[PATCH 2/3][RFC] Batch getthrds() calls in get_signaled_thread

[PATCH 2/3][RFC] Batch getthrds() calls in get_signaled_thread

[Aditya Vidyadhar Kamath]

From: Aditya Vidyadhar Kamath <aditya.kamath1@ibm.com>

The original code called getthrds() with a count of 1, fetching one
kernel thread descriptor per syscall.  With N threads in the process
that costs N syscalls per stop just to find the signaled thread.

Change to fetch GETTHRDS_BATCH (64) descriptors per call and scan
the batch in a loop.  This reduces the syscall count from O(N) to
O(N/64).

Also done the benchmark here who results are pasted below. To analyse
where time is being spent. For now, this patch request for comment/draft and will
remove the same and this paragraph from the commit message after
feedback.

==========
Benchmark C code
// cat ../bench_next.c
/* Spawns NUM_THREADS worker threads that spin, then steps through
   10 assignments in do_steps() so GDB can time each next.

   Also tests two scenarios from the bug report:
     1. "next" over a call that spawns a thread (spawn_thread_here()).
     2. A background thread exits while "next" is in progress
        (exit_thread_here()).  */
#include <stdio.h>
#include <stdlib.h>
#include <pthread.h>
#include <unistd.h>

#ifndef NUM_THREADS
#define NUM_THREADS 3
#endif

static volatile int keep_running = 1;
static volatile int extra_thread_started = 0;

static void *
worker (void *arg)
{
  while (keep_running)
    sched_yield ();
  return NULL;
}

/* Thread that exits on its own after signalling it has started.  */
static void *
short_lived (void *arg)
{
  extra_thread_started = 1;
  return NULL;
}

/* GDB will "next" over this call -- a new thread is spawned inside.  */
static void
spawn_thread_here (void)
{
  pthread_t t;
  pthread_create (&t, NULL, short_lived, NULL);
  pthread_join (t, NULL);
}

/* GDB will "next" over this call -- the short-lived background thread
   is in the process of exiting while we step.  */
static void
exit_thread_here (void)
{
  /* Just a visible marker so GDB can stop here.  */
  volatile int x = 42;
  (void)x;
}

void
do_steps (void)
{
  volatile int a = 1;
  volatile int b = 2;
  volatile int c = a + b;
  volatile int d = c * 2;
  volatile int e = d - a;
  volatile int f = e + c;
  volatile int g = f / 2;
  volatile int h = g + 1;
  volatile int i2 = h * h;
  volatile int j = i2 - b;
  (void)j;
}

int
main (void)
{
  pthread_t threads[NUM_THREADS];
  pthread_t bg;
  int i;

  /* Start background spinning threads.  */
  for (i = 0; i < NUM_THREADS; i++)
    pthread_create (&threads[i], NULL, worker, NULL);

  /* Next over a call that spawns a thread */
  spawn_thread_here ();

  /* Next while a background thread is exiting
     Launch a thread that exits quickly; "next" through exit_thread_here()
     while that exit is in flight.  */
  pthread_create (&bg, NULL, short_lived, NULL);
  exit_thread_here ();
  pthread_join (bg, NULL);

  do_steps ();

  keep_running = 0;
  for (i = 0; i < NUM_THREADS; i++)
    pthread_join (threads[i], NULL);

  return 0;
}
============
Benchmark GDB code:

// cat ../bench_next.gdb 
set pagination off
set confirm off
maintenance set per-command time on

