Merge branch 'lukas/erts/msacc'

* lukas/erts/msacc:
  Update preloaded modules
  erts: Make msacc alloctor type thread safe
  Silence compiler
  erts: Fix msacc testcase on some windowses
  erts: Add power saving cpu feature tests and use them
  erts: Refactor perf counter internal interface
  erts: Add rdtscp instruction check
  erts: Fix hrtime for windows
  erts: use correct function for perf counter on non-x86
  erts: Fix msacc win32 debug compile error
  erts: Add microstate accounting
  erts, kernel: Add os:perf_counter function
  erts: Add ERTS_WRITE_UNLIKELY

1 files changed, 175 insertions, 13 deletions

diff --git a/erts/doc/src/erlang.xml b/erts/doc/src/erlang.xml
index d3aad94660..803da382ed 100644
--- a/erts/doc/src/erlang.xml
+++ b/erts/doc/src/erlang.xml
@@ -131,6 +131,17 @@
 	  </note>
 	  </item>
 
+	  <tag><c>perf_counter</c></tag>
+	  <item><p>Symbolic representation of the performance counter
+          time unit used by the Erlang runtime system.</p>
+
+          <p>The <c>perf_counter</c> time unit behaves much in the same way
+          as the <c>native</c> time unit. That is it might differ inbetween
+          run-time restarts. You get values of this type by calling
+          <seealso marker="kernel:os#perf_counter/0"><c>os:perf_counter()</c></seealso>
+          </p>
+          </item>
+
 	</taglist>
 
 	<p>The <c>time_unit/0</c> type may be extended. Use
@@ -5859,6 +5870,146 @@ true</pre>
 
