* benchmark/bm_vm4_pipe.rb: Add two new benchmark for GVL

performance. They was written by Koichi Sasada.
* benchmark/bm_vm4_thread_pass.rb: ditto.



git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@31379 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

ChangeLog

@@ -1,3 +1,9 @@
+Fri Apr 29 10:43:09 2011  KOSAKI Motohiro  <kosaki.motohiro@gmail.com>
+
+	* benchmark/bm_vm4_pipe.rb: Add two new benchmark for GVL
+	  performance. They was written by Koichi Sasada.
+	* benchmark/bm_vm4_thread_pass.rb: ditto.
+
 Fri Apr 29 10:25:31 2011  KOSAKI Motohiro  <kosaki.motohiro@gmail.com>
 
 	* vm_method.c (rb_clear_cache_by_class): Revert r29673. It made

benchmark/bm_vm4_pipe.rb

@@ -0,0 +1,17 @@
+# Mesure small and plenty pipe read/write.
+# A performance may depend on GVL implementation.
+
+lmax = 1_000_000
+r, w = IO.pipe
+[Thread.new{
+  lmax.times{
+    w.write('a')
+  }
+  p "w:exit"
+}, Thread.new{
+  lmax.times{
+    r.read(1)
+  }
+  p "r:exit"
+}].each{|t| t.join}
+

benchmark/bm_vm4_thread_pass.rb

@@ -0,0 +1,15 @@
+# Plenty Thtread.pass
+# A performance may depend on GVL implementation.
+
+tmax = (ARGV.shift || 2).to_i
+lmax = 2_000_000 / tmax
+
+(1..tmax).map{
+  Thread.new{
+    lmax.times{
+      Thread.pass
+    }
+  }
+}.each{|t| t.join}
+
+