# Scenario 1: next over a call that spawns a thread
break spawn_thread_here
run
next

# Scenario 2: next while a background thread is exiting
break exit_thread_here
continue
next

break do_steps
continue

next
next
next
next
next
next
next
next
next
next

quit

===============
# cd /current_gdb && time /current_gdb/binutils-gdb/gdb/gdb -batch -x bench_next.gdb ./bench_next_bin 2>&1 | grep -E "^(real|user|sys)|volatile int
|pd_update total|sync_thread|pthdb_session"
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "pd_update total": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "pd_update total": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.146, user 0.081, sys 0.085, user+sys 0.166, 14.5 % CPU
Time for "pd_update total": wall 1.146, user 0.081, sys 0.085, user+sys 0.166, 14.5 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.025, user 0.092, sys 0.092, user+sys 0.184, 18.0 % CPU
Time for "pd_update total": wall 1.025, user 0.092, sys 0.092, user+sys 0.184, 18.0 % CPU
25        volatile int a = 1;
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 0.962, user 0.098, sys 0.095, user+sys 0.193, 20.1 % CPU
Time for "pd_update total": wall 0.962, user 0.098, sys 0.095, user+sys 0.193, 20.1 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 0.965, user 0.098, sys 0.094, user+sys 0.192, 19.9 % CPU
Time for "pd_update total": wall 0.965, user 0.098, sys 0.094, user+sys 0.192, 19.9 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 0.954, user 0.098, sys 0.094, user+sys 0.192, 20.1 % CPU
Time for "pd_update total": wall 0.954, user 0.098, sys 0.094, user+sys 0.192, 20.1 % CPU
26        volatile int b = 2;
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 0.949, user 0.098, sys 0.094, user+sys 0.192, 20.2 % CPU
Time for "pd_update total": wall 0.949, user 0.098, sys 0.094, user+sys 0.192, 20.2 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 0.957, user 0.098, sys 0.094, user+sys 0.192, 20.1 % CPU
Time for "pd_update total": wall 0.957, user 0.098, sys 0.094, user+sys 0.192, 20.1 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.033, user 0.101, sys 0.098, user+sys 0.199, 19.3 % CPU
Time for "pd_update total": wall 1.033, user 0.101, sys 0.098, user+sys 0.199, 19.3 % CPU
27        volatile int c = a + b;
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.031, user 0.101, sys 0.098, user+sys 0.199, 19.3 % CPU
Time for "pd_update total": wall 1.031, user 0.101, sys 0.098, user+sys 0.199, 19.3 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.028, user 0.100, sys 0.098, user+sys 0.198, 19.3 % CPU
Time for "pd_update total": wall 1.028, user 0.100, sys 0.098, user+sys 0.198, 19.3 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 0.996, user 0.100, sys 0.098, user+sys 0.198, 19.9 % CPU
Time for "pd_update total": wall 0.996, user 0.100, sys 0.098, user+sys 0.198, 19.9 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.000, user 0.098, sys 0.098, user+sys 0.196, 19.6 % CPU
Time for "pd_update total": wall 1.001, user 0.098, sys 0.098, user+sys 0.196, 19.6 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 0.998, user 0.098, sys 0.098, user+sys 0.196, 19.6 % CPU
Time for "pd_update total": wall 0.998, user 0.098, sys 0.098, user+sys 0.196, 19.6 % CPU
28        volatile int d = c * 2;
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 0.999, user 0.098, sys 0.098, user+sys 0.196, 19.6 % CPU
Time for "pd_update total": wall 0.999, user 0.098, sys 0.098, user+sys 0.196, 19.6 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.001, user 0.099, sys 0.098, user+sys 0.197, 19.7 % CPU
Time for "pd_update total": wall 1.001, user 0.099, sys 0.098, user+sys 0.197, 19.7 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 0.969, user 0.100, sys 0.095, user+sys 0.195, 20.1 % CPU
Time for "pd_update total": wall 0.969, user 0.100, sys 0.095, user+sys 0.195, 20.1 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 0.980, user 0.099, sys 0.096, user+sys 0.195, 19.9 % CPU
Time for "pd_update total": wall 0.980, user 0.099, sys 0.096, user+sys 0.195, 19.9 % CPU
29        volatile int e = d - a;
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 0.984, user 0.100, sys 0.096, user+sys 0.196, 19.9 % CPU
Time for "pd_update total": wall 0.984, user 0.100, sys 0.096, user+sys 0.196, 19.9 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.094, user 0.098, sys 0.094, user+sys 0.192, 17.6 % CPU
Time for "pd_update total": wall 1.094, user 0.099, sys 0.094, user+sys 0.193, 17.6 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.019, user 0.106, sys 0.097, user+sys 0.203, 19.9 % CPU
Time for "pd_update total": wall 1.019, user 0.106, sys 0.097, user+sys 0.203, 19.9 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.002, user 0.106, sys 0.096, user+sys 0.202, 20.2 % CPU
Time for "pd_update total": wall 1.002, user 0.106, sys 0.096, user+sys 0.202, 20.2 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.003, user 0.105, sys 0.096, user+sys 0.201, 20.0 % CPU
