erts: Add microstate accounting

Microstate accounting is a way to track which state the
different threads within ERTS are in. The main usage area
is to pin point performance bottlenecks by checking which
states the threads are in and then from there figuring out
why and where to optimize.

Since checking whether microstate accounting is on or off is
relatively expensive if done in a short loop only a few of the
states are enabled by default and more states can be enabled
through configure.

I've done some benchmarking and the overhead with it turned off
is not noticible and with it on it is a fraction of a percent.
If you enable the extra states, depending on the benchmark,
the ovehead when turned off is about 1% and when turned on
somewhere inbetween 5-15%.

OTP-12345

2 files changed, 32 insertions, 8 deletions

diff --git a/erts/emulator/sys/common/erl_poll.c b/erts/emulator/sys/common/erl_poll.c
index 21fa214364..36ee94111c 100644
--- a/erts/emulator/sys/common/erl_poll.c
+++ b/erts/emulator/sys/common/erl_poll.c
@@ -74,6 +74,7 @@
 #include "erl_thr_progress.h"
 #include "erl_driver.h"
 #include "erl_alloc.h"
+#include "erl_msacc.h"
 
 #if !defined(ERTS_POLL_USE_EPOLL) \
     && !defined(ERTS_POLL_USE_DEVPOLL)  \
@@ -2238,6 +2239,7 @@ static ERTS_INLINE int
 check_fd_events(ErtsPollSet ps, ErtsMonotonicTime timeout_time, int max_res)
 {
     int res;
+    ERTS_MSACC_PUSH_STATE_M();
     if (erts_smp_atomic_read_nob(&ps->no_of_user_fds) == 0
 	&& timeout_time == ERTS_POLL_NO_TIMEOUT) {
 	/* Nothing to poll and zero timeout; done... */
@@ -2259,6 +2261,7 @@ check_fd_events(ErtsPollSet ps, ErtsMonotonicTime timeout_time, int max_res)
 #ifdef ERTS_SMP
                     erts_thr_progress_prepare_wait(NULL);
 #endif
+                    ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
                     timerfd_set(ps, &its);
                     res = epoll_wait(ps->kp_fd, ps->res_events, max_res, -1);
                     res = timerfd_clear(ps, res, max_res);
@@ -2268,10 +2271,12 @@ check_fd_events(ErtsPollSet ps, ErtsMonotonicTime timeout_time, int max_res)
             }
 #else /* !ERTS_POLL_USE_TIMERFD */
 	    timeout = (int) get_timeout(ps, 1000, timeout_time);
+            if (timeout) {
 #ifdef ERTS_SMP
-	    if (timeout)
 		erts_thr_progress_prepare_wait(NULL);
 #endif
+                ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
+            }
 	    res = epoll_wait(ps->kp_fd, ps->res_events, max_res, timeout);
 #endif /* !ERTS_POLL_USE_TIMERFD */
 #elif ERTS_POLL_USE_KQUEUE	/* --- kqueue ------------------------------ */
@@ -2279,10 +2284,12 @@ check_fd_events(ErtsPollSet ps, ErtsMonotonicTime timeout_time, int max_res)
 	    if (max_res > ps->res_events_len)
 		grow_res_events(ps, max_res);
 	    timeout = get_timeout_timespec(ps, &ts, timeout_time);
+            if (timeout) {
 #ifdef ERTS_SMP
-	    if (timeout)
 		erts_thr_progress_prepare_wait(NULL);
 #endif
+		ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
+            }
 	    res = kevent(ps->kp_fd, NULL, 0, ps->res_events, max_res, &ts);
 #endif				/* ----------------------------------------- */
 	}
@@ -2306,26 +2313,33 @@ check_fd_events(ErtsPollSet ps, ErtsMonotonicTime timeout_time, int max_res)
 	    if (poll_res.dp_nfds > ps->res_events_len)
 		grow_res_events(ps, poll_res.dp_nfds);
 	    poll_res.dp_fds = ps->res_events;
+	    if (timeout) {
 #ifdef ERTS_SMP
-	    if (timeout)
 		erts_thr_progress_prepare_wait(NULL);
 #endif
+                ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
+            }
 	    poll_res.dp_timeout = timeout;
 	    res = ioctl(ps->kp_fd, DP_POLL, &poll_res);
 #elif ERTS_POLL_USE_POLL && defined(HAVE_PPOLL)	/* --- ppoll ---------------- */
             struct timespec ts;
 	    timeout = get_timeout_timespec(ps, &ts, timeout_time);
+            if (timeout) {
 #ifdef ERTS_SMP
-	    if (timeout)
 		erts_thr_progress_prepare_wait(NULL);
 #endif
+		ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
+            }
             res = ppoll(ps->poll_fds, ps->no_poll_fds, &ts, NULL);
 #elif ERTS_POLL_USE_POLL        /* --- poll --------------------------------- */
 	    timeout = (int) get_timeout(ps, 1000, timeout_time);
+
+	    if (timeout) {
 #ifdef ERTS_SMP
-	    if (timeout)
 		erts_thr_progress_prepare_wait(NULL);
 #endif
+		ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
+            }
 	    res = poll(ps->poll_fds, ps->no_poll_fds, timeout);
 #elif ERTS_POLL_USE_SELECT	/* --- select ------------------------------ */
 	    SysTimeval to;
@@ -2334,18 +2348,22 @@ check_fd_events(ErtsPollSet ps, ErtsMonotonicTime timeout_time, int max_res)
 	    ERTS_FD_COPY(&ps->input_fds, &ps->res_input_fds);
 	    ERTS_FD_COPY(&ps->output_fds, &ps->res_output_fds);
 
