diff --git a/include/ceres/manifold_test_utils.h b/include/ceres/manifold_test_utils.h
index dd1f931..f8ed557 100644
--- a/include/ceres/manifold_test_utils.h
+++ b/include/ceres/manifold_test_utils.h
@@ -76,17 +76,20 @@
 //      EXPECT_THAT_MANIFOLD_INVARIANTS_HOLD(manifold, x, delta, y, kTolerance);
 //    }
 #define EXPECT_THAT_MANIFOLD_INVARIANTS_HOLD(manifold, x, delta, y, tolerance) \
-  ::ceres::Vector zero_tangent = ::ceres::Vector::Zero(manifold.TangentSize()); \
-  EXPECT_THAT(manifold, ::ceres::XPlusZeroIsXAt(x, tolerance));         \
-  EXPECT_THAT(manifold, ::ceres::XMinusXIsZeroAt(x, tolerance));        \
-  EXPECT_THAT(manifold, ::ceres::MinusPlusIsIdentityAt(x, delta, tolerance)); \
-  EXPECT_THAT(manifold, ::ceres::MinusPlusIsIdentityAt(x, zero_tangent, tolerance)); \
-  EXPECT_THAT(manifold, ::ceres::PlusMinusIsIdentityAt(x, x, tolerance)); \
-  EXPECT_THAT(manifold, ::ceres::PlusMinusIsIdentityAt(x, y, tolerance)); \
-  EXPECT_THAT(manifold, ::ceres::HasCorrectPlusJacobianAt(x, tolerance)); \
-  EXPECT_THAT(manifold, ::ceres::HasCorrectMinusJacobianAt(x, tolerance)); \
+  ::ceres::Vector zero_tangent =                                               \
+      ::ceres::Vector::Zero(manifold.TangentSize());                           \
+  EXPECT_THAT(manifold, ::ceres::XPlusZeroIsXAt(x, tolerance));                \
+  EXPECT_THAT(manifold, ::ceres::XMinusXIsZeroAt(x, tolerance));               \
+  EXPECT_THAT(manifold, ::ceres::MinusPlusIsIdentityAt(x, delta, tolerance));  \
+  EXPECT_THAT(manifold,                                                        \
+              ::ceres::MinusPlusIsIdentityAt(x, zero_tangent, tolerance));     \
+  EXPECT_THAT(manifold, ::ceres::PlusMinusIsIdentityAt(x, x, tolerance));      \
+  EXPECT_THAT(manifold, ::ceres::PlusMinusIsIdentityAt(x, y, tolerance));      \
+  EXPECT_THAT(manifold, ::ceres::HasCorrectPlusJacobianAt(x, tolerance));      \
+  EXPECT_THAT(manifold, ::ceres::HasCorrectMinusJacobianAt(x, tolerance));     \
   EXPECT_THAT(manifold, ::ceres::MinusPlusJacobianIsIdentityAt(x, tolerance)); \
-  EXPECT_THAT(manifold, ::ceres::HasCorrectRightMultiplyByPlusJacobianAt(x, tolerance));
+  EXPECT_THAT(manifold,                                                        \
+              ::ceres::HasCorrectRightMultiplyByPlusJacobianAt(x, tolerance));
 
 // Checks that the invariant Plus(x, 0) == x holds.
 MATCHER_P2(XPlusZeroIsXAt, x, tolerance, "") {
diff --git a/include/ceres/problem.h b/include/ceres/problem.h
index fac57c9..7b157d1 100644
--- a/include/ceres/problem.h
+++ b/include/ceres/problem.h
@@ -538,7 +538,7 @@
       double** jacobians) const;
 
   // Returns reference to the options with which the Problem was constructed.
-  const Options & options() const;
+  const Options& options() const;
 