     <func>
       <name name="statistics" arity="1" clause_i="6"/>
+      <fsummary>Information about microstate accounting.</fsummary>
+      <desc>
+        <marker id="statistics_microstate_accounting"></marker>
+        <p>
+          Microstate accounting can be used to measure how much time the Erlang
+          runtime system spends doing various tasks. It is designed to be as
+          lightweight as possible, but there will be some overhead when this
+          is enabled. Microstate accounting is meant to be a profiling tool
+          to help figure out performance bottlenecks.
+          To <c>start</c>/<c>stop</c>/<c>reset</c> microstate_accounting you use
+          the system_flag
+          <seealso marker="#system_flag_microstate_accounting">
+            <c>microstate_accounting</c></seealso>.
+        </p>
+        <p>
+          <c>erlang:statistics(microstate_accounting)</c> returns a list of maps
+          representing some of the OS threads within ERTS. Each map contains
+          <c>type</c> and <c>id</c> fields that can be used to identify what
+          thread it is, and also a counters field that contains data about how
+          much time has been spent in the various states.</p>
+        <pre>
+> <input>erlang:statistics(microstate_accounting).</input>
+[#{counters => #{aux => 1899182914,
+                 check_io => 2605863602,
+                 emulator => 45731880463,
+                 gc => 1512206910,
+                 other => 5421338456,
+                 port => 221631,
+                 sleep => 5150294100},
+   id => 1,
+   type => scheduler}|...]
+        </pre>
+        <p>The time unit is the same as returned by
+        <seealso marker="kernel:os#perf_counter/0">
+          <c>os:perf_counter/0</c></seealso>.
+        So to convert it to milliseconds you could do something like this:</p>
+        <pre>
+lists:map(
+  fun(#{ counters := Cnt } = M) ->
+          MsCnt = maps:map(fun(_K, PerfCount) ->
+                                   erlang:convert_time_unit(PerfCount, perf_counter, 1000)
+                           end, Cnt),
+         M#{ counters := MsCnt }
+  end, erlang:statistics(microstate_accounting)).
+        </pre>
+        <p>
+          It is important to note that these values are not guaranteed to be
+          the exact time spent in each state. This is because of various
+          optimisation done in order to keep the overhead as small as possible.
+        </p>
+
+        <p>Currently the following <c><anno>MSAcc_Thread_Type</anno></c> are available:</p>
+        <taglist>
+          <tag><c>scheduler</c></tag>
+          <item>The main execution threads that do most of the work.</item>
+          <tag><c>async</c></tag><item>Async threads are used by various
+          linked-in drivers (mainly the file drivers) do offload non-cpu
+          intensive work.</item>
+          <tag><c>aux</c></tag><item>Takes care of any work that is not
+          specifically assigned to a scheduler.</item>
+        </taglist>
+        <p>Currently the following <c><anno>MSAcc_Thread_State</anno></c>s are available.
+        All states are exclusive, meaning that a thread cannot be in two states
+        at once. So if you add the numbers of all counters in a thread
+        you will get the total run-time for that thread.</p>
+        <taglist>
+          <tag><c>aux</c></tag>
+          <item>Time spent handling auxiliary jobs.</item>
+          <tag><c>check_io</c></tag>
+          <item>Time spent checking for new I/O events.</item>
+          <tag><c>emulator</c></tag>
+          <item>Time spent executing erlang processes.</item>
+          <tag><c>gc</c></tag>
+          <item>Time spent doing garbage collection. When extra states are
+          enabled this is the time spent doing non-fullsweep garbage
+          collections.</item>
+          <tag><c>other</c></tag>
+          <item>Time spent doing unaccounted things.</item>
+          <tag><c>port</c></tag>
+          <item>Time spent executing ports.</item>
+          <tag><c>sleep</c></tag>
+          <item>Time spent sleeping.</item>
+        </taglist>
+        <p>It is possible to add more fine grained <c><anno>MSAcc_Thread_State</anno></c>s
+        through configure.
+        (e.g. <c>./configure --with-microstate-accounting=extra</c>).
+        Enabling these states will cause a performance degradation when
+        microstate accounting is turned off and increase the overhead when
+        it is turned on.</p>
+        <taglist>
+          <tag><c>alloc</c></tag>
+          <item>Time spent managing memory. Without extra states this time is
+          spread out over all other states.</item>
+          <tag><c>bif</c></tag>
+          <item>Time spent in bifs. Without extra states this time is part of
+          the <c>emulator</c> state.</item>
+          <tag><c>busy_wait</c></tag>
+          <item>Time spent busy waiting. This is also the state where a
+          scheduler no longer reports that it is active when using
+          <seealso marker="#statistics_scheduler_wall_time">
+            <c>erlang:statistics(scheduler_wall_time)</c></seealso>.
+          So if you add all other states but this and sleep and then divide that
+          by all time in the thread you should get something very similar to the
+          scheduler_wall_time fraction. Without extra states this time is part
+          of the <c>other</c> state.</item>
+          <tag><c>ets</c></tag>
+          <item>Time spent executing ETS bifs. Without extra states this time is
+          part of the <c>emulator</c> state.</item>
+          <tag><c>gc_full</c></tag>
+          <item>Time spent doing fullsweep garbage collection. Without extra
+          states this time is part of the <c>gc</c> state.</item>
+          <tag><c>nif</c></tag>
+          <item>Time spent in nifs. Without extra states this time is part of
+          the <c>emulator</c> state.</item>
+          <tag><c>send</c></tag>
+          <item>Time spent sending messages (processes only). Without extra
+          states this time is part of the <c>emulator</c> state.</item>
+          <tag><c>timers</c></tag>
+          <item>Time spent managing timers. Without extra states this time is
+          part of the <c>other</c> state.</item>
+        </taglist>
+        <p>There is a utility module called
+        <seealso marker="runtime_tools:msacc"><c>msacc</c></seealso> in
+        runtime_tools that can be used to more easily analyse these
+        statistics.</p>
+
+        <p>
+          Returns <c>undefined</c> if the system flag
+          <seealso marker="#system_flag_microstate_accounting">
+            <c>microstate_accounting</c></seealso>
+          is turned off.
+	</p>
+        <p>The list of thread information is unsorted and may appear in
+        different order between calls.</p>
+        <note><p>The threads and states are subject to change without any
+        prior notice.</p></note>
+      </desc>
+    </func>
+    <func>
+      <name name="statistics" arity="1" clause_i="7"/>
       <fsummary>Information about reductions.</fsummary>
       <desc>
         <marker id="statistics_reductions"></marker>
@@ -5876,7 +6027,7 @@ true</pre>
     </func>
 
     <func>
-      <name name="statistics" arity="1" clause_i="7"/>
+      <name name="statistics" arity="1" clause_i="8"/>
       <fsummary>Information about the run-queues.</fsummary>
       <desc><marker id="statistics_run_queue"></marker>
         <p>
@@ -5892,7 +6043,7 @@ true</pre>
     </func>
 
     <func>
-      <name name="statistics" arity="1" clause_i="8"/>
+      <name name="statistics" arity="1" clause_i="9"/>
       <fsummary>Information about the run-queue lengths.</fsummary>
       <desc><marker id="statistics_run_queue_lengths"></marker>
 	<p>
@@ -5912,7 +6063,7 @@ true</pre>
     </func>
 
     <func>
-      <name name="statistics" arity="1" clause_i="9"/>
+      <name name="statistics" arity="1" clause_i="10"/>
       <fsummary>Information about runtime.</fsummary>
       <desc>
         <p>Returns information about runtime, in milliseconds.</p>
@@ -5927,7 +6078,7 @@ true</pre>
     </func>
 