+	    if (timeout) {
 #ifdef ERTS_SMP
-	    if (timeout)
 		erts_thr_progress_prepare_wait(NULL);
 #endif
+		ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
+	    }
 	    res = ERTS_SELECT(ps->max_fd + 1,
 			      &ps->res_input_fds,
 			      &ps->res_output_fds,
 			      NULL,
 			      &to);
 #ifdef ERTS_SMP
-	    if (timeout)
+	    if (timeout) {
 		erts_thr_progress_finalize_wait(NULL);
+		ERTS_MSACC_POP_STATE_M();
+	    }
 	    if (res < 0
 		&& errno == EBADF
 		&& ERTS_POLLSET_HAVE_UPDATE_REQUESTS(ps)) {
@@ -2380,10 +2398,12 @@ check_fd_events(ErtsPollSet ps, ErtsMonotonicTime timeout_time, int max_res)
 	    return res;
 #endif				/* ----------------------------------------- */
 	}
+	if (timeout) {
 #ifdef ERTS_SMP
-	if (timeout)
 	    erts_thr_progress_finalize_wait(NULL);
 #endif
+	    ERTS_MSACC_POP_STATE_M();
+	}
 	return res;
     }
 }
diff --git a/erts/emulator/sys/win32/erl_poll.c b/erts/emulator/sys/win32/erl_poll.c
index 466f4a3b48..660ded297a 100644
--- a/erts/emulator/sys/win32/erl_poll.c
+++ b/erts/emulator/sys/win32/erl_poll.c
@@ -27,6 +27,7 @@
 #include "erl_alloc.h"
 #include "erl_poll.h"
 #include "erl_time.h"
+#include "erl_msacc.h"
 
 /*
  * Some debug macros 
@@ -1188,16 +1189,19 @@ int erts_poll_wait(ErtsPollSet ps,
     if (timeout > 0 && !erts_atomic32_read_nob(&break_waiter_state)) {
 	HANDLE harr[2] = {ps->event_io_ready, break_happened_event};
 	int num_h = 2;
+        ERTS_MSACC_PUSH_STATE_M();
 
 	HARDDEBUGF(("Start waiting %d [%d]",num_h, (int) timeout));
 	ERTS_POLLSET_UNLOCK(ps);
 #ifdef ERTS_SMP
 	erts_thr_progress_prepare_wait(NULL);
 #endif
+        ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
 	WaitForMultipleObjects(num_h, harr, FALSE, timeout);
 #ifdef ERTS_SMP
 	erts_thr_progress_finalize_wait(NULL);
 #endif
+        ERTS_MSACC_POP_STATE_M();
 	ERTS_POLLSET_LOCK(ps);
 	HARDDEBUGF(("Stop waiting %d [%d]",num_h, (int) timeout));
 	woke_up(ps);