Time for "pd_update total": wall 1.003, user 0.105, sys 0.096, user+sys 0.201, 20.0 % CPU
30        volatile int f = e + c;
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.011, user 0.106, sys 0.097, user+sys 0.203, 20.1 % CPU
Time for "pd_update total": wall 1.011, user 0.106, sys 0.097, user+sys 0.203, 20.1 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.010, user 0.107, sys 0.096, user+sys 0.203, 20.1 % CPU
Time for "pd_update total": wall 1.010, user 0.107, sys 0.096, user+sys 0.203, 20.1 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.026, user 0.110, sys 0.096, user+sys 0.206, 20.1 % CPU
Time for "pd_update total": wall 1.026, user 0.110, sys 0.096, user+sys 0.206, 20.1 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.200, user 0.090, sys 0.085, user+sys 0.175, 14.6 % CPU
Time for "pd_update total": wall 1.201, user 0.090, sys 0.085, user+sys 0.175, 14.6 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.212, user 0.090, sys 0.091, user+sys 0.181, 14.9 % CPU
Time for "pd_update total": wall 1.212, user 0.090, sys 0.091, user+sys 0.181, 14.9 % CPU
31        volatile int g = f / 2;
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 0.993, user 0.103, sys 0.096, user+sys 0.199, 20.0 % CPU
Time for "pd_update total": wall 0.993, user 0.103, sys 0.096, user+sys 0.199, 20.0 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.258, user 0.079, sys 0.082, user+sys 0.161, 12.8 % CPU
Time for "pd_update total": wall 1.258, user 0.079, sys 0.082, user+sys 0.161, 12.8 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.255, user 0.080, sys 0.081, user+sys 0.161, 12.8 % CPU
Time for "pd_update total": wall 1.255, user 0.080, sys 0.081, user+sys 0.161, 12.8 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.259, user 0.080, sys 0.081, user+sys 0.161, 12.8 % CPU
Time for "pd_update total": wall 1.259, user 0.080, sys 0.081, user+sys 0.161, 12.8 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.264, user 0.080, sys 0.082, user+sys 0.162, 12.8 % CPU
Time for "pd_update total": wall 1.265, user 0.080, sys 0.082, user+sys 0.162, 12.8 % CPU
32        volatile int h = g + 1;
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.261, user 0.079, sys 0.083, user+sys 0.162, 12.8 % CPU
Time for "pd_update total": wall 1.261, user 0.079, sys 0.083, user+sys 0.162, 12.8 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.232, user 0.076, sys 0.081, user+sys 0.157, 12.7 % CPU
Time for "pd_update total": wall 1.232, user 0.076, sys 0.081, user+sys 0.157, 12.7 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.250, user 0.078, sys 0.083, user+sys 0.161, 12.9 % CPU
Time for "pd_update total": wall 1.250, user 0.078, sys 0.083, user+sys 0.161, 12.9 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.258, user 0.080, sys 0.084, user+sys 0.164, 13.0 % CPU
Time for "pd_update total": wall 1.258, user 0.080, sys 0.084, user+sys 0.164, 13.0 % CPU
33        volatile int i2 = h * h;
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.276, user 0.082, sys 0.082, user+sys 0.164, 12.9 % CPU
Time for "pd_update total": wall 1.276, user 0.082, sys 0.082, user+sys 0.164, 12.9 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.271, user 0.083, sys 0.081, user+sys 0.164, 12.9 % CPU
Time for "pd_update total": wall 1.271, user 0.083, sys 0.081, user+sys 0.164, 12.9 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.280, user 0.079, sys 0.086, user+sys 0.165, 12.9 % CPU
Time for "pd_update total": wall 1.280, user 0.079, sys 0.086, user+sys 0.165, 12.9 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.211, user 0.075, sys 0.081, user+sys 0.156, 12.9 % CPU
Time for "pd_update total": wall 1.211, user 0.075, sys 0.081, user+sys 0.156, 12.9 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.264, user 0.077, sys 0.086, user+sys 0.163, 12.9 % CPU
Time for "pd_update total": wall 1.264, user 0.077, sys 0.086, user+sys 0.163, 12.9 % CPU
34        volatile int j = i2 - b;
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.240, user 0.078, sys 0.081, user+sys 0.159, 12.8 % CPU
Time for "pd_update total": wall 1.240, user 0.078, sys 0.081, user+sys 0.159, 12.8 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.256, user 0.078, sys 0.082, user+sys 0.160, 12.7 % CPU
Time for "pd_update total": wall 1.257, user 0.078, sys 0.082, user+sys 0.160, 12.7 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.267, user 0.078, sys 0.082, user+sys 0.160, 12.6 % CPU
Time for "pd_update total": wall 1.267, user 0.078, sys 0.082, user+sys 0.160, 12.6 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.254, user 0.079, sys 0.081, user+sys 0.160, 12.8 % CPU
Time for "pd_update total": wall 1.255, user 0.079, sys 0.081, user+sys 0.160, 12.7 % CPU
Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
Time for "sync_threadlists": wall 1.257, user 0.078, sys 0.082, user+sys 0.160, 12.7 % CPU
Time for "pd_update total": wall 1.257, user 0.078, sys 0.082, user+sys 0.160, 12.7 % CPU