   // Returns pointer to Problem implementation
   internal::ProblemImpl* mutable_impl();
diff --git a/internal/ceres/block_jacobi_preconditioner.h b/internal/ceres/block_jacobi_preconditioner.h
index a39ad12..533b764 100644
--- a/internal/ceres/block_jacobi_preconditioner.h
+++ b/internal/ceres/block_jacobi_preconditioner.h
@@ -48,6 +48,10 @@
 
 // This version of the preconditioner is for use with BlockSparseMatrix
 // Jacobians.
+//
+// TODO(https://github.com/ceres-solver/ceres-solver/issues/936):
+// BlockSparseJacobiPreconditioner::RightMultiply will benefit from
+// multithreading
 class CERES_NO_EXPORT BlockSparseJacobiPreconditioner
     : public BlockSparseMatrixPreconditioner {
  public:
diff --git a/internal/ceres/block_sparse_matrix.cc b/internal/ceres/block_sparse_matrix.cc
index c81ea39..f49f229 100644
--- a/internal/ceres/block_sparse_matrix.cc
+++ b/internal/ceres/block_sparse_matrix.cc
@@ -153,7 +153,8 @@
               });
 }
 
-// TODO: This method might benefit from caching column-block partition
+// TODO(https://github.com/ceres-solver/ceres-solver/issues/933): This method
+// might benefit from caching column-block partition
 void BlockSparseMatrix::LeftMultiplyAndAccumulate(const double* x,
                                                   double* y,
                                                   ContextImpl* context,
@@ -244,7 +245,8 @@
   }
 }
 
-// TODO: This method might benefit from caching column-block partition
+// TODO(https://github.com/ceres-solver/ceres-solver/issues/933): This method
+// might benefit from caching column-block partition
 void BlockSparseMatrix::SquaredColumnNorm(double* x,
                                           ContextImpl* context,
                                           int num_threads) const {
@@ -295,7 +297,8 @@
   }
 }
 
-// TODO: This method might benefit from caching column-block partition
+// TODO(https://github.com/ceres-solver/ceres-solver/issues/933): This method
+// might benefit from caching column-block partition
 void BlockSparseMatrix::ScaleColumns(const double* scale,
                                      ContextImpl* context,
                                      int num_threads) {
diff --git a/internal/ceres/block_sparse_matrix_test.cc b/internal/ceres/block_sparse_matrix_test.cc
index 0529a2d..5036aa2 100644
--- a/internal/ceres/block_sparse_matrix_test.cc
+++ b/internal/ceres/block_sparse_matrix_test.cc
@@ -126,15 +126,15 @@
     std::unique_ptr<LinearLeastSquaresProblem> problem =
         CreateLinearLeastSquaresProblemFromId(2);
     CHECK(problem != nullptr);
-    A_.reset(down_cast<BlockSparseMatrix*>(problem->A.release()));
+    a_.reset(down_cast<BlockSparseMatrix*>(problem->A.release()));
 
     problem = CreateLinearLeastSquaresProblemFromId(1);
     CHECK(problem != nullptr);
-    B_.reset(down_cast<TripletSparseMatrix*>(problem->A.release()));
+    b_.reset(down_cast<TripletSparseMatrix*>(problem->A.release()));
 
-    CHECK_EQ(A_->num_rows(), B_->num_rows());
-    CHECK_EQ(A_->num_cols(), B_->num_cols());
-    CHECK_EQ(A_->num_nonzeros(), B_->num_nonzeros());
+    CHECK_EQ(a_->num_rows(), b_->num_rows());
+    CHECK_EQ(a_->num_cols(), b_->num_cols());
+    CHECK_EQ(a_->num_nonzeros(), b_->num_nonzeros());
     context_.EnsureMinimumThreads(kNumThreads);
 
     BlockSparseMatrix::RandomMatrixOptions options;
@@ -147,67 +147,67 @@
     options.block_density = 0.05;
 
     std::mt19937 rng;
-    C_ = BlockSparseMatrix::CreateRandomMatrix(options, rng);
+    c_ = BlockSparseMatrix::CreateRandomMatrix(options, rng);
   }
 
