Files
ceres-solver/internal/ceres/types.cc
T
Joydeep Biswas 829089053e CUDA CGNR, Part 4: CudaCgnrSolver
* Added CudaCgnrSolver, a new CUDA-accelerated CGNR.
* To use CudaCgnrSolver, the user must select CGNR as the linear_solver
  and CUDA_SPARSE as the sparse_linear_algebra_library.
* Updated ConjugateGradientSolver to work with an array of pointers to
  scratch to support CudaVectors as scratch.
* Moved CUDA initialization to run in Solver::Solve as needed.

Some performance comparisons on an Ubuntu 20.04 desktop with an
Intel i9-9940X CPU @ 3.30GHz, and an nVidia Quadro RTX 6000,
all configurations run with 24 threads, and 10 iterations.

=================================================
CGNR + CUDA_SPARSE + IDENTITY Preconditioner
problem-1778-993923-pre.txt
=================================================
Cost:
Initial                          2.563973e+08
Final                            1.724755e+06
Change                           2.546725e+08

Minimizer iterations                       11
Successful steps                            7
Unsuccessful steps                          4

Time (in seconds):
Preprocessor                         4.020158

  Residual only evaluation           1.567092 (10)
  Jacobian & residual evaluation     7.847130 (7)
  Linear solver                     31.688898 (10)
Minimizer                           46.834987

Postprocessor                        0.353974
Total                               51.209120

=================================================
SPARSE_SCHUR (CPU) + SUITE_SPARSE + AMD
problem-1778-993923-pre.txt
=================================================
Cost:
Initial                          2.563973e+08
Final                            1.651617e+06
Change                           2.547457e+08

Minimizer iterations                       11
Successful steps                           11
Unsuccessful steps                          0

Time (in seconds):
Preprocessor                        35.812003

  Residual only evaluation           1.658980 (10)
  Jacobian & residual evaluation    12.218799 (11)
  Linear solver                     76.409992 (10)
Minimizer                           98.809773

Postprocessor                        0.372712
Total                              134.994489

=================================================
ITERATIVE_SCHUR (CPU) + JACOBI Preconditioner
problem-1778-993923-pre.txt
=================================================
Cost:
Initial                          2.563973e+08
Final                            1.684447e+06
Change                           2.547128e+08

Minimizer iterations                       11
Successful steps                            8
Unsuccessful steps                          3

Time (in seconds):
Preprocessor                        15.331614

  Residual only evaluation           1.606114 (10)
  Jacobian & residual evaluation     8.502166 (8)
  Linear solver                    351.910080 (10)
Minimizer                          368.797327

Postprocessor                        0.363536
Total                              384.492478

=================================================
CGNR + CUDA_SPARSE + IDENTITY Preconditioner
problem-13682-4456117-pre.txt
=================================================
Cost:
Initial                          1.126372e+09
Final                            2.269329e+07
Change                           1.103678e+09

Minimizer iterations                       11
Successful steps                            7
Unsuccessful steps                          4

Time (in seconds):
Preprocessor                        19.140087

  Residual only evaluation           8.721920 (10)
  Jacobian & residual evaluation    41.955923 (7)
  Linear solver                    214.121861 (10)
Minimizer                          296.636890

Postprocessor                        1.971827
Total                              317.748804