real    0m52.611s
user    0m5.930s
sys     0m8.012s
---
 gdb/aix-thread.c | 60 ++++++++++++++++++++++++++++++++++--------------
 1 file changed, 43 insertions(+), 17 deletions(-)

diff --git a/gdb/aix-thread.c b/gdb/aix-thread.c
index aa14f8ff045..2355f87db54 100644
--- a/gdb/aix-thread.c
+++ b/gdb/aix-thread.c
@@ -44,6 +44,7 @@
 #include "inferior.h"
 #include "regcache.h"
 #include "cli/cli-cmds.h"
+#include "maint.h"
 #include "ppc-tdep.h"
 #include "observable.h"
 #include "objfiles.h"
@@ -706,30 +707,35 @@ state2str (pthdb_state_t state)
     }
 }
 
+/* Number of thread descriptors to fetch per getthrds() call.
+   Fetching in batches reduces syscall overhead to one call per
+   GETTHRDS_BATCH threads.  */
+constexpr int GETTHRDS_BATCH = 64;
+
 /* Search through the list of all kernel threads for the thread
-   that has stopped on a SIGTRAP signal, and return its TID.
+   that has stopped on a signal, and return its TID.
    Return 0 if none found.  */
 
 static pthdb_tid_t
 get_signaled_thread (int pid)
 {
-  struct thrdsinfo64 thrinf;
+  struct thrdsinfo64 thrinf[GETTHRDS_BATCH];
   tid_t ktid = 0;
 
-  while (1)
-    {
-      if (getthrds (pid, &thrinf,
-		    sizeof (thrinf), &ktid, 1) != 1)
-	break;
+  scoped_time_it time_it ("get_signaled_thread");
 
+  for (int count;
+       (count = getthrds (pid, thrinf, sizeof (thrinf[0]), &ktid,
+			  GETTHRDS_BATCH)) > 0;)
+    {
       /* We also need to keep in mind Trap and interrupt or any
 	 signal that needs to be handled in pd_update ().  */
-
-      if (thrinf.ti_cursig)
-	return thrinf.ti_tid;
+      for (int i = 0; i < count; i++)
+	if (thrinf[i].ti_cursig)
+	  return thrinf[i].ti_tid;
     }
 
-  /* Didn't find any thread stopped on a SIGTRAP signal.  */
+  /* Didn't find any thread stopped on a signal.  */
   return 0;
 }
 
@@ -763,18 +769,30 @@ sync_threadlists (pid_t pid)
 
   for (cmd = PTHDB_LIST_FIRST;; cmd = PTHDB_LIST_NEXT)
     {
-      status = pthdb_pthread (data->pd_session, &pdtid, cmd);
+      {
+	scoped_time_it time_it ("pthdb_pthread");
+	status = pthdb_pthread (data->pd_session, &pdtid, cmd);
+      }
       if (status != PTHDB_SUCCESS || pdtid == PTHDB_INVALID_PTHREAD)
 	break;
 
-      status = pthdb_pthread_ptid (data->pd_session, pdtid, &pthid);
+      {
+	scoped_time_it time_it ("pthdb_pthread_ptid");
+	status = pthdb_pthread_ptid (data->pd_session, pdtid, &pthid);
+      }
       if (status != PTHDB_SUCCESS || pthid == PTHDB_INVALID_PTID)
 	continue;
 