-  std::unique_ptr<BlockSparseMatrix> A_;
-  std::unique_ptr<TripletSparseMatrix> B_;
-  std::unique_ptr<BlockSparseMatrix> C_;
+  std::unique_ptr<BlockSparseMatrix> a_;
+  std::unique_ptr<TripletSparseMatrix> b_;
+  std::unique_ptr<BlockSparseMatrix> c_;
   ContextImpl context_;
 };
 
 TEST_F(BlockSparseMatrixTest, SetZeroTest) {
-  A_->SetZero();
-  EXPECT_EQ(13, A_->num_nonzeros());
+  a_->SetZero();
+  EXPECT_EQ(13, a_->num_nonzeros());
 }
 
 TEST_F(BlockSparseMatrixTest, RightMultiplyAndAccumulateTest) {
-  Vector y_a = Vector::Zero(A_->num_rows());
-  Vector y_b = Vector::Zero(A_->num_rows());
-  for (int i = 0; i < A_->num_cols(); ++i) {
-    Vector x = Vector::Zero(A_->num_cols());
+  Vector y_a = Vector::Zero(a_->num_rows());
+  Vector y_b = Vector::Zero(a_->num_rows());
+  for (int i = 0; i < a_->num_cols(); ++i) {
+    Vector x = Vector::Zero(a_->num_cols());
     x[i] = 1.0;
-    A_->RightMultiplyAndAccumulate(x.data(), y_a.data());
-    B_->RightMultiplyAndAccumulate(x.data(), y_b.data());
+    a_->RightMultiplyAndAccumulate(x.data(), y_a.data());
+    b_->RightMultiplyAndAccumulate(x.data(), y_b.data());
     EXPECT_LT((y_a - y_b).norm(), 1e-12);
   }
 }
 
 TEST_F(BlockSparseMatrixTest, RightMultiplyAndAccumulateParallelTest) {
-  Vector y_0 = Vector::Random(A_->num_rows());
+  Vector y_0 = Vector::Random(a_->num_rows());
   Vector y_s = y_0;
   Vector y_p = y_0;
 
-  Vector x = Vector::Random(A_->num_cols());
-  A_->RightMultiplyAndAccumulate(x.data(), y_s.data());
+  Vector x = Vector::Random(a_->num_cols());
+  a_->RightMultiplyAndAccumulate(x.data(), y_s.data());
 
-  A_->RightMultiplyAndAccumulate(x.data(), y_p.data(), &context_, kNumThreads);
+  a_->RightMultiplyAndAccumulate(x.data(), y_p.data(), &context_, kNumThreads);
 
   // Current parallel implementation is expected to be bit-exact
   EXPECT_EQ((y_s - y_p).norm(), 0.);
 }
 
 TEST_F(BlockSparseMatrixTest, LeftMultiplyAndAccumulateTest) {
-  Vector y_a = Vector::Zero(A_->num_cols());
-  Vector y_b = Vector::Zero(A_->num_cols());
-  for (int i = 0; i < A_->num_rows(); ++i) {
-    Vector x = Vector::Zero(A_->num_rows());
+  Vector y_a = Vector::Zero(a_->num_cols());
+  Vector y_b = Vector::Zero(a_->num_cols());
+  for (int i = 0; i < a_->num_rows(); ++i) {
+    Vector x = Vector::Zero(a_->num_rows());
     x[i] = 1.0;
-    A_->LeftMultiplyAndAccumulate(x.data(), y_a.data());
-    B_->LeftMultiplyAndAccumulate(x.data(), y_b.data());
+    a_->LeftMultiplyAndAccumulate(x.data(), y_a.data());
+    b_->LeftMultiplyAndAccumulate(x.data(), y_b.data());
     EXPECT_LT((y_a - y_b).norm(), 1e-12);
   }
 }
 
