Merge branch 'master' of github.com:patflick/mxx

patflick · Apr 25, 2016 · f96d491 · f96d491
2 parents 09a2ca0 + f777498
commit f96d491
Show file tree

Hide file tree

Showing 5 changed files with 10 additions and 8 deletions.
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -7,7 +7,7 @@ project(mxx)
 
 set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -Wuninitialized --std=c++11")
 set(CMAKE_CXX_FLAGS_DEBUG "${CMAKE_CXX_FLAGS_DEBUG} -g -O0")
-set(CMAKE_CXX_FLAGS_RELEASE "-O3 -DNDEBUG -msse3 -march=native -funroll-loops")
+set(CMAKE_CXX_FLAGS_RELEASE "-O3 -DNDEBUG -march=native -funroll-loops")
 set(CMAKE_CXX_FLAGS_RELWITHDEBINFO "${CMAKE_CXX_FLAGS_RELEASE} -g")
 
 # Add these standard paths to the search paths for FIND_LIBRARY

diff --git a/gtest/mxx_gtest_main.cpp b/gtest/mxx_gtest_main.cpp
@@ -34,6 +34,7 @@ int main(int argc, char* argv[]) {
  ::testing::InitGoogleTest(&argc, argv);
 
  mxx::env e(argc, argv);
+ mxx::env::set_exception_on_error();
  mxx::comm c = mxx::comm();
  //MPI_Init(&argc, &argv);
 

diff --git a/include/mxx/algos.hpp b/include/mxx/algos.hpp
@@ -228,7 +228,7 @@ std::vector<size_t> bucketing(std::vector<T>& input, Func key_func, size_t num_b
 
  // [1st pass]: counting the number of elements per bucket
  for (auto it = input.begin(); it != input.end(); ++it) {
- MXX_ASSERT(0 <= key_func(*it) && key_func(*it) < num_buckets);
+ MXX_ASSERT(0 <= key_func(*it) && (size_t)key_func(*it) < num_buckets);
  bucket_counts[key_func(*it)]++;
  }
 
@@ -278,7 +278,7 @@ std::vector<size_t> bucketing_inplace(std::vector<T>& input, Func key_func, size
 
  // [1st pass]: counting the number of elements per bucket
  for (auto it = input.begin(); it != input.end(); ++it) {
- MXX_ASSERT(0 <= key_func(*it) && key_func(*it) < num_buckets);
+ MXX_ASSERT(0 <= key_func(*it) && (size_t)key_func(*it) < num_buckets);
  bucket_counts[key_func(*it)]++;
  }
  // get exclusive prefix sum

diff --git a/include/mxx/benchmark.hpp b/include/mxx/benchmark.hpp
@@ -92,7 +92,7 @@ class hybrid_comm {
  func(*this);
  }
  } else {
- hybrid_comm hc(std::move(split_by_node(participate)));
+ hybrid_comm hc(split_by_node(participate));
  if (participate) {
  func(hc);
  }

diff --git a/include/mxx/timer.hpp b/include/mxx/timer.hpp
@@ -97,14 +97,15 @@ class section_timer_impl {
  time_rep elapsed = duration(stop-start).count();
  // TODO: reduce rather than allreduce
  // TODO: use single reduction with custom operator
- time_rep min = mxx::allreduce(elapsed, [](time_rep x, time_rep y){return std::min(x,y);}, comm);
- time_rep max = mxx::allreduce(elapsed, [](time_rep x, time_rep y){return std::max(x,y);}, comm);
- time_rep sum = mxx::allreduce(elapsed, std::plus<time_rep>(), comm);
+ double delapsed = elapsed;
+ double min = mxx::allreduce(delapsed, mxx::min<double>(), comm);
+ double max = mxx::allreduce(delapsed, mxx::max<double>(), comm);
+ double sum = mxx::allreduce(delapsed, std::plus<time_rep>(), comm);
  // calc mean:
  int p, rank;
  MPI_Comm_rank(comm, &rank);
  MPI_Comm_size(comm, &p);
- time_rep mean = sum / (double)p;
+ double mean = sum / (double)p;
  // only root process outputs the timings
  if (rank == root)
  ostr << std::setprecision(3) << std::scientific << "TIMER" << sep << min << sep << mean << sep << max << sep << depth << sep << name << std::endl;