erts: Pass thread progress data where possible

The poll thread does a lot of waking up and then going
back to sleep. A large part of the waking up is managing
thread progress and a large part of that was using thread
specific data to get the thread progress data pointer.
With this refactor the tpd is passed to each of the functions
which greatly decreases the number of ethr_get_tsd calls
which in turn halves the CPU usage of the poller thread in
certain scenarios.

5 files changed, 21 insertions, 12 deletions

diff --git a/erts/emulator/sys/common/erl_check_io.c b/erts/emulator/sys/common/erl_check_io.c
index 9f115706dc..748555165f 100644
--- a/erts/emulator/sys/common/erl_check_io.c
+++ b/erts/emulator/sys/common/erl_check_io.c
@@ -105,6 +105,7 @@ typedef struct erts_poll_thread
 {
     ErtsPollSet *ps;
     ErtsPollResFd *pollres;
+    ErtsThrPrgrData *tpd;
     int pollres_len;
 } ErtsPollThread;
 
@@ -1516,7 +1517,8 @@ erts_check_io_interrupt(ErtsPollThread *psi, int set)
 }
 
 ErtsPollThread *
-erts_create_pollset_thread(int id) {
+erts_create_pollset_thread(int id, ErtsThrPrgrData *tpd) {
+    psiv[id].tpd = tpd;
     return psiv+id;
 }
 
@@ -1538,12 +1540,12 @@ erts_check_io(ErtsPollThread *psi)
 #if ERTS_POLL_USE_FALLBACK
     if (psi->ps == get_fallback()) {
 
-        poll_ret = erts_poll_wait_flbk(psi->ps, psi->pollres, &pollres_len);
+        poll_ret = erts_poll_wait_flbk(psi->ps, psi->pollres, &pollres_len, psi->tpd);
 
     } else
 #endif
     {
-        poll_ret = erts_poll_wait(psi->ps, psi->pollres, &pollres_len);
+        poll_ret = erts_poll_wait(psi->ps, psi->pollres, &pollres_len, psi->tpd);
     }
 
 #ifdef ERTS_ENABLE_LOCK_CHECK
diff --git a/erts/emulator/sys/common/erl_check_io.h b/erts/emulator/sys/common/erl_check_io.h
index 443ef1264c..16aba8a5f3 100644
--- a/erts/emulator/sys/common/erl_check_io.h
+++ b/erts/emulator/sys/common/erl_check_io.h
@@ -90,8 +90,11 @@ void erts_check_io_interrupt(struct erts_poll_thread *pt, int set);
 /**
  * Create a new poll thread structure that is associated with the number no.
  * It is the callers responsibility that no is unique.
+ *
+ * @param no the id of the pollset thread, -2 = aux thread, -1 = scheduler
+ * @param tpd the thread progress data of the pollset thread
  */
-struct erts_poll_thread* erts_create_pollset_thread(int no);
+struct erts_poll_thread* erts_create_pollset_thread(int no, ErtsThrPrgrData *tpd);
 #ifdef ERTS_ENABLE_LOCK_COUNT
 /**
  * Toggle lock counting on all check io locks
diff --git a/erts/emulator/sys/common/erl_poll.c b/erts/emulator/sys/common/erl_poll.c
index b4d1575ee5..5a5874ddfa 100644
--- a/erts/emulator/sys/common/erl_poll.c
+++ b/erts/emulator/sys/common/erl_poll.c
@@ -75,6 +75,7 @@
 #  define WANT_NONBLOCKING
 #endif
 
