We have a thread with priority 255 loops forever to receive pulse event and post semaphore to notify work thread. The 
looping pseudo code lists below.

while (true) {
    MsgReceivePulse();
    ret = seg_getvalue(sem, value);
    if (!((0 == ret) && (value > 0))) {
        sem_post(sem);
    }
}

The running time of the above code is abort 10us. But we found the running time was nearly 3ms sometimes, when we used 
QNX traceevent to catch some events. See the attachment.

This thread's events list below:
696441, 613ms 438us, 0x265209E82, 7, safety Patie#ty_2TimTh, Process and Thread, Running, policy_name FIFO pid 9424945 
tid 34 priority 255 policy 1 partition 7 sched_flags 0
696442, 613ms 438us, 0x26520A3C0, 7, safety Patie#ty_2TimTh, Communication, Receive Pulse, scoid 0x10005 pid 9424945 
process safety
696443, 613ms 438us, 0x26520A4FA, 7, safety Patie#ty_2TimTh, Kernel Calls, MsgReceivePulsev (64) Exit, rcvid 0x0 rmsg0 
0x0 rmsg1 0x0 rmsg2 0x0 info->nd 0 info->srcnd 0 info->pid 0 info->tid 0 info->chid 0 info->scoid 0 info->msglen 0 info-
>srcmsglen 0 info->dstmsglen 0 info->priority 0 info->flags 0 info->reserved 0
696885, 613ms 762us, 0x265306EF0, 7, safety Patie#ty_2TimTh, System, IPI (64), ipicmd 0x20 CHECK_INTR ipi 0x257f781317 
tid 34 pid 9424945 process safety
696903, 613ms 777us, 0x2653128D0, 7, safety Patie#ty_2TimTh, Process and Thread, Running, policy_name FIFO pid 9424945 
tid 34 priority 255 policy 1 partition 7 sched_flags 0
698114, 614ms 762us, 0x265612411, 7, safety Patie#ty_2TimTh, System, IPI (64), ipicmd 0x20 CHECK_INTR ipi 0x257f781317 
tid 34 pid 9424945 process safety
698134, 614ms 781us, 0x265620D94, 7, safety Patie#ty_2TimTh, Process and Thread, Running, policy_name FIFO pid 9424945 
tid 34 priority 255 policy 1 partition 7 sched_flags 0
698691, 615ms 181us, 0x26575868B, 7, safety Patie#ty_2TimTh, Kernel Calls, SyncSemPost (64) Enter, sync_ptr 0x1e3b814a7c
 count 0 owner 0xfffffffc
698693, 615ms 183us, 0x26575A3DE, 7, safety Patie#ty_2TimTh, Kernel Calls, SyncSemPost Exit, ret_val 0x0
699423, 615ms 763us, 0x26591DC0B, 7, safety Patie#ty_2TimTh, System, IPI (64), ipicmd 0x20 CHECK_INTR ipi 0x257f7810ed 
tid 34 pid 9424945 process safety
699451, 615ms 786us, 0x26592FE48, 7, safety Patie#ty_2TimTh, Process and Thread, Running, policy_name FIFO pid 9424945 
tid 34 priority 255 policy 1 partition 7 sched_flags 0
699846, 616ms 99us, 0x265A23AE1, 7, safety Patie#ty_2TimTh, Kernel Calls, MsgReceivePulsev (64) Enter, chid 0x9 rparts 1

 
From the attachment, we found this thread is always running and there are many Control events, such as.
696454, 613ms 446us, 0x265210A7F, 1, , Control Events, Buffer, sequence_number 6216609 event_count 0
696455, 613ms 447us, 0x2652112FE, 1, , Control Events, Time, msb 0x174d lsb 0x652112fe
696459, 613ms 451us, 0x265214063, 9, , Control Events, Time, msb 0x174d lsb 0x65214063

Our system has 12 CPU core, and this thread is running on CPU 7/8/9 , the aps partition is 20% (less than 3/12=25%).

So why this thread runs so long? Is it because traceevent? Or do aps partition exceed 20%?
Thanks very much.