stats: Add support for hierarchical stats

This change makes the stat system aware of the hierarchical nature of
stats. The aim is to achieve the following goals:

  * Make the SimObject hierarchy explicit in the stat system (i.e.,
    get rid of name() + ".foo"). This makes stat naming less fragile
    and makes it possible to implement hierarchical formats like
    XML/HDF5/JSON in a clean way.

  * Make it more convenient to split stats into a separate
    struct/class that can be bound to a SimObject. This makes the
    namespace cleaner and makes stat accesses a bit more obvious.

  * Make it possible to build groups of stats in C++ that can be used
    in subcomponents in a SimObject (similar to what we do for
    checkpoint sections). This makes it easier to structure large
    components.

  * Enable partial stat dumps. Some of our internal users have been
    asking for this since a full stat dump can be large.

  * Enable better stat access from Python.

This changeset implements solves the first three points by introducing
a class (Stats::Group) that owns statistics belonging to the same
object. SimObjects inherit from Stats::Group since they typically have
statistics.

New-style statistics need to be associated with a parent group at
instantiation time. Instantiation typically sets the name and the
description, other parameters need to be set by overriding
Group::regStats() just like with legacy stats. Simple objects with
scalar stats can typically avoid implementing regStats() altogether
since the stat name and description are both specified in the
constructor.

For convenience reasons, statistics groups can be merged into other
groups. This means that a SimObject can create a stat struct that
inherits from Stats::Group and merge it into the parent group
(SimObject). This can make the code cleaner since statistics tracking
gets grouped into a single object.

Stat visitors have a new API to expose the group structure. The
Output::beginGroup(name) method is called at the beginning of a group
and the Output::endGroup() method is called when all stats, and
sub-groups, have been visited. Flat formats (e.g., the text format)
typically need to maintain a stack to track the full path to a stat.

Legacy, flat, statistics are still supported after applying this
change. These stats don't belong to any group and stat visitors will
not see a Output::beginGroup(name) call before their corresponding
Output::visit() methods are called.

Change-Id: I9025d61dfadeabcc8ecf30813ab2060def455648
Signed-off-by: Andreas Sandberg <andreas.sandberg@arm.com>
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/19368
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <jason@lowepower.com>
Reviewed-by: Daniel Carvalho <odanrc@yahoo.com.br>

src/python/m5/SimObject.py

@@ -1086,7 +1086,7 @@ class SimObject(object):
     abstract = True
 
     cxx_header = "sim/sim_object.hh"
-    cxx_extra_bases = [ "Drainable", "Serializable" ]
+    cxx_extra_bases = [ "Drainable", "Serializable", "Stats::Group" ]
     eventq_index = Param.UInt32(Parent.eventq_index, "Event Queue Index")
 
     cxx_exports = [
@@ -1094,8 +1094,6 @@ class SimObject(object):
         PyBindMethod("initState"),
         PyBindMethod("memInvalidate"),
         PyBindMethod("memWriteback"),
-        PyBindMethod("regStats"),
-        PyBindMethod("resetStats"),
         PyBindMethod("regProbePoints"),
         PyBindMethod("regProbeListeners"),
         PyBindMethod("startup"),

src/python/m5/simulate.py

@@ -124,7 +124,8 @@ def instantiate(ckpt_dir=None):
     for obj in root.descendants(): obj.init()
 
     # Do a third pass to initialize statistics
-    for obj in root.descendants(): obj.regStats()
+    stats._bindStatHierarchy(root)
+    root.regStats()
 
     # Do a fourth pass to initialize probe points
     for obj in root.descendants(): obj.regProbePoints()

src/python/m5/stats/__init__.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2017 ARM Limited
+# Copyright (c) 2017, 2019 Arm Limited
 # All rights reserved.
 #
 # The license below extends only to copyright in the software and shall
@@ -158,6 +158,29 @@ def initSimStats():
     _m5.stats.initSimStats()
     _m5.stats.registerPythonStatsHandlers()
 
+def _visit_groups(root, visitor):
+    for group in root.getStatGroups().values():
+        visitor(group)
+        _visit_groups(group, visitor)
+
+def _visit_stats(root, visitor):
+    def for_each_stat(g):
+        for stat in g.getStats():
+            visitor(g, stat)
+    _visit_groups(root, for_each_stat)
+
+def _bindStatHierarchy(root):
+    def _bind_obj(name, obj):
+        if m5.SimObject.isSimObjectVector(obj):
+            for idx, obj in enumerate(obj):
+                _bind_obj("{}{}".format(name, idx), obj)
+        else:
+            root.addStatGroup(name, obj.getCCObject())
+            _bindStatHierarchy(obj)
+
+    for name, obj in root._children.items():
+        _bind_obj(name, obj)
+
 names = []
 stats_dict = {}
 stats_list = []
@@ -166,10 +189,7 @@ def enable():
     enabled, all statistics must be created and initialized and once
     the package is enabled, no more statistics can be created.'''
 