 TEST_F(BlockSparseMatrixTest, LeftMultiplyAndAccumulateParallelTest) {
-  Vector y_0 = Vector::Random(A_->num_cols());
+  Vector y_0 = Vector::Random(a_->num_cols());
   Vector y_s = y_0;
   Vector y_p = y_0;
 
-  Vector x = Vector::Random(A_->num_rows());
-  A_->LeftMultiplyAndAccumulate(x.data(), y_s.data());
+  Vector x = Vector::Random(a_->num_rows());
+  a_->LeftMultiplyAndAccumulate(x.data(), y_s.data());
 
-  A_->LeftMultiplyAndAccumulate(x.data(), y_p.data(), &context_, kNumThreads);
+  a_->LeftMultiplyAndAccumulate(x.data(), y_p.data(), &context_, kNumThreads);
 
   // Parallel implementation for left products uses a different order of
   // traversal, thus results might be different
@@ -215,49 +215,49 @@
 }
 
 TEST_F(BlockSparseMatrixTest, SquaredColumnNormTest) {
-  Vector y_a = Vector::Zero(A_->num_cols());
-  Vector y_b = Vector::Zero(A_->num_cols());
-  A_->SquaredColumnNorm(y_a.data());
-  B_->SquaredColumnNorm(y_b.data());
+  Vector y_a = Vector::Zero(a_->num_cols());
+  Vector y_b = Vector::Zero(a_->num_cols());
+  a_->SquaredColumnNorm(y_a.data());
+  b_->SquaredColumnNorm(y_b.data());
   EXPECT_LT((y_a - y_b).norm(), 1e-12);
 }
 
 TEST_F(BlockSparseMatrixTest, SquaredColumnNormParallelTest) {
-  Vector y_a = Vector::Zero(C_->num_cols());
-  Vector y_b = Vector::Zero(C_->num_cols());
-  C_->SquaredColumnNorm(y_a.data());
+  Vector y_a = Vector::Zero(c_->num_cols());
+  Vector y_b = Vector::Zero(c_->num_cols());
+  c_->SquaredColumnNorm(y_a.data());
 
-  C_->SquaredColumnNorm(y_b.data(), &context_, kNumThreads);
+  c_->SquaredColumnNorm(y_b.data(), &context_, kNumThreads);
   EXPECT_LT((y_a - y_b).norm(), 1e-12);
 }
 
 TEST_F(BlockSparseMatrixTest, ScaleColumnsTest) {
-  const Vector scale = Vector::Random(C_->num_cols()).cwiseAbs();
+  const Vector scale = Vector::Random(c_->num_cols()).cwiseAbs();
 
-  const Vector x = Vector::Random(C_->num_rows());
-  Vector y_expected = Vector::Zero(C_->num_cols());
-  C_->LeftMultiplyAndAccumulate(x.data(), y_expected.data());
+  const Vector x = Vector::Random(c_->num_rows());
+  Vector y_expected = Vector::Zero(c_->num_cols());
+  c_->LeftMultiplyAndAccumulate(x.data(), y_expected.data());
   y_expected.array() *= scale.array();
 
-  C_->ScaleColumns(scale.data());
-  Vector y_observed = Vector::Zero(C_->num_cols());
-  C_->LeftMultiplyAndAccumulate(x.data(), y_observed.data());
+  c_->ScaleColumns(scale.data());
+  Vector y_observed = Vector::Zero(c_->num_cols());
+  c_->LeftMultiplyAndAccumulate(x.data(), y_observed.data());
 
   EXPECT_GT(y_expected.norm(), 1.);
   EXPECT_LT((y_observed - y_expected).norm(), 1e-12 * y_expected.norm());
 }
 
 TEST_F(BlockSparseMatrixTest, ScaleColumnsParallelTest) {
-  const Vector scale = Vector::Random(C_->num_cols()).cwiseAbs();
+  const Vector scale = Vector::Random(c_->num_cols()).cwiseAbs();
 