-      status = pthdb_pthread_tid (data->pd_session, pdtid, &tid);
+      {
+	scoped_time_it time_it ("pthdb_pthread_tid");
+	status = pthdb_pthread_tid (data->pd_session, pdtid, &tid);
+      }
       ptid_t ptid (pid, tid, pthid);
 
-      status = pthdb_pthread_state (data->pd_session, pdtid, &state);
+      {
+	scoped_time_it time_it ("pthdb_pthread_state");
+	status = pthdb_pthread_state (data->pd_session, pdtid, &state);
+      }
       in_queue_threads.insert (pdtid);
 
       /* If this thread has reported and exited, do not add it again.  */
@@ -853,7 +871,12 @@ pd_update (pid_t pid)
   if (!data->pd_active)
     return ptid_t (pid);
 
-  status = pthdb_session_update (data->pd_session);
+  scoped_time_it total_time_it ("pd_update total");
+
+  {
+    scoped_time_it time_it ("pthdb_session_update");
+    status = pthdb_session_update (data->pd_session);
+  }
   if (status != PTHDB_SUCCESS)
     return ptid_t (pid);
 
@@ -861,7 +884,10 @@ pd_update (pid_t pid)
      and not core files since list of threads never change for core
      files.  */
   if (target_has_execution ())
-    sync_threadlists (pid);
+    {
+      scoped_time_it time_it ("sync_threadlists");
+      sync_threadlists (pid);
+    }
 
   /* Define "current thread" as one that just received a trap signal.  */
 
-- 
2.51.2

Re: [PATCH 2/3][RFC] Batch getthrds() calls in get_signaled_thread

[Simon Marchi]

On 2026-09-16 07:46, Aditya Vidyadhar Kamath wrote:
> From: Aditya Vidyadhar Kamath <aditya.kamath1@ibm.com>
> 
> The original code called getthrds() with a count of 1, fetching one
> kernel thread descriptor per syscall.  With N threads in the process
> that costs N syscalls per stop just to find the signaled thread.
> 
> Change to fetch GETTHRDS_BATCH (64) descriptors per call and scan
> the batch in a loop.  This reduces the syscall count from O(N) to
> O(N/64).
> 
> Also done the benchmark here who results are pasted below. To analyse
> where time is being spent. For now, this patch request for comment/draft and will
> remove the same and this paragraph from the commit message after
> feedback.
> 
> ==========
> Benchmark C code
> // cat ../bench_next.c
> /* Spawns NUM_THREADS worker threads that spin, then steps through
>    10 assignments in do_steps() so GDB can time each next.
> 
>    Also tests two scenarios from the bug report:
>      1. "next" over a call that spawns a thread (spawn_thread_here()).
>      2. A background thread exits while "next" is in progress
>         (exit_thread_here()).  */
> #include <stdio.h>
> #include <stdlib.h>
> #include <pthread.h>
> #include <unistd.h>
> 
> #ifndef NUM_THREADS
> #define NUM_THREADS 3
> #endif
> 
> static volatile int keep_running = 1;
> static volatile int extra_thread_started = 0;
> 
> static void *
> worker (void *arg)
> {
>   while (keep_running)
>     sched_yield ();
>   return NULL;
> }
> 
> /* Thread that exits on its own after signalling it has started.  */
> static void *
> short_lived (void *arg)
> {
>   extra_thread_started = 1;
>   return NULL;
> }
> 
> /* GDB will "next" over this call -- a new thread is spawned inside.  */
> static void
> spawn_thread_here (void)
> {
>   pthread_t t;
>   pthread_create (&t, NULL, short_lived, NULL);
>   pthread_join (t, NULL);
> }
> 
> /* GDB will "next" over this call -- the short-lived background thread
>    is in the process of exiting while we step.  */
> static void
> exit_thread_here (void)
> {
>   /* Just a visible marker so GDB can stop here.  */
>   volatile int x = 42;
>   (void)x;
> }
> 
> void
> do_steps (void)
> {
>   volatile int a = 1;
>   volatile int b = 2;
>   volatile int c = a + b;
>   volatile int d = c * 2;
>   volatile int e = d - a;
>   volatile int f = e + c;
>   volatile int g = f / 2;
>   volatile int h = g + 1;
>   volatile int i2 = h * h;
>   volatile int j = i2 - b;
>   (void)j;
> }
> 
> int
> main (void)
> {
>   pthread_t threads[NUM_THREADS];
>   pthread_t bg;
>   int i;
> 
>   /* Start background spinning threads.  */
>   for (i = 0; i < NUM_THREADS; i++)
>     pthread_create (&threads[i], NULL, worker, NULL);
> 
>   /* Next over a call that spawns a thread */
>   spawn_thread_here ();
> 
>   /* Next while a background thread is exiting
>      Launch a thread that exits quickly; "next" through exit_thread_here()
>      while that exit is in flight.  */
>   pthread_create (&bg, NULL, short_lived, NULL);
>   exit_thread_here ();
>   pthread_join (bg, NULL);
> 
>   do_steps ();
> 
>   keep_running = 0;
>   for (i = 0; i < NUM_THREADS; i++)
>     pthread_join (threads[i], NULL);
> 
>   return 0;
> }