Change-Id: I3a09f31aa6903f661e91f595afd39d427583e856
2022-08-15 23:44:34 -05:00

468 lines
11 KiB
C++

// Ceres Solver - A fast non-linear least squares minimizer
// Copyright 2015 Google Inc. All rights reserved.
// http://ceres-solver.org/
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are met:
//
// * Redistributions of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
// * Neither the name of Google Inc. nor the names of its contributors may be
// used to endorse or promote products derived from this software without
// specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
// ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
// LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
// CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
// SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
// INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
// CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
// ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
// POSSIBILITY OF SUCH DAMAGE.
//
// Author: sameeragarwal@google.com (Sameer Agarwal)
#include "ceres/types.h"
#include <algorithm>
#include <cctype>
#include <string>
#include "ceres/internal/config.h"
#include "glog/logging.h"
namespace ceres {
using std::string;
// clang-format off
#define CASESTR(x) case x: return #x
#define STRENUM(x) if (value == #x) { *type = x; return true; }
// clang-format on
static void UpperCase(string* input) {
std::transform(input->begin(), input->end(), input->begin(), ::toupper);
}
const char* LinearSolverTypeToString(LinearSolverType type) {
switch (type) {
CASESTR(DENSE_NORMAL_CHOLESKY);
CASESTR(DENSE_QR);
CASESTR(SPARSE_NORMAL_CHOLESKY);
CASESTR(DENSE_SCHUR);
CASESTR(SPARSE_SCHUR);
CASESTR(ITERATIVE_SCHUR);
CASESTR(CGNR);
default:
return "UNKNOWN";
}
}
bool StringToLinearSolverType(string value, LinearSolverType* type) {
UpperCase(&value);
STRENUM(DENSE_NORMAL_CHOLESKY);
STRENUM(DENSE_QR);
STRENUM(SPARSE_NORMAL_CHOLESKY);
STRENUM(DENSE_SCHUR);
STRENUM(SPARSE_SCHUR);
STRENUM(ITERATIVE_SCHUR);
STRENUM(CGNR);
return false;
}
const char* PreconditionerTypeToString(PreconditionerType type) {
switch (type) {
CASESTR(IDENTITY);
CASESTR(JACOBI);
CASESTR(SCHUR_JACOBI);
CASESTR(SCHUR_POWER_SERIES_EXPANSION);
CASESTR(CLUSTER_JACOBI);
CASESTR(CLUSTER_TRIDIAGONAL);
CASESTR(SUBSET);
default:
return "UNKNOWN";
}
}
bool StringToPreconditionerType(string value, PreconditionerType* type) {
UpperCase(&value);
STRENUM(IDENTITY);
STRENUM(JACOBI);
STRENUM(SCHUR_JACOBI);
STRENUM(SCHUR_POWER_SERIES_EXPANSION);
STRENUM(CLUSTER_JACOBI);
STRENUM(CLUSTER_TRIDIAGONAL);
STRENUM(SUBSET);
return false;
}
const char* SparseLinearAlgebraLibraryTypeToString(
SparseLinearAlgebraLibraryType type) {
switch (type) {
CASESTR(SUITE_SPARSE);
CASESTR(EIGEN_SPARSE);
CASESTR(ACCELERATE_SPARSE);
CASESTR(CUDA_SPARSE);
CASESTR(NO_SPARSE);
default:
return "UNKNOWN";
}
}
bool StringToSparseLinearAlgebraLibraryType(
string value, SparseLinearAlgebraLibraryType* type) {
UpperCase(&value);
STRENUM(SUITE_SPARSE);
STRENUM(EIGEN_SPARSE);
STRENUM(ACCELERATE_SPARSE);