+#include "erl_thr_progress.h"
 #include "erl_poll.h"
 #if ERTS_POLL_USE_KQUEUE
 #  include <sys/types.h>
@@ -95,7 +96,6 @@
 #    include <limits.h>
 #  endif
 #endif
-#include "erl_thr_progress.h"
 #include "erl_driver.h"
 #include "erl_alloc.h"
 #include "erl_msacc.h"
@@ -1629,7 +1629,8 @@ check_fd_events(ErtsPollSet *ps, ErtsPollResFd pr[], int do_wait, int max_res)
 int
 ERTS_POLL_EXPORT(erts_poll_wait)(ErtsPollSet *ps,
 				 ErtsPollResFd pr[],
-				 int *len)
+				 int *len,
+                                 ErtsThrPrgrData *tpd)
 {
     int res, no_fds, used_fds = 0;
     int ebadf = 0;
@@ -1659,7 +1660,7 @@ ERTS_POLL_EXPORT(erts_poll_wait)(ErtsPollSet *ps,
     DEBUG_PRINT_WAIT("Entering %s(), do_wait=%d", ps, __FUNCTION__, do_wait);
 
     if (do_wait) {
-        erts_thr_progress_prepare_wait(NULL);
+        erts_thr_progress_prepare_wait(tpd);
         ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_SLEEP);
     }
 
@@ -1703,7 +1704,7 @@ ERTS_POLL_EXPORT(erts_poll_wait)(ErtsPollSet *ps,
     }
 
     if (do_wait) {
-        erts_thr_progress_finalize_wait(NULL);
+        erts_thr_progress_finalize_wait(tpd);
         ERTS_MSACC_UPDATE_CACHE();
         ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_CHECK_IO);
     }
diff --git a/erts/emulator/sys/common/erl_poll_api.h b/erts/emulator/sys/common/erl_poll_api.h
index 1170a549b9..f35f64a9f3 100644
--- a/erts/emulator/sys/common/erl_poll_api.h
+++ b/erts/emulator/sys/common/erl_poll_api.h
@@ -72,11 +72,13 @@ ErtsPollEvents ERTS_POLL_EXPORT(erts_poll_control)(ErtsPollSet *ps,
  * @param res an array of fd results that the ready fds are put in.
  * @param[in] length the length of the res array
  * @param[out] length the number of ready events returned in res
+ * @param tpd the thread progress data to note sleep state in
  * @return 0 on success, else the ERRNO of the error that happened.
  */
 int ERTS_POLL_EXPORT(erts_poll_wait)(ErtsPollSet *ps,
                                      ErtsPollResFd res[],
-                                     int *length);
+                                     int *length,
+                                     ErtsThrPrgrData *tpd);
 /**
  * Interrupt the thread waiting in the pollset. This function should be called
  * with set = 0 before any thread calls erts_poll_wait in order to clear any
diff --git a/erts/emulator/sys/win32/erl_poll.c b/erts/emulator/sys/win32/erl_poll.c
index 39bb4d515e..5d832f4d34 100644
--- a/erts/emulator/sys/win32/erl_poll.c
+++ b/erts/emulator/sys/win32/erl_poll.c
@@ -1017,7 +1017,8 @@ ErtsPollEvents erts_poll_control(ErtsPollSet *ps,
 
 int erts_poll_wait(ErtsPollSet *ps,
 		   ErtsPollResFd pr[],
-		   int *len)
+		   int *len,
+                   ErtsThrPrgrData *tpd)
 {
     int no_fds;
     DWORD timeout = INFINITE;
@@ -1056,10 +1057,10 @@ int erts_poll_wait(ErtsPollSet *ps,
 
 	HARDDEBUGF(("Start waiting %d [%d]",num_h, (int) timeout));
 	ERTS_POLLSET_UNLOCK(ps);
-	erts_thr_progress_prepare_wait(NULL);
+	erts_thr_progress_prepare_wait(tpd);
         ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_SLEEP);
 	handle = WaitForMultipleObjects(num_h, harr, FALSE, timeout);
-	erts_thr_progress_finalize_wait(NULL);
+	erts_thr_progress_finalize_wait(tpd);
         ERTS_MSACC_POP_STATE();
 	ERTS_POLLSET_LOCK(ps);
 	HARDDEBUGF(("Stop waiting %d [%d]",num_h, (int) timeout));