I would expect the benchmark program to have a ton of
threads, to show what a worst case scenario of get_signaled_thread
looks like.  This program appears to spawn 3 threads max, that means
before the patch you'd get 3-4 getthrds calls, and after the patch you'd
get 1.  I suppose it would be more interesting to see what happens when
you have e.g. 1000 threads, in which case the number of getthrds calls
goes from 1000 to 16.

Also, since there is an early return in get_signaled_thread, if you want
to show the worst case, you'd need to set things up such that the
signaled thread is in the last batch, ideally the last thread (assuming
the order in which the kernel returns threads is predictable).

> ============
> Benchmark GDB code:
> 
> // cat ../bench_next.gdb 
> set pagination off
> set confirm off
> maintenance set per-command time on
> 
> # Scenario 1: next over a call that spawns a thread
> break spawn_thread_here
> run
> next
> 
> # Scenario 2: next while a background thread is exiting
> break exit_thread_here
> continue
> next
> 
> break do_steps
> continue
> 
> next
> next
> next
> next
> next
> next
> next
> next
> next
> next
> 
> quit
> 
> ===============
> # cd /current_gdb && time /current_gdb/binutils-gdb/gdb/gdb -batch -x bench_next.gdb ./bench_next_bin 2>&1 | grep -E "^(real|user|sys)|volatile int
> |pd_update total|sync_thread|pthdb_session"
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "pd_update total": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "pd_update total": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.146, user 0.081, sys 0.085, user+sys 0.166, 14.5 % CPU
> Time for "pd_update total": wall 1.146, user 0.081, sys 0.085, user+sys 0.166, 14.5 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.025, user 0.092, sys 0.092, user+sys 0.184, 18.0 % CPU
> Time for "pd_update total": wall 1.025, user 0.092, sys 0.092, user+sys 0.184, 18.0 % CPU
> 25        volatile int a = 1;
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 0.962, user 0.098, sys 0.095, user+sys 0.193, 20.1 % CPU
> Time for "pd_update total": wall 0.962, user 0.098, sys 0.095, user+sys 0.193, 20.1 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 0.965, user 0.098, sys 0.094, user+sys 0.192, 19.9 % CPU
> Time for "pd_update total": wall 0.965, user 0.098, sys 0.094, user+sys 0.192, 19.9 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 0.954, user 0.098, sys 0.094, user+sys 0.192, 20.1 % CPU
> Time for "pd_update total": wall 0.954, user 0.098, sys 0.094, user+sys 0.192, 20.1 % CPU
> 26        volatile int b = 2;
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 0.949, user 0.098, sys 0.094, user+sys 0.192, 20.2 % CPU
> Time for "pd_update total": wall 0.949, user 0.098, sys 0.094, user+sys 0.192, 20.2 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 0.957, user 0.098, sys 0.094, user+sys 0.192, 20.1 % CPU
> Time for "pd_update total": wall 0.957, user 0.098, sys 0.094, user+sys 0.192, 20.1 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.033, user 0.101, sys 0.098, user+sys 0.199, 19.3 % CPU
> Time for "pd_update total": wall 1.033, user 0.101, sys 0.098, user+sys 0.199, 19.3 % CPU
> 27        volatile int c = a + b;
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.031, user 0.101, sys 0.098, user+sys 0.199, 19.3 % CPU
> Time for "pd_update total": wall 1.031, user 0.101, sys 0.098, user+sys 0.199, 19.3 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.028, user 0.100, sys 0.098, user+sys 0.198, 19.3 % CPU
> Time for "pd_update total": wall 1.028, user 0.100, sys 0.098, user+sys 0.198, 19.3 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 0.996, user 0.100, sys 0.098, user+sys 0.198, 19.9 % CPU
