remove annotations

eth-cscs · Sep 29, 2023 · a36baa8 · a36baa8
1 parent e8d3fb5
commit a36baa8
Show file tree

Hide file tree

Showing 2 changed files with 0 additions and 18 deletions.
diff --git a/include/dlaf/eigensolver/tridiag_solver/merge.h b/include/dlaf/eigensolver/tridiag_solver/merge.h
@@ -1200,8 +1200,6 @@ void solveRank1ProblemDist(CommSender&& row_comm, CommSender&& col_comm, const S
                                                         auto& barrier_ptr) {
                  using dlaf::comm::internal::transformMPI;
 
-                 pika::scoped_annotation _("solveRank1ProblemDist");
-
                  common::Pipeline<comm::Communicator> row_comm_chain(row_comm_wrapper.get());
                  const dlaf::comm::Communicator& col_comm = col_comm_wrapper.get();
 
@@ -1220,7 +1218,6 @@ void solveRank1ProblemDist(CommSender&& row_comm, CommSender&& col_comm, const S
                  // to be dropped soon.
                  // Note: use last threads that in principle should have less work to do
                  if (k < n && thread_idx == nthreads - 1) {
-                   pika::scoped_annotation _("fill1_and_xevals");
                    const T* eval_initial_ptr = d_tiles_futs[0].get().ptr();
                    T* eval_ptr = eval_tiles[0].ptr();
 
@@ -1261,7 +1258,6 @@ void solveRank1ProblemDist(CommSender&& row_comm, CommSender&& col_comm, const S
 
                  // STEP 0b: Initialize workspaces (single-thread)
                  if (thread_idx == 0) {
-                   pika::scoped_annotation _("init_workspaces");
                    // Note:
                    // - nthreads are used for both LAED4 and weight calculation (one per worker thread)
                    // - last one is used for reducing weights from all workers
@@ -1291,7 +1287,6 @@ void solveRank1ProblemDist(CommSender&& row_comm, CommSender&& col_comm, const S
 
                  // STEP 1: LAED4 (multi-thread)
                  {
-                   pika::scoped_annotation _("laed4");
                    common::internal::SingleThreadedBlasScope single;  // TODO needed also for laed?
 
                    T* eval_ptr = eval_tiles[0].ptr();
@@ -1367,7 +1362,6 @@ void solveRank1ProblemDist(CommSender&& row_comm, CommSender&& col_comm, const S
                  // permutations
                  // during the switch from col axis to row axis we must keep the matching between eigenvectors
                  if (thread_idx == 0) {
-                   pika::scoped_annotation _("copy-diag");
                    for (SizeType ieg_el_lc = 0; ieg_el_lc < m_el_lc; ++ieg_el_lc) {
                      const SizeType ieg_el =
                          dist_sub.globalElementFromLocalElement<Coord::Row>(ieg_el_lc);
@@ -1403,8 +1397,6 @@ void solveRank1ProblemDist(CommSender&& row_comm, CommSender&& col_comm, const S
 
                  // STEP 2b: compute weights
                  {
-                   pika::scoped_annotation _("compute-weights");
-
                    for (SizeType jeg_el_lc = begin; jeg_el_lc < end; ++jeg_el_lc) {
                      const SizeType jeg_el =
                          dist_sub.globalElementFromLocalElement<Coord::Col>(jeg_el_lc);
@@ -1449,8 +1441,6 @@ void solveRank1ProblemDist(CommSender&& row_comm, CommSender&& col_comm, const S
 
                  // STEP 2c: reduce, then finalize computation with sign and square root (single-thread)
                  if (thread_idx == 0) {
-                   pika::scoped_annotation _("reduce-weights");
-
                    // local reduction from all bulk workers
                    for (SizeType i = 0; i < m_el_lc; ++i) {
                      for (std::size_t tidx = 1; tidx < nthreads; ++tidx) {
@@ -1481,7 +1471,6 @@ void solveRank1ProblemDist(CommSender&& row_comm, CommSender&& col_comm, const S
 
                  // STEP 3a: Form evecs using weights vector and compute (local) sum of squares
                  {
-                   pika::scoped_annotation _("form-evecs");
                    common::internal::SingleThreadedBlasScope single;
 
                    const T* w = ws_cols[nthreads]();
@@ -1526,7 +1515,6 @@ void solveRank1ProblemDist(CommSender&& row_comm, CommSender&& col_comm, const S
 
                  // STEP 3b: Reduce to get the sum of all squares on all ranks
                  if (thread_idx == 0) {
-                   pika::scoped_annotation _("reduce-squares-norm");
                    // TODO it can be limited to k_lc
                    tt::sync_wait(ex::just(std::cref(col_comm), MPI_SUM,
                                           common::make_data(ws_row(), n_el_lc)) |
@@ -1537,7 +1525,6 @@ void solveRank1ProblemDist(CommSender&& row_comm, CommSender&& col_comm, const S
 
                  // STEP 3c: Normalize (compute norm of each column and scale column vector)
                  {
-                   pika::scoped_annotation _("normalize-evecs");
                    common::internal::SingleThreadedBlasScope single;
 
                    const T* sum_squares = ws_row();
@@ -1576,8 +1563,6 @@ void mergeDistSubproblems(comm::CommunicatorGrid grid,
                           DistWorkSpaceHostMirror<T, D>& ws_hm) {
   namespace ex = pika::execution::experimental;
 
-  pika::scoped_annotation _("mergeDistSubproblems");
-
   const matrix::Distribution& dist_evecs = ws.e0.distribution();
 
   // Calculate the size of the upper subproblem

diff --git a/include/dlaf/permutations/general/impl.h b/include/dlaf/permutations/general/impl.h
@@ -15,7 +15,6 @@
 #include <mpi.h>
 
 #include <pika/algorithm.hpp>
-#include <pika/threading_base/scoped_annotation.hpp>
 
 #include <dlaf/blas/tile.h>
 #include <dlaf/common/assert.h>
@@ -216,8 +215,6 @@ void permuteJustLocal(const SizeType i_begin, const SizeType i_end, Matrix<const
                       Matrix<const T, D>& mat_in, Matrix<T, D>& mat_out) {
   static_assert(C == Coord::Col, "Just column permutation");
 
-  pika::scoped_annotation _("permuteJustLocal");
-
   namespace ut = matrix::util;
   namespace ex = pika::execution::experimental;
   namespace di = dlaf::internal;