STRENUM(CUDA_SPARSE);
STRENUM(NO_SPARSE);
return false;
}
const char* LinearSolverOrderingTypeToString(LinearSolverOrderingType type) {
switch (type) {
CASESTR(AMD);
CASESTR(NESDIS);
default:
return "UNKNOWN";
}
}
bool StringToLinearSolverOrderingType(string value,
LinearSolverOrderingType* type) {
UpperCase(&value);
STRENUM(AMD);
STRENUM(NESDIS);
return false;
}
const char* DenseLinearAlgebraLibraryTypeToString(
DenseLinearAlgebraLibraryType type) {
switch (type) {
CASESTR(EIGEN);
CASESTR(LAPACK);
CASESTR(CUDA);
default:
return "UNKNOWN";
}
}
bool StringToDenseLinearAlgebraLibraryType(
string value, DenseLinearAlgebraLibraryType* type) {
UpperCase(&value);
STRENUM(EIGEN);
STRENUM(LAPACK);
STRENUM(CUDA);
return false;
}
const char* TrustRegionStrategyTypeToString(TrustRegionStrategyType type) {
switch (type) {
CASESTR(LEVENBERG_MARQUARDT);
CASESTR(DOGLEG);
default:
return "UNKNOWN";
}
}
bool StringToTrustRegionStrategyType(string value,
TrustRegionStrategyType* type) {
UpperCase(&value);
STRENUM(LEVENBERG_MARQUARDT);
STRENUM(DOGLEG);
return false;
}
const char* DoglegTypeToString(DoglegType type) {
switch (type) {
CASESTR(TRADITIONAL_DOGLEG);
CASESTR(SUBSPACE_DOGLEG);
default:
return "UNKNOWN";
}
}
bool StringToDoglegType(string value, DoglegType* type) {
UpperCase(&value);
STRENUM(TRADITIONAL_DOGLEG);
STRENUM(SUBSPACE_DOGLEG);
return false;
}
const char* MinimizerTypeToString(MinimizerType type) {
switch (type) {
CASESTR(TRUST_REGION);
CASESTR(LINE_SEARCH);
default:
return "UNKNOWN";
}
}
bool StringToMinimizerType(string value, MinimizerType* type) {
UpperCase(&value);
STRENUM(TRUST_REGION);
STRENUM(LINE_SEARCH);
return false;
}
const char* LineSearchDirectionTypeToString(LineSearchDirectionType type) {
switch (type) {
CASESTR(STEEPEST_DESCENT);
CASESTR(NONLINEAR_CONJUGATE_GRADIENT);
CASESTR(LBFGS);
CASESTR(BFGS);
default:
return "UNKNOWN";
}
}
bool StringToLineSearchDirectionType(string value,
LineSearchDirectionType* type) {
UpperCase(&value);
STRENUM(STEEPEST_DESCENT);
STRENUM(NONLINEAR_CONJUGATE_GRADIENT);
STRENUM(LBFGS);
STRENUM(BFGS);
return false;
}
const char* LineSearchTypeToString(LineSearchType type) {
switch (type) {
CASESTR(ARMIJO);
CASESTR(WOLFE);
default:
return "UNKNOWN";
}
}
bool StringToLineSearchType(string value, LineSearchType* type) {
UpperCase(&value);
STRENUM(ARMIJO);
STRENUM(WOLFE);
return false;
}
const char* LineSearchInterpolationTypeToString(
LineSearchInterpolationType type) {
switch (type) {
CASESTR(BISECTION);
CASESTR(QUADRATIC);
CASESTR(CUBIC);
default:
return "UNKNOWN";
}
}
bool StringToLineSearchInterpolationType(string value,
LineSearchInterpolationType* type) {
UpperCase(&value);
STRENUM(BISECTION);
STRENUM(QUADRATIC);
STRENUM(CUBIC);
return false;
}
const char* NonlinearConjugateGradientTypeToString(
NonlinearConjugateGradientType type) {
switch (type) {
CASESTR(FLETCHER_REEVES);
CASESTR(POLAK_RIBIERE);
CASESTR(HESTENES_STIEFEL);
default:
return "UNKNOWN";
}
}
bool StringToNonlinearConjugateGradientType(
string value, NonlinearConjugateGradientType* type) {
UpperCase(&value);
STRENUM(FLETCHER_REEVES);