> Time for "pd_update total": wall 0.996, user 0.100, sys 0.098, user+sys 0.198, 19.9 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.000, user 0.098, sys 0.098, user+sys 0.196, 19.6 % CPU
> Time for "pd_update total": wall 1.001, user 0.098, sys 0.098, user+sys 0.196, 19.6 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 0.998, user 0.098, sys 0.098, user+sys 0.196, 19.6 % CPU
> Time for "pd_update total": wall 0.998, user 0.098, sys 0.098, user+sys 0.196, 19.6 % CPU
> 28        volatile int d = c * 2;
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 0.999, user 0.098, sys 0.098, user+sys 0.196, 19.6 % CPU
> Time for "pd_update total": wall 0.999, user 0.098, sys 0.098, user+sys 0.196, 19.6 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.001, user 0.099, sys 0.098, user+sys 0.197, 19.7 % CPU
> Time for "pd_update total": wall 1.001, user 0.099, sys 0.098, user+sys 0.197, 19.7 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 0.969, user 0.100, sys 0.095, user+sys 0.195, 20.1 % CPU
> Time for "pd_update total": wall 0.969, user 0.100, sys 0.095, user+sys 0.195, 20.1 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 0.980, user 0.099, sys 0.096, user+sys 0.195, 19.9 % CPU
> Time for "pd_update total": wall 0.980, user 0.099, sys 0.096, user+sys 0.195, 19.9 % CPU
> 29        volatile int e = d - a;
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 0.984, user 0.100, sys 0.096, user+sys 0.196, 19.9 % CPU
> Time for "pd_update total": wall 0.984, user 0.100, sys 0.096, user+sys 0.196, 19.9 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.094, user 0.098, sys 0.094, user+sys 0.192, 17.6 % CPU
> Time for "pd_update total": wall 1.094, user 0.099, sys 0.094, user+sys 0.193, 17.6 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.019, user 0.106, sys 0.097, user+sys 0.203, 19.9 % CPU
> Time for "pd_update total": wall 1.019, user 0.106, sys 0.097, user+sys 0.203, 19.9 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.002, user 0.106, sys 0.096, user+sys 0.202, 20.2 % CPU
> Time for "pd_update total": wall 1.002, user 0.106, sys 0.096, user+sys 0.202, 20.2 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.003, user 0.105, sys 0.096, user+sys 0.201, 20.0 % CPU
> Time for "pd_update total": wall 1.003, user 0.105, sys 0.096, user+sys 0.201, 20.0 % CPU
> 30        volatile int f = e + c;
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.011, user 0.106, sys 0.097, user+sys 0.203, 20.1 % CPU
> Time for "pd_update total": wall 1.011, user 0.106, sys 0.097, user+sys 0.203, 20.1 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.010, user 0.107, sys 0.096, user+sys 0.203, 20.1 % CPU
> Time for "pd_update total": wall 1.010, user 0.107, sys 0.096, user+sys 0.203, 20.1 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.026, user 0.110, sys 0.096, user+sys 0.206, 20.1 % CPU
> Time for "pd_update total": wall 1.026, user 0.110, sys 0.096, user+sys 0.206, 20.1 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.200, user 0.090, sys 0.085, user+sys 0.175, 14.6 % CPU
> Time for "pd_update total": wall 1.201, user 0.090, sys 0.085, user+sys 0.175, 14.6 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.212, user 0.090, sys 0.091, user+sys 0.181, 14.9 % CPU
> Time for "pd_update total": wall 1.212, user 0.090, sys 0.091, user+sys 0.181, 14.9 % CPU
> 31        volatile int g = f / 2;
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 0.993, user 0.103, sys 0.096, user+sys 0.199, 20.0 % CPU
> Time for "pd_update total": wall 0.993, user 0.103, sys 0.096, user+sys 0.199, 20.0 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.258, user 0.079, sys 0.082, user+sys 0.161, 12.8 % CPU
