1.07e+04 5.30e+03 1 7.50e-01 2.96e+00
3 4.335397e+07 -3.97e+07 1.84e+08 2.74e+03 -1.16e+01 1.32e+03 1 7.13e-01 3.67e+00
4 1.345488e+06 2.34e+06 4.12e+07 1.55e+03 6.87e-01 1.40e+03 1 9.01e-01 4.57e+00
5 5.376653e+05 8.08e+05 9.99e+06 6.64e+02 7.46e-01 1.59e+03 1 9.36e-01 5.51e+00
Change-Id: Iae538fe089be07c7bb219337a6f1392f7213acfe
2020-09-28 12:19:45 -07:00
d3f66d77f4
fix formatting generated files (best effort)
...
- update file generator scripts / templates so generated files adhere
to clang-format
- A few exceptions are not fixed, where the file generation results in
lines of different width. To properly fix this would make the code
more complicated and it's not that important for generated files
anyway.
- note that generated files are excluded in ./scripts/format_all.sh
Change-Id: I4f42c83d1fec01242eada5e7ce6c1a5192234d37
2020-09-28 02:24:43 +02:00
a9c7361c8d
minor formatting fix (wrongly updated in earlier commit)
...
Change-Id: I544635fd936cb5b7f7bd9255876641cd5a9590c6
2020-09-28 02:19:02 +02:00
7b8f675bfd
fix formatting for (non-generated) internal source files
...
- Change formatting standard to Cpp11. Main difference is not having
the space between two closing >> for nested templates. We don't
choose c++14, because older versions of clang-format (version 9
and earlier) don't know this value yet, and it doesn't make a
difference in the formatting.
- Apply clang-format to all (non generated) internal source files.
- Manually fix some code sections (clang-format on/off) and c-strings
- Exclude some embedded external files with very different formatting
(gtest/gmock)
- Add script to format all source files
Change-Id: Ic6cea41575ad6e37c9e136dbce176b0d505dc44d
2020-09-21 02:52:07 +02:00
921368ce31
Fix a number of typos in covariance.h
...
Also some minor cleanups in covariance_impl.h
Thanks to Lorenzo Lamia for pointing these out.
Change-Id: Icb4012a367fdd1f249bc1e7019e0114c868e45b6
2020-09-09 09:39:42 -07:00
8cbd721c19
Add erf and erfc to jet.h, including tests in jet_test.cc
...
erf is necessary for evaluating Gaussian functions.
erfc was added because it is so similar to erf.
Change-Id: I5e470dbe013cc938fabb87cde3b0ebf26a90fff4
2020-09-03 17:54:20 +02:00
31366cff29
Benchmarks for dynamic autodiff.
...
This patch is from Clement Courbet. courbet@google.com
Change-Id: I886390663644733bfa5b7b52b0c883079e793726
2020-09-01 09:25:02 -07:00
41d05f13d0
Fix lint errors in evaluation_callback_test.cc
...
Change-Id: I63eb069544ad0d8f495490fe4caa07b9f04f7ec2
2020-08-04 14:56:33 -07:00
4b67903c1f
Remove unused variables from problem_test.cc
...
Change-Id: Ia1a13cfc6e462f6d249dcbf169ad34831dd93ec2
2020-08-04 14:40:50 -07:00
Sergiu Deitsch