Run mojo_benchmark on the perf bot.

mojo/tools/get_test_list.py

Index: mojo/tools/get_test_list.py
diff --git a/mojo/tools/get_test_list.py b/mojo/tools/get_test_list.py
index c59602c10244b174054ca4e0120d7a7db2ceaa19..9488636efc8cc887e104e0145943773b665e4693 100755
--- a/mojo/tools/get_test_list.py
+++ b/mojo/tools/get_test_list.py
@@ -212,6 +212,17 @@ def GetTestList(config, verbose_count=0):
 
       AddEntry(test_name, command)
 
+  # Benchmarks -----------------------------------------------------------------
+
+  if target_os == Config.OS_LINUX and ShouldRunTest(Config.TEST_TYPE_PERF):

[viettrungluu, 2015/10/01 15:22:06]

Seems kind of pointless to run them and not record results.

[ppi, 2015/10/01 15:25:04]

This is meant to catch breakages, e.g. benchmarks no longer producing results.

Before we record results we need to teach mojo_benchmark about the compatible
format, I was planning to do that separately.

[viettrungluu, 2015/10/01 19:38:39]

This suggests to me that you need actual tests for the benchmarks. Just running
this here will only catch relatively trivial errors. Moreover, this isn't run on
the trybots, etc.

+    command = ([os.path.join("mojo", "devtools", "common", "mojo_benchmark"),
+               os.path.join("mojo", "tools", "data", "benchmarks")] +
+               verbose_flags)
+    if not config.is_debug:
+      command.append("--release")
+
+    AddEntry("benchmarks", command)
+
   # Integration tests ----------------------------------------------------------
 
   if target_os == Config.OS_ANDROID and ShouldRunTest(