commit b9f15a59361c609ffc4a328aea9be3d265b5da81
author Sameer Agarwal <sameeragarwal@google.com> Sat Aug 18 13:06:19 2012 -0700
committer Sameer Agarwal <sameeragarwal@google.com> Sun Aug 19 14:47:38 2012 -0700
tree 881d1e7e545e1a9c3bf3701b8034169f73ce0135
parent b3fa009435acf476cd373052e62988f6437970b1

Add a dense Cholesky factorization based linear solver.

For problems with a small number of variables, but a large
number of residuals, it is sometimes beneficial to use the
Cholesky factorization on the normal equations, instead of
the dense QR factorization of the Jacobian, even though it
is numerically the better thing to do.

Change-Id: I3506b006195754018deec964e6e190b7e8c9ac8f

internal/ceres/unsymmetric_linear_solver_test.cc

diff --git a/internal/ceres/unsymmetric_linear_solver_test.cc b/internal/ceres/unsymmetric_linear_solver_test.cc
index f13c5a4..0b0d593 100644
--- a/internal/ceres/unsymmetric_linear_solver_test.cc
+++ b/internal/ceres/unsymmetric_linear_solver_test.cc
@@ -73,7 +73,8 @@
 
     scoped_ptr<SparseMatrix> transformed_A;
 
-    if (linear_solver_type == DENSE_QR) {
+    if (linear_solver_type == DENSE_QR ||
+        linear_solver_type == DENSE_NORMAL_CHOLESKY) {
       transformed_A.reset(new DenseSparseMatrix(*A_));
     } else if (linear_solver_type == SPARSE_NORMAL_CHOLESKY) {
       transformed_A.reset(new   CompressedRowSparseMatrix(*A_));
@@ -118,6 +119,10 @@
   TestSolver(DENSE_QR, SUITE_SPARSE);
 }
 
+TEST_F(UnsymmetricLinearSolverTest, DenseNormalCholesky) {
+  TestSolver(DENSE_NORMAL_CHOLESKY, SUITE_SPARSE);
+}
+
 #ifndef CERES_NO_SUITESPARSE
 TEST_F(UnsymmetricLinearSolverTest, SparseNormalCholeskyUsingSuiteSparse) {
   TestSolver(SPARSE_NORMAL_CHOLESKY, SUITE_SPARSE);