> Time for "pd_update total": wall 1.258, user 0.079, sys 0.082, user+sys 0.161, 12.8 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.255, user 0.080, sys 0.081, user+sys 0.161, 12.8 % CPU
> Time for "pd_update total": wall 1.255, user 0.080, sys 0.081, user+sys 0.161, 12.8 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.259, user 0.080, sys 0.081, user+sys 0.161, 12.8 % CPU
> Time for "pd_update total": wall 1.259, user 0.080, sys 0.081, user+sys 0.161, 12.8 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.264, user 0.080, sys 0.082, user+sys 0.162, 12.8 % CPU
> Time for "pd_update total": wall 1.265, user 0.080, sys 0.082, user+sys 0.162, 12.8 % CPU
> 32        volatile int h = g + 1;
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.261, user 0.079, sys 0.083, user+sys 0.162, 12.8 % CPU
> Time for "pd_update total": wall 1.261, user 0.079, sys 0.083, user+sys 0.162, 12.8 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.232, user 0.076, sys 0.081, user+sys 0.157, 12.7 % CPU
> Time for "pd_update total": wall 1.232, user 0.076, sys 0.081, user+sys 0.157, 12.7 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.250, user 0.078, sys 0.083, user+sys 0.161, 12.9 % CPU
> Time for "pd_update total": wall 1.250, user 0.078, sys 0.083, user+sys 0.161, 12.9 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.258, user 0.080, sys 0.084, user+sys 0.164, 13.0 % CPU
> Time for "pd_update total": wall 1.258, user 0.080, sys 0.084, user+sys 0.164, 13.0 % CPU
> 33        volatile int i2 = h * h;
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.276, user 0.082, sys 0.082, user+sys 0.164, 12.9 % CPU
> Time for "pd_update total": wall 1.276, user 0.082, sys 0.082, user+sys 0.164, 12.9 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.271, user 0.083, sys 0.081, user+sys 0.164, 12.9 % CPU
> Time for "pd_update total": wall 1.271, user 0.083, sys 0.081, user+sys 0.164, 12.9 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.280, user 0.079, sys 0.086, user+sys 0.165, 12.9 % CPU
> Time for "pd_update total": wall 1.280, user 0.079, sys 0.086, user+sys 0.165, 12.9 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.211, user 0.075, sys 0.081, user+sys 0.156, 12.9 % CPU
> Time for "pd_update total": wall 1.211, user 0.075, sys 0.081, user+sys 0.156, 12.9 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.264, user 0.077, sys 0.086, user+sys 0.163, 12.9 % CPU
> Time for "pd_update total": wall 1.264, user 0.077, sys 0.086, user+sys 0.163, 12.9 % CPU
> 34        volatile int j = i2 - b;
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.240, user 0.078, sys 0.081, user+sys 0.159, 12.8 % CPU
> Time for "pd_update total": wall 1.240, user 0.078, sys 0.081, user+sys 0.159, 12.8 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.256, user 0.078, sys 0.082, user+sys 0.160, 12.7 % CPU
> Time for "pd_update total": wall 1.257, user 0.078, sys 0.082, user+sys 0.160, 12.7 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.267, user 0.078, sys 0.082, user+sys 0.160, 12.6 % CPU
> Time for "pd_update total": wall 1.267, user 0.078, sys 0.082, user+sys 0.160, 12.6 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.254, user 0.079, sys 0.081, user+sys 0.160, 12.8 % CPU
> Time for "pd_update total": wall 1.255, user 0.079, sys 0.081, user+sys 0.160, 12.7 % CPU
> Time for "pthdb_session_update": wall 0.000, user 0.000, sys 0.000, user+sys 0.000, NaNQ % CPU
> Time for "sync_threadlists": wall 1.257, user 0.078, sys 0.082, user+sys 0.160, 12.7 % CPU
> Time for "pd_update total": wall 1.257, user 0.078, sys 0.082, user+sys 0.160, 12.7 % CPU

I don't see any "Time for" line for get_signaled_thread, which is the
function you are optimizing in this patch... am I missing something?

Simon