-  const Vector x = Vector::Random(C_->num_rows());
-  Vector y_expected = Vector::Zero(C_->num_cols());
-  C_->LeftMultiplyAndAccumulate(x.data(), y_expected.data());
+  const Vector x = Vector::Random(c_->num_rows());
+  Vector y_expected = Vector::Zero(c_->num_cols());
+  c_->LeftMultiplyAndAccumulate(x.data(), y_expected.data());
   y_expected.array() *= scale.array();
 
-  C_->ScaleColumns(scale.data(), &context_, kNumThreads);
-  Vector y_observed = Vector::Zero(C_->num_cols());
-  C_->LeftMultiplyAndAccumulate(x.data(), y_observed.data());
+  c_->ScaleColumns(scale.data(), &context_, kNumThreads);
+  Vector y_observed = Vector::Zero(c_->num_cols());
+  c_->LeftMultiplyAndAccumulate(x.data(), y_observed.data());
 
   EXPECT_GT(y_expected.norm(), 1.);
   EXPECT_LT((y_observed - y_expected).norm(), 1e-12 * y_expected.norm());
@@ -266,8 +266,8 @@
 TEST_F(BlockSparseMatrixTest, ToDenseMatrixTest) {
   Matrix m_a;
   Matrix m_b;
-  A_->ToDenseMatrix(&m_a);
-  B_->ToDenseMatrix(&m_b);
+  a_->ToDenseMatrix(&m_a);
+  b_->ToDenseMatrix(&m_b);
   EXPECT_LT((m_a - m_b).norm(), 1e-12);
 }
 
@@ -276,25 +276,25 @@
       CreateLinearLeastSquaresProblemFromId(2);
   std::unique_ptr<BlockSparseMatrix> m(
       down_cast<BlockSparseMatrix*>(problem->A.release()));
-  A_->AppendRows(*m);
-  EXPECT_EQ(A_->num_rows(), 2 * m->num_rows());
-  EXPECT_EQ(A_->num_cols(), m->num_cols());
+  a_->AppendRows(*m);
+  EXPECT_EQ(a_->num_rows(), 2 * m->num_rows());
+  EXPECT_EQ(a_->num_cols(), m->num_cols());
 
   problem = CreateLinearLeastSquaresProblemFromId(1);
   std::unique_ptr<TripletSparseMatrix> m2(
       down_cast<TripletSparseMatrix*>(problem->A.release()));
-  B_->AppendRows(*m2);
+  b_->AppendRows(*m2);
 
-  Vector y_a = Vector::Zero(A_->num_rows());
-  Vector y_b = Vector::Zero(A_->num_rows());
-  for (int i = 0; i < A_->num_cols(); ++i) {
-    Vector x = Vector::Zero(A_->num_cols());
+  Vector y_a = Vector::Zero(a_->num_rows());
+  Vector y_b = Vector::Zero(a_->num_rows());
+  for (int i = 0; i < a_->num_cols(); ++i) {
+    Vector x = Vector::Zero(a_->num_cols());
     x[i] = 1.0;
     y_a.setZero();
     y_b.setZero();
 
-    A_->RightMultiplyAndAccumulate(x.data(), y_a.data());
-    B_->RightMultiplyAndAccumulate(x.data(), y_b.data());
+    a_->RightMultiplyAndAccumulate(x.data(), y_a.data());
+    b_->RightMultiplyAndAccumulate(x.data(), y_b.data());
     EXPECT_LT((y_a - y_b).norm(), 1e-12);
   }
 }
@@ -304,7 +304,7 @@
   std::unique_ptr<BlockSparseMatrix> m(
       down_cast<BlockSparseMatrix*>(problem->A.release()));
 
-  auto block_structure = A_->block_structure();
+  auto block_structure = a_->block_structure();
 
