Nat Goodspeed:

- Change some doc references from 'algorithm' to 'sched_algorithm'.
- Initial cut at supporting arbitrary user-coded scheduler properties.
- Set fiber_properties::sched_algo_ every time through awakened().
- Define sched_algorithm methods on fiber_base*, not worker_fiber*.
- Simplify detail::fifo by making tail_ point to last link pointer.
- Reimplement waiting_queue::push() using pointer-to-pointer trick.
- Reimplement waiting_queue::move_to() using fiber_base** scan.
- Make bounded_queue::tail_ a ptr* to simplify appending new nodes.
- Make unbounded_queue::tail_ a ptr* to simplify linking new nodes.
- Remove thread_affinity flag and access methods.
- Re-add thread_affinity specific to workstealing_round_robin.
- Remove 'priority' for every fiber, and its support methods.

examples/cpp03/migration/workstealing_round_robin.cpp

@@ -12,47 +12,71 @@
 #  include BOOST_ABI_PREFIX
 #endif
 
+workstealing_round_robin::workstealing_round_robin():
+    rhead_(0),
+    rtail_(&rhead_)
+{}
+
 void
-workstealing_round_robin::awakened( boost::fibers::detail::worker_fiber * f)
+workstealing_round_robin::awakened( boost::fibers::fiber_base * f)
 {
+    // forward this call to base-class method
+    boost::fibers::sched_algorithm_with_properties<affinity>::awakened(f);
+
     boost::mutex::scoped_lock lk( mtx_);
-    rqueue_.push_back( f);
+    // append this fiber_base* to ready queue
+    BOOST_ASSERT(! f->nxt_);
+    *rtail_ = f;
+    rtail_ = &f->nxt_;
 }
 
-boost::fibers::detail::worker_fiber *
+boost::fibers::fiber_base *
 workstealing_round_robin::pick_next()
 {
     boost::mutex::scoped_lock lk( mtx_);
-    boost::fibers::detail::worker_fiber * f = 0;
-    if ( ! rqueue_.empty() )
+    boost::fibers::fiber_base * f = 0;
+    if ( rhead_ )
     {
-        f = rqueue_.front();
-        rqueue_.pop_front();
+        f = rhead_;
+        // pop head item from ready queue
+        rhead_ = rhead_->nxt_;
+        f->nxt_ = 0;
+        // if that was the last item, reset tail_
+        if (! rhead_)
+            rtail_ = &rhead_;
     }
     return f;
 }
 
-void
-workstealing_round_robin::priority( boost::fibers::detail::worker_fiber * f, int prio) BOOST_NOEXCEPT
-{
-    BOOST_ASSERT( f);
-
-    // set only priority to fiber
-    // round-robin does not respect priorities
-    f->priority( prio);
-}
-
 boost::fibers::fiber
 workstealing_round_robin::steal() BOOST_NOEXCEPT
 {
     boost::mutex::scoped_lock lk( mtx_);
-    boost::fibers::detail::worker_fiber * f = 0;
-    if ( ! rqueue_.empty() )
+
+    // Search the queue for the LAST fiber_base that's willing to migrate,
+    // in other words (! thread_affinity).
+    boost::fibers::fiber_base ** fp = &rhead_, ** found = 0;
+    for ( ; *fp; fp = &(*fp)->nxt_)
     {
-        f = rqueue_.back();
-        rqueue_.pop_back();
+        // do not consider any fiber whose thread_affinity is set
+        if (! properties(*fp).thread_affinity)
+            found = fp;
     }
-    return boost::fibers::fiber( f);
+    if (! found)
+    {
+        // either the queue is completely empty or all current entries have
+        // thread_affinity set
+        return boost::fibers::fiber(static_cast<boost::fibers::fiber_base*>(0));
+    }
+    // We found at least one fiber_base whose thread_affinity is NOT set;
+    // *found points to the last of these. Unlink and return it.
+    boost::fibers::fiber_base* ret = *found;
+    *found = ret->nxt_;
+    ret->nxt_ = 0;
+    // if that was the last item, reset tail_
+    if (! *found)
+        rtail_ = &rhead_;
+    return boost::fibers::fiber(ret);
 }
 
 #ifdef BOOST_HAS_ABI_HEADERS