     <func>
-      <name name="statistics" arity="1" clause_i="10"/>
+      <name name="statistics" arity="1" clause_i="11"/>
       <fsummary>Information about each schedulers work time.</fsummary>
       <desc>
       <marker id="statistics_scheduler_wall_time"></marker>
@@ -5998,7 +6149,7 @@ ok
     </func>
 
     <func>
-      <name name="statistics" arity="1" clause_i="11"/>
+      <name name="statistics" arity="1" clause_i="12"/>
       <fsummary>Information about active processes and ports.</fsummary>
       <desc><marker id="statistics_total_active_tasks"></marker>
 	<p>
@@ -6016,7 +6167,7 @@ ok
     </func>
 
     <func>
-      <name name="statistics" arity="1" clause_i="12"/>
+      <name name="statistics" arity="1" clause_i="13"/>
       <fsummary>Information about the run-queue lengths.</fsummary>
       <desc><marker id="statistics_total_run_queue_lengths"></marker>
         <p>
@@ -6035,7 +6186,7 @@ ok
     </func>
 
     <func>
-      <name name="statistics" arity="1" clause_i="13"/>
+      <name name="statistics" arity="1" clause_i="14"/>
       <fsummary>Information about wall clock.</fsummary>
       <desc>
         <p>Returns information about wall clock. <c>wall_clock</c> can
@@ -6269,6 +6420,17 @@ ok
 
     <func>
       <name name="system_flag" arity="2" clause_i="5"/>
+      <fsummary>Set system flag microstate_accounting</fsummary>
+      <desc><p><marker id="system_flag_microstate_accounting"></marker>
+      Turns on/off microstate accounting measurements. By passing reset it is possible to reset
+      all counters to 0.</p>
+      <p>For more information see,
+      <seealso marker="#statistics_microstate_accounting">erlang:statistics(microstate_accounting)</seealso>.
+      </p>
+      </desc>
+    </func>
+    <func>
+      <name name="system_flag" arity="2" clause_i="6"/>
       <fsummary>Sets system flag <c>min_heap_size</c>.</fsummary>
       <desc>
         <p>Sets the default minimum heap size for processes. The size
@@ -6283,7 +6445,7 @@ ok
     </func>
 
     <func>
-      <name name="system_flag" arity="2" clause_i="6"/>
+      <name name="system_flag" arity="2" clause_i="7"/>
       <fsummary>Sets system flag <c>min_bin_vheap_size</c>.</fsummary>
       <desc>
         <p>Sets the default minimum binary virtual heap size for
@@ -6300,7 +6462,7 @@ ok
     </func>
 
     <func>
-      <name name="system_flag" arity="2" clause_i="7"/>
+      <name name="system_flag" arity="2" clause_i="8"/>
       <fsummary>Sets system flag <c>multi_scheduling</c>.</fsummary>
       <desc>
         <p><marker id="system_flag_multi_scheduling"></marker>
@@ -6338,7 +6500,7 @@ ok
     </func>
 
     <func>
-      <name name="system_flag" arity="2" clause_i="8"/>
+      <name name="system_flag" arity="2" clause_i="9"/>
       <fsummary>Sets system flag <c>scheduler_bind_type</c>.</fsummary>
       <type name="scheduler_bind_type"/>
       <desc>
@@ -6456,7 +6618,7 @@ ok
     </func>
 
     <func>
-      <name name="system_flag" arity="2" clause_i="9"/>
+      <name name="system_flag" arity="2" clause_i="10"/>
       <fsummary>Sets system flag <c>scheduler_wall_time</c>.</fsummary>
       <desc><p><marker id="system_flag_scheduler_wall_time"></marker>
       Turns on or off scheduler wall time measurements.</p>
@@ -6466,7 +6628,7 @@ ok
     </func>
 
     <func>
-      <name name="system_flag" arity="2" clause_i="10"/>
+      <name name="system_flag" arity="2" clause_i="11"/>
       <fsummary>Sets system flag <c>schedulers_online</c>.</fsummary>
       <desc>
         <p><marker id="system_flag_schedulers_online"></marker>
@@ -6491,7 +6653,7 @@ ok
     </func>
 
     <func>
-      <name name="system_flag" arity="2" clause_i="11"/>
+      <name name="system_flag" arity="2" clause_i="12"/>
       <fsummary>Sets system flag <c>trace_control_word</c>.</fsummary>
       <desc>
         <p>Sets the value of the node trace control word to