   // Several AppendRows and DeleteRowBlocks operations are applied to matrix,
   // with regular and transpose block structures being compared after each
@@ -315,14 +315,14 @@
   const int num_row_blocks_to_delete[] = {0, -1, 1, -1, 8, -1, 10};
   for (auto& t : num_row_blocks_to_delete) {
     if (t == -1) {
-      A_->AppendRows(*m);
+      a_->AppendRows(*m);
     } else if (t > 0) {
       CHECK_GE(block_structure->rows.size(), t);
-      A_->DeleteRowBlocks(t);
+      a_->DeleteRowBlocks(t);
     }
 
-    auto block_structure = A_->block_structure();
-    auto transpose_block_structure = A_->transpose_block_structure();
+    auto block_structure = a_->block_structure();
+    auto transpose_block_structure = a_->transpose_block_structure();
     ASSERT_NE(block_structure, nullptr);
     ASSERT_NE(transpose_block_structure, nullptr);
 
@@ -378,7 +378,7 @@
 }
 
 TEST_F(BlockSparseMatrixTest, AppendAndDeleteBlockDiagonalMatrix) {
-  const std::vector<Block>& column_blocks = A_->block_structure()->cols;
+  const std::vector<Block>& column_blocks = a_->block_structure()->cols;
   const int num_cols =
       column_blocks.back().size + column_blocks.back().position;
   Vector diagonal(num_cols);
@@ -388,39 +388,39 @@
   std::unique_ptr<BlockSparseMatrix> appendage(
       BlockSparseMatrix::CreateDiagonalMatrix(diagonal.data(), column_blocks));
 
-  A_->AppendRows(*appendage);
+  a_->AppendRows(*appendage);
   Vector y_a, y_b;
-  y_a.resize(A_->num_rows());
-  y_b.resize(A_->num_rows());
-  for (int i = 0; i < A_->num_cols(); ++i) {
-    Vector x = Vector::Zero(A_->num_cols());
+  y_a.resize(a_->num_rows());
+  y_b.resize(a_->num_rows());
+  for (int i = 0; i < a_->num_cols(); ++i) {
+    Vector x = Vector::Zero(a_->num_cols());
     x[i] = 1.0;
     y_a.setZero();
     y_b.setZero();
 
-    A_->RightMultiplyAndAccumulate(x.data(), y_a.data());
-    B_->RightMultiplyAndAccumulate(x.data(), y_b.data());
-    EXPECT_LT((y_a.head(B_->num_rows()) - y_b.head(B_->num_rows())).norm(),
+    a_->RightMultiplyAndAccumulate(x.data(), y_a.data());
+    b_->RightMultiplyAndAccumulate(x.data(), y_b.data());
+    EXPECT_LT((y_a.head(b_->num_rows()) - y_b.head(b_->num_rows())).norm(),
               1e-12);
-    Vector expected_tail = Vector::Zero(A_->num_cols());
+    Vector expected_tail = Vector::Zero(a_->num_cols());
     expected_tail(i) = diagonal(i);
-    EXPECT_LT((y_a.tail(A_->num_cols()) - expected_tail).norm(), 1e-12);
+    EXPECT_LT((y_a.tail(a_->num_cols()) - expected_tail).norm(), 1e-12);
   }
 
-  A_->DeleteRowBlocks(column_blocks.size());
-  EXPECT_EQ(A_->num_rows(), B_->num_rows());
-  EXPECT_EQ(A_->num_cols(), B_->num_cols());
+  a_->DeleteRowBlocks(column_blocks.size());
+  EXPECT_EQ(a_->num_rows(), b_->num_rows());
+  EXPECT_EQ(a_->num_cols(), b_->num_cols());
 
-  y_a.resize(A_->num_rows());
-  y_b.resize(A_->num_rows());
-  for (int i = 0; i < A_->num_cols(); ++i) {
-    Vector x = Vector::Zero(A_->num_cols());
+  y_a.resize(a_->num_rows());
+  y_b.resize(a_->num_rows());
+  for (int i = 0; i < a_->num_cols(); ++i) {
+    Vector x = Vector::Zero(a_->num_cols());
     x[i] = 1.0;
     y_a.setZero();
     y_b.setZero();
 
