Use Crystal::PointerLinkedList instead of Deque in Mutex (#15330)

Extracts the undocumented `Fiber::Waiting` struct from `WaitGroup` that acts as
the node in a linked list, replacing a `Deque` to store the waiting fibers.

The flat array doesn't have much impact on performance: we only reach the head
or the tail once to dequeue/dequeue one fiber at a time. This however spares a
number of GC allocations since the Deque has to be allocated plus its buffer
that will have to be reallocated sometimes (and will only ever grow, never shrink).

Author: ysbaddaden

src/fiber/pointer_linked_list_node.cr

@@ -0,0 +1,15 @@
+require "crystal/pointer_linked_list"
+
+class Fiber
+  # :nodoc:
+  struct PointerLinkedListNode
+    include Crystal::PointerLinkedList::Node
+
+    def initialize(@fiber : Fiber)
+    end
+
+    def enqueue : Nil
+      @fiber.enqueue
+    end
+  end
+end

src/mutex.cr

@@ -1,3 +1,4 @@
+require "fiber/pointer_linked_list_node"
 require "crystal/spin_lock"
 
 # A fiber-safe mutex.
@@ -22,7 +23,7 @@ class Mutex
   @state = Atomic(Int32).new(UNLOCKED)
   @mutex_fiber : Fiber?
   @lock_count = 0
-  @queue = Deque(Fiber).new
+  @queue = Crystal::PointerLinkedList(Fiber::PointerLinkedListNode).new
   @queue_count = Atomic(Int32).new(0)
   @lock = Crystal::SpinLock.new
 
@@ -59,6 +60,8 @@ class Mutex
     loop do
       break if try_lock
 
+      waiting = Fiber::PointerLinkedListNode.new(Fiber.current)
+
       @lock.sync do
         @queue_count.add(1)
 
@@ -71,7 +74,7 @@ class Mutex
           end
         end
 
-        @queue.push Fiber.current
+        @queue.push pointerof(waiting)
       end
 
       Fiber.suspend
@@ -116,17 +119,18 @@ class Mutex
       return
     end
 
-    fiber = nil
+    waiting = nil
     @lock.sync do
       if @queue_count.get == 0
         return
       end
 
-      if fiber = @queue.shift?
+      if waiting = @queue.shift?
         @queue_count.add(-1)
       end
     end
-    fiber.enqueue if fiber
+
+    waiting.try(&.value.enqueue)
   end
 
   def synchronize(&)

src/wait_group.cr

@@ -1,6 +1,6 @@
 require "fiber"
+require "fiber/pointer_linked_list_node"
 require "crystal/spin_lock"
-require "crystal/pointer_linked_list"
 
 # Suspend execution until a collection of fibers are finished.
 #
@@ -31,17 +31,6 @@ require "crystal/pointer_linked_list"
 # wg.wait
 # ```
 class WaitGroup
-  private struct Waiting
-    include Crystal::PointerLinkedList::Node
-
-    def initialize(@fiber : Fiber)
-    end
-
-    def enqueue : Nil
-      @fiber.enqueue
-    end
-  end
-
   # Yields a `WaitGroup` instance and waits at the end of the block for all of
   # the work enqueued inside it to complete.
   #
@@ -59,7 +48,7 @@ class WaitGroup
   end
 
   def initialize(n : Int32 = 0)
-    @waiting = Crystal::PointerLinkedList(Waiting).new
+    @waiting = Crystal::PointerLinkedList(Fiber::PointerLinkedListNode).new
     @lock = Crystal::SpinLock.new
     @counter = Atomic(Int32).new(n)
   end
@@ -128,7 +117,7 @@ class WaitGroup
   def wait : Nil
     return if done?
 
-    waiting = Waiting.new(Fiber.current)
+    waiting = Fiber::PointerLinkedListNode.new(Fiber.current)
 
     @lock.sync do
       # must check again to avoid a race condition where #done may have