STRENUM(POLAK_RIBIERE);
STRENUM(HESTENES_STIEFEL);
return false;
}
const char* CovarianceAlgorithmTypeToString(CovarianceAlgorithmType type) {
switch (type) {
CASESTR(DENSE_SVD);
CASESTR(SPARSE_QR);
default:
return "UNKNOWN";
}
}
bool StringToCovarianceAlgorithmType(string value,
CovarianceAlgorithmType* type) {
UpperCase(&value);
STRENUM(DENSE_SVD);
STRENUM(SPARSE_QR);
return false;
}
const char* NumericDiffMethodTypeToString(NumericDiffMethodType type) {
switch (type) {
CASESTR(CENTRAL);
CASESTR(FORWARD);
CASESTR(RIDDERS);
default:
return "UNKNOWN";
}
}
bool StringToNumericDiffMethodType(string value, NumericDiffMethodType* type) {
UpperCase(&value);
STRENUM(CENTRAL);
STRENUM(FORWARD);
STRENUM(RIDDERS);
return false;
}
const char* VisibilityClusteringTypeToString(VisibilityClusteringType type) {
switch (type) {
CASESTR(CANONICAL_VIEWS);
CASESTR(SINGLE_LINKAGE);
default:
return "UNKNOWN";
}
}
bool StringToVisibilityClusteringType(string value,
VisibilityClusteringType* type) {
UpperCase(&value);
STRENUM(CANONICAL_VIEWS);
STRENUM(SINGLE_LINKAGE);
return false;
}
const char* TerminationTypeToString(TerminationType type) {
switch (type) {
CASESTR(CONVERGENCE);
CASESTR(NO_CONVERGENCE);
CASESTR(FAILURE);
CASESTR(USER_SUCCESS);
CASESTR(USER_FAILURE);
default:
return "UNKNOWN";
}
}
const char* LoggingTypeToString(LoggingType type) {
switch (type) {
CASESTR(SILENT);
CASESTR(PER_MINIMIZER_ITERATION);
default:
return "UNKNOWN";
}
}
bool StringtoLoggingType(std::string value, LoggingType* type) {
UpperCase(&value);
STRENUM(SILENT);
STRENUM(PER_MINIMIZER_ITERATION);
return false;
}
const char* DumpFormatTypeToString(DumpFormatType type) {
switch (type) {
CASESTR(CONSOLE);
CASESTR(TEXTFILE);
default:
return "UNKNOWN";
}
}
bool StringtoDumpFormatType(std::string value, DumpFormatType* type) {
UpperCase(&value);
STRENUM(CONSOLE);
STRENUM(TEXTFILE);
return false;
}
#undef CASESTR
#undef STRENUM
bool IsSchurType(LinearSolverType type) {
// clang-format off
return ((type == SPARSE_SCHUR) ||
(type == DENSE_SCHUR) ||
(type == ITERATIVE_SCHUR));
// clang-format on
}
bool IsSparseLinearAlgebraLibraryTypeAvailable(
SparseLinearAlgebraLibraryType type) {
if (type == SUITE_SPARSE) {
#ifdef CERES_NO_SUITESPARSE
return false;
#else
return true;
#endif
}
if (type == ACCELERATE_SPARSE) {
#ifdef CERES_NO_ACCELERATE_SPARSE
return false;
#else
return true;
#endif
}
if (type == EIGEN_SPARSE) {
#ifdef CERES_USE_EIGEN_SPARSE
return true;
#else
return false;
#endif
}
if (type == CUDA_SPARSE) {
#ifdef CERES_NO_CUDA
return false;
#else
return true;
#endif
}
if (type == NO_SPARSE) {
return true;
}
LOG(WARNING) << "Unknown sparse linear algebra library " << type;
return false;
}
bool IsDenseLinearAlgebraLibraryTypeAvailable(
DenseLinearAlgebraLibraryType type) {
if (type == EIGEN) {
return true;
}
if (type == LAPACK) {
#ifdef CERES_NO_LAPACK
return false;
#else
return true;
#endif
}
if (type == CUDA) {
#ifdef CERES_NO_CUDA
return false;
#else
return true;
#endif
}
LOG(WARNING) << "Unknown dense linear algebra library " << type;
return false;
}
} // namespace ceres