-    A_->RightMultiplyAndAccumulate(x.data(), y_a.data());
-    B_->RightMultiplyAndAccumulate(x.data(), y_b.data());
+    a_->RightMultiplyAndAccumulate(x.data(), y_a.data());
+    b_->RightMultiplyAndAccumulate(x.data(), y_b.data());
     EXPECT_LT((y_a - y_b).norm(), 1e-12);
   }
 }
diff --git a/internal/ceres/cgnr_solver.cc b/internal/ceres/cgnr_solver.cc
index a1a3c6e..3f867dc 100644
--- a/internal/ceres/cgnr_solver.cc
+++ b/internal/ceres/cgnr_solver.cc
@@ -160,8 +160,6 @@
     preconditioner_options.context = options_.context;
 
     if (options_.preconditioner_type == JACOBI) {
-      // TODO: BlockSparseJacobiPreconditioner::RightMultiply will benefit from
-      // multithreading
       preconditioner_ = std::make_unique<BlockSparseJacobiPreconditioner>(
           preconditioner_options, *A);
     } else if (options_.preconditioner_type == SUBSET) {
diff --git a/internal/ceres/iterative_schur_complement_solver.cc b/internal/ceres/iterative_schur_complement_solver.cc
index 16b20a7..ace075e 100644
--- a/internal/ceres/iterative_schur_complement_solver.cc
+++ b/internal/ceres/iterative_schur_complement_solver.cc
@@ -186,17 +186,10 @@
     case SCHUR_POWER_SERIES_EXPANSION:
       // Ignoring the value of spse_tolerance to ensure preconditioner stays
       // fixed during the iterations of cg.
-      // TODO: In PowerSeriesExpansionPreconditioner::RightMultiplyAndAccumulate
-      // only operations performed via ImplicitSchurComplement are threaded.
-      // PowerSeriesExpansionPreconditioner will benefit from multithreading
-      // applied to remaning operations (block-sparse right product and several
-      // vector operations)
       preconditioner_ = std::make_unique<PowerSeriesExpansionPreconditioner>(
           schur_complement_.get(), options_.max_num_spse_iterations, 0);
       break;
     case SCHUR_JACOBI:
-      // TODO: SchurJacobiPreconditioner::RightMultiply will benefit from
-      // multithreading
       preconditioner_ = std::make_unique<SchurJacobiPreconditioner>(
           *A->block_structure(), preconditioner_options);
       break;
diff --git a/internal/ceres/parallel_for_cxx.cc b/internal/ceres/parallel_for_cxx.cc
index 3f47436..44b0be5 100644
--- a/internal/ceres/parallel_for_cxx.cc
+++ b/internal/ceres/parallel_for_cxx.cc
@@ -28,11 +28,13 @@
 //
 // Author: vitus@google.com (Michael Vitus)
 
+#include <algorithm>
 #include <atomic>
 #include <cmath>
 #include <condition_variable>
 #include <memory>
 #include <mutex>
+#include <tuple>
 
 #include "ceres/internal/config.h"
 #include "ceres/parallel_for.h"
diff --git a/internal/ceres/parallel_for_test.cc b/internal/ceres/parallel_for_test.cc
index 058b0a3..6c342aa 100644
--- a/internal/ceres/parallel_for_test.cc
+++ b/internal/ceres/parallel_for_test.cc
@@ -36,6 +36,7 @@
 #include <numeric>
 #include <random>
 #include <thread>
+#include <tuple>
 #include <vector>
 
