Thread context switching.

Since I'm working with some real-time applications I'm interested in the thread switching mechanisms of Java ME.
So my question is when does the context switching take place?
If a thread A changes the priority of thread B to higher than its own priority will thread B preempt A instantly or is it neccessary to cause the current thread to interrupt?
How do same priority threads preempt each other?
What priority is given to the garbage collector?
Mikael

Hi Mikael,
I'm not a ME VM expert, so do not have the deep knowledge. But from what I know:
- thread switching can take place at any point during execution of java code. That's in interpreted mode. In compiled mode there are some limitations on when it can happen but it still could be described with the words "at almost any point"
- the scheduler is called to have "fair" policy. That is - all threads get their share of CPU time, the priority defines how big this share is
- the current thread get preempted when it's exhaused it's currently allocated share of CPU time. next thread is scheduled at that point
- I don't know whether rescheduling happens when thread priority is modified
- GC is not a thread and it does not have any priority. It gets invoked according to internal logic which does not depend on logic of the scheduler but rather on heap parameters (configuration and usage)
Regards,
Andrey

Similar Messages

High thread context switching for java web application

We have been load testing our java web application and observe high cpu usage with 50 users (which doesn't seem practical). The CPU shoots up above 80%. While profiling it with java flight recording (JFR) we see that the context switch rate is 8400 per second (as seen in the Hot threads tab on java mission control). Analyzing the hot threads in jfr, it seems the cpu usage is distributed across the application threads with each thread using less than 3% cpu.
Increasing the user load to 100, 150 or 200 users we see the cpu shooting up above 90%, the throughput (transactions per second) remaining constant (as seen for 50 users load) while the response time crosses the acceptable threshold values (3 sec). Decreasing the user load to 20 users shows the cpu usage averages out to be above 55%. It certainly isn't true that the application threads are using up the cpu since our application is not a CPU bound application. The Hot Packages tab under Code tab group confirms this by showing that most of the time the application spends in is executing database queries.
We use glassfish 3.1.2.2 as our application server where the max thread pool is configured to be of 100. Oracle Linux Server release 6.4 is our operating system with linux kernel version as 2.6.39-400.214.4.el6uek.x86_64. I tried executing linux commands namely "watch -n0.5 pidstat -w -I -p " and "watch -n.5 grep ctxt /proc//status" to see the voluntary and involuntary thread context switching at OS level but they don't give any results.
Suspecting that high context switching could be causing the cpu to shoot up, do you have guidelines on what could be done to confirm that thread context switching is the cause of high cpu and what are there ways to tune the jvm or the application if that's the cause?
Thanks!

Kelum -
We just saw this issue today for the first time. Have you been able to find a cause?
We upgraded our 32bit Windows operating systems this weekend to use the /3GB flag. Since then, we have seen that our servers have ample heap space, but are dangerously low in PTE memory.
But when we've been diagnosing the state of the server that produced this error (we run 2 nodes on 3 different computers; only 1 produced this error; the other 5 are working normally), everything looked fine. The server was reporting sufficient PTE availablility, plenty of heap space, and around 172 threads (we expect to be able to run many more than that).
When we restarted the node, it came up fine and everything appeared to be working normally.
So I'm looking for any clue as to the root cause, and what kind of resolution to explore. Any clues or pointers would be greatly appreciated.
Paul Christmann

Context switching / Threads

Hello !
The following program is for 3 Threads which do context switching.
Often we get '0' zero for the low priority thread when we run this program.
MY QUESTION IS WHY DO WE GET ZERO ?
As far as my understanding is concerned; even if preemptive multitasking is done by the threads,
the low priority thread should have run through few iterations and thus giving some value other
than zero '0'.
Secondly most of the time we get negative values for the high priority thread. Why is that ?
Is it because of the fact that volatile sets the variable 'running' to some different value ?
The speed of my processor is 1.5GHz.
GOD BLESS YOU.
NADEEM.
// Demonstrates threads priorities.
class Clicker implements Runnable {
int click = 0;
String name ;
Thread t ;
private boolean running = true ;
public Clicker(String tname, int p) {
   name = tname ;
   t = new Thread(this, name);
   t.setPriority(p);
   System.out.println("Current Thread is " + t + " " + t.getPriority());
public void start() {
   t.start();
public void run() {
   while (running) {
    click++;
public void stop() {
   running = false ;
class HLPriority {
public static void main(String args[]) {
   System.out.println("Active Count : " + Thread.activeCount());
   Thread.currentThread().setPriority(Thread.MAX_PRIORITY);
   Clicker hi = new Clicker("Hi", Thread.NORM_PRIORITY + 2);
   Clicker lo = new Clicker("Lo", Thread.NORM_PRIORITY - 2);
   System.out.println("Active Count : " + Thread.activeCount());
   lo.start();
   hi.start();
   try {
     System.out.println("Sleeping Thread : " + Thread.currentThread());
     Thread.sleep(10000);
   } catch (InterruptedException e) {
     System.out.println("Main Thread interrupted : " + e);
    hi.stop();
    lo.stop();
   try {
    hi.t.join();
    lo.t.join();
   } catch (InterruptedException e) {
     System.out.println("Interrupted Exception caught : " + e);
   System.out.println("Low priority thread : " + lo.click);
   System.out.println("Hi priority thread : " + hi.click);
}

