updated parallel_for and taskflow.hpp

tsung-wei-huang · tsung-wei-huang · commit aa83f77040d8 · 2018-06-05T10:58:23.000-05:00
diff --git a/example/parallel_for.cpp b/example/parallel_for.cpp
@@ -1,5 +1,6 @@
 #include "taskflow.hpp"
 #include <cassert>
+#include <numeric>
 
 // Function: fib
 int fib(int n) {
@@ -23,13 +24,13 @@ void sequential(int N) {
 
 // Procedure: taskflow
 void taskflow(int N) {
+
+  std::vector<int> range(N);
+  std::iota(range.begin(), range.end(), 0);
+
   auto tbeg = std::chrono::steady_clock::now();
   tf::Taskflow tf;
-  std::vector<int> range(N);
-  for(int n=0; n<N; ++n) {
-    range[n] = n;
-  }
-  tf.parallel_for(range.begin(), range.end(), [&] (int& i) { 
+  tf.parallel_for(range, [&] (const int i) { 
     printf("fib[%d]=%d\n", i, fib(i));
   });
   tf.wait_for_all();
@@ -42,10 +43,14 @@ void taskflow(int N) {
 
 // Procedure: openmp
 void openmp(int N) {
+  
+  std::vector<int> range(N);
+  std::iota(range.begin(), range.end(), 0);
+
   auto tbeg = std::chrono::steady_clock::now();
   #pragma omp parallel for
-  for(int i=0; i<N; ++i) {
-    printf("fib[%d]=%d\n", i, fib(i));
+  for(size_t i=0; i<N; ++i) {
+    printf("fib[%d]=%d\n", range[i], fib(range[i]));
   }
   auto tend = std::chrono::steady_clock::now();
   std::cout << "openmp version takes " 
diff --git a/taskflow.hpp b/taskflow.hpp
@@ -438,10 +438,10 @@ class BasicTaskflow {
     auto wait_for_all();
 
     template <typename I, typename C>
-    auto parallel_for(I, I, C&&, ssize_t = 1);
+    auto parallel_for(I, I, C&&, size_t = 1);
 
     template <typename T, typename C, std::enable_if_t<is_iterable_v<T>, void>* = nullptr>
-    auto parallel_for(T&, C&&, ssize_t = 1);
+    auto parallel_for(T&, C&&, size_t = 1);
 
     //template<typename I, class C>
     //auto parallel_range(const I, const I, C&&, ssize_t = 1);
@@ -847,28 +847,25 @@ auto BasicTaskflow<F>::emplace(C&&... cs) {
 // Function: parallel_for    
 template <typename F>
 template <typename I, class C>
-auto BasicTaskflow<F>::parallel_for(I beg, I end, C&& c, ssize_t group) {
+auto BasicTaskflow<F>::parallel_for(I beg, I end, C&& c, size_t group) {
 
   if(group <= 0) {
     group = 1;
   }
 
   auto source = placeholder();
   auto target = placeholder();
-  auto len = std::distance(beg, end);
 
-  for(; beg != end;) {
+  for(; beg != end; ) {
     auto e = beg;
-    std::advance(e, group < len ? group : len);
-    len -= group;
-
-    auto task = silent_emplace([c, itr=beg, e]() mutable { 
-      for(;itr!=e; itr++){
-        c(*itr); 
+    for(size_t i=0; i<group && e != end; ++e, ++i);
+    auto task = silent_emplace([c, beg, e] () mutable {
+      for(auto itr = beg; itr != e; ++itr) {
+        c(*itr);
       }
     });
     beg = e;
-
+    
     source.precede(task);
     task.precede(target);
   }
@@ -879,7 +876,7 @@ auto BasicTaskflow<F>::parallel_for(I beg, I end, C&& c, ssize_t group) {
 // Function: parallel_for
 template <typename F>
 template <typename T, typename C, std::enable_if_t<is_iterable_v<T>, void>*>
-auto BasicTaskflow<F>::parallel_for(T& t, C&& c, ssize_t group) {
+auto BasicTaskflow<F>::parallel_for(T& t, C&& c, size_t group) {
   return parallel_for(t.begin(), t.end(), std::forward<C>(c), group);
 }
 