 #include "ceres/context_impl.h"
diff --git a/internal/ceres/parallel_vector_operations_benchmark.cc b/internal/ceres/parallel_vector_operations_benchmark.cc
index 7180b54..31a5c7a 100644
--- a/internal/ceres/parallel_vector_operations_benchmark.cc
+++ b/internal/ceres/parallel_vector_operations_benchmark.cc
@@ -25,6 +25,9 @@
 // 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.
+
+#include <algorithm>
+
 #include "benchmark/benchmark.h"
 #include "ceres/eigen_vector_ops.h"
 #include "ceres/parallel_for.h"
diff --git a/internal/ceres/power_series_expansion_preconditioner.h b/internal/ceres/power_series_expansion_preconditioner.h
index a8bb9a6..20be767 100644
--- a/internal/ceres/power_series_expansion_preconditioner.h
+++ b/internal/ceres/power_series_expansion_preconditioner.h
@@ -41,7 +41,14 @@
 // This is a preconditioner via power series expansion of Schur
 // complement inverse based on "Weber et al, Power Bundle Adjustment for
 // Large-Scale 3D Reconstruction".
-
+//
+//
+// TODO(https://github.com/ceres-solver/ceres-solver/issues/934): In
+// PowerSeriesExpansionPreconditioner::RightMultiplyAndAccumulate only
+// operations performed via ImplicitSchurComplement are threaded.
+// PowerSeriesExpansionPreconditioner will benefit from multithreading
+// applied to remaning operations (block-sparse right product and several
+// vector operations)
 class CERES_NO_EXPORT PowerSeriesExpansionPreconditioner
     : public Preconditioner {
  public:
diff --git a/internal/ceres/problem.cc b/internal/ceres/problem.cc
index e51fdd1..e8589be 100644
--- a/internal/ceres/problem.cc
+++ b/internal/ceres/problem.cc
@@ -212,7 +212,7 @@
   impl_->GetResidualBlocksForParameterBlock(values, residual_blocks);
 }
 
-const Problem::Options & Problem::options() const { return impl_->options(); }
+const Problem::Options& Problem::options() const { return impl_->options(); }
 
 internal::ProblemImpl* Problem::mutable_impl() { return impl_.get(); }
 
diff --git a/internal/ceres/problem_impl.h b/internal/ceres/problem_impl.h
index 22b37b6..b51c48d 100644
--- a/internal/ceres/problem_impl.h
+++ b/internal/ceres/problem_impl.h
@@ -166,7 +166,7 @@
     return residual_block_set_;
   }
 
-  const Problem::Options & options() const { return options_; }
+  const Problem::Options& options() const { return options_; }
 
   ContextImpl* context() { return context_impl_; }
 
diff --git a/internal/ceres/schur_jacobi_preconditioner.h b/internal/ceres/schur_jacobi_preconditioner.h
index ddf471c..fbdeab4 100644
--- a/internal/ceres/schur_jacobi_preconditioner.h
+++ b/internal/ceres/schur_jacobi_preconditioner.h
@@ -73,6 +73,8 @@
 //   preconditioner.Update(A, nullptr);
 //   preconditioner.RightMultiplyAndAccumulate(x, y);
 //
+// TODO(https://github.com/ceres-solver/ceres-solver/issues/935):
+// SchurJacobiPreconditioner::RightMultiply will benefit from multithreading
 class CERES_NO_EXPORT SchurJacobiPreconditioner
     : public BlockSparseMatrixPreconditioner {
  public:
diff --git a/internal/ceres/sparse_matrix.cc b/internal/ceres/sparse_matrix.cc
index 584ac1d..77c28e7 100644
--- a/internal/ceres/sparse_matrix.cc
+++ b/internal/ceres/sparse_matrix.cc
@@ -42,12 +42,12 @@
   SquaredColumnNorm(x);
 }
 
-void SparseMatrix::ScaleColumns(const double* x,
+void SparseMatrix::ScaleColumns(const double* scale,
                                 ContextImpl* context,
                                 int num_threads) {
   (void)context;
   (void)num_threads;
-  ScaleColumns(x);
+  ScaleColumns(scale);
 }
 
 }  // namespace ceres::internal