-    global stats_list
-    stats_list = list(_m5.stats.statsList())
-
-    for stat in stats_list:
+    def check_stat(group, stat):
         if not stat.check() or not stat.baseCheck():
             fatal("statistic '%s' (%d) was not properly initialized " \
                   "by a regStats() function\n", stat.name, stat.id)
@@ -177,21 +197,57 @@ def enable():
         if not (stat.flags & flags.display):
             stat.name = "__Stat%06d" % stat.id
 
+
+    # Legacy stat
+    global stats_list
+    stats_list = list(_m5.stats.statsList())
+
+    for stat in stats_list:
+        check_stat(None, stat)
+
     stats_list.sort(key=lambda s: s.name.split('.'))
     for stat in stats_list:
         stats_dict[stat.name] = stat
         stat.enable()
 
+
+    # New stats
+    _visit_stats(Root.getInstance(), check_stat)
+    _visit_stats(Root.getInstance(), lambda g, s: s.enable())
+
     _m5.stats.enable();
 
 def prepare():
     '''Prepare all stats for data access.  This must be done before
     dumping and serialization.'''
 
+    # Legacy stats
     for stat in stats_list:
         stat.prepare()
 
+    # New stats
+    _visit_stats(Root.getInstance(), lambda g, s: s.prepare())
+
 lastDump = 0
+
+def _dump_to_visitor(visitor):
+    # Legacy stats
+    for stat in stats_list:
+        stat.visit(visitor)
+
+    # New stats
+    def dump_group(group):
+        for stat in group.getStats():
+            stat.visit(visitor)
+
+        for n, g in group.getStatGroups().items():
+            visitor.beginGroup(n)
+            dump_group(g)
+            visitor.endGroup()
+
+    dump_group(Root.getInstance())
+
+
 def dump():
     '''Dump all statistics data to the registered outputs'''
 
@@ -210,8 +266,7 @@ def dump():
     for output in outputList:
         if output.valid():
             output.begin()
-            for stat in stats_list:
-                stat.visit(output)
+            _dump_to_visitor(output)
             output.end()
 
 def reset():
@@ -220,9 +275,9 @@ def reset():
     # call reset stats on all SimObjects
     root = Root.getInstance()
     if root:
-        for obj in root.descendants(): obj.resetStats()
+        root.resetStats()
 
-    # call any other registered stats reset callbacks
+    # call any other registered legacy stats reset callbacks
     for stat in stats_list:
         stat.reset()
 

src/python/pybind11/stats.cc

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017 ARM Limited
+ * Copyright (c) 2017, 2019 ARM Limited
  * All rights reserved
  *
  * The license below extends only to copyright in the software and shall
@@ -93,9 +93,12 @@ pybind_init_stats(py::module &m_native)
         .def("begin", &Stats::Output::begin)
         .def("end", &Stats::Output::end)
         .def("valid", &Stats::Output::valid)
+        .def("beginGroup", &Stats::Output::beginGroup)
+        .def("endGroup", &Stats::Output::endGroup)
         ;
 
-    py::class_<Stats::Info>(m, "Info")
+    py::class_<Stats::Info, std::unique_ptr<Stats::Info, py::nodelete>>(
+        m, "Info")
         .def_readwrite("name", &Stats::Info::name)
         .def_readonly("desc", &Stats::Info::desc)
         .def_readonly("id", &Stats::Info::id)
@@ -110,4 +113,13 @@ pybind_init_stats(py::module &m_native)
         .def("zero", &Stats::Info::zero)
         .def("visit", &Stats::Info::visit)
         ;
+
+    py::class_<Stats::Group, std::unique_ptr<Stats::Group, py::nodelete>>(
+        m, "Group")
+        .def("regStats", &Stats::Group::regStats)
+        .def("resetStats", &Stats::Group::resetStats)
+        .def("getStats", &Stats::Group::getStats)
+        .def("getStatGroups", &Stats::Group::getStatGroups)
+        .def("addStatGroup", &Stats::Group::addStatGroup)
+        ;
 }