Hello !
The following program is for 3 Threads which do
context switching.
Often we get '0' zero for the low priority thread when
we run this program.
MY QUESTION IS WHY DO WE GET ZERO ?Presumably because the low priority thread gets no CPU time.
>
As far as my understanding is concerned; even if
preemptive multitasking is done by the threads,
the low priority thread should have run through few
iterations and thus giving some value other
than zero '0'.You can't make ANY assumptions about when or how much CPU time a given thread will get. Why don't you let your main thread sleep longer--a minute or 5 or ten--and see if LO gets some cycles then.
Secondly most of the time we get negative values for
the high priority thread. Why is that ?count = Integer.MAX_VALUE;
count++; // --> Integer.MIN_VALUE (-2^31)
I guess 10 seconds is enough time for a thread in a tight loop to count to 2 billion.
Is it because of the fact that volatile sets the
variable 'running' to some different value ?Volatile does nothing of the sort, and, in any case, isn't even in your code.
GOD BLESS YOU.I didn't sneeze.

Forcing context switching between Labview threads

Hello,
I have 2 threads interfacing with 2 serial ports on Labview ( Lets say Thread1 , is responsible for polling COM1 and Thread2 for polling COM2 ).
By using occurrences, Thread1 executes before Thread2 and context switching is done according to Labview between the threads. Now, I'm looking to do the following :
As soon as a specific block of code is executed in Thread1, I need Labview to force context switching to execute Thread2. I want to do this in order to synchronize the data received from the 2 threads.
1- So, can this be done, or are there other ways for synchronization ?
2- What if I want Labview to run as a real-time process or a higher priority thread on Windows XP in order to emulate the real-time effect on a Windows XP and not get any delays ? ( I already changed the process priority from the Process explorer, but there seems no effect. I also changed the Labview threads priority to real-time)
3- Are there better approached for (1) & (2) ?
Thank you,
Walid F. Abdelfatah

wfarid wrote:
I already used occurrences for Thread1 to execute before Thread2. Does occurrences guarantee that Labview will switch the context to Thread2, as soon as the context is fired ?
-- Walid
NO!
LV depends on the OS to schedule threads. If you are sticking with Non-RT then you would be better off getting thread one to do the work while it has teh CPU.
Ben
Ben Rayner
I am currently active on.. MainStream Preppers
Rayner's Ridge is under construction

Context switching.

Theory: Firewalls essentially partition the Java Card platform’s object system into separate
protected object spaces called contexts. The
firewall is the boundary between one context and another. The Java Card RE shall
allocate and manage a context for each Java API package containing applets1. All
applet instances within a single Java API package share the same context. There is
no firewall between individual applet instances within the same package. That is, an
applet instance can freely access objects belonging to another applet instance that
resides in the same package.
That is the theory. What happens in my case. My Java Card project contains three packages and in one there is one Java Card applet. Splitting to three package was necessary because the application is large. What about the object instances from other packages. Are they assigned to other context and what happens when java card applet instance access these objects? Is context switching is happening?

Patrick,
Don't worry about context switching. Build a good load test. Run it against
a "best guess" number of exec threads. Increase the number of threads and
run again. If overall throughput drops, then decrease the number of threads
and run again. Start with coarse increments (5 threads?) and work from there
until you get the best setting.
Peace,
Cameron Purdy
Tangosol, Inc.
http://www.tangosol.com/coherence.jsp
Tangosol Coherence: Clustered Replicated Cache for Weblogic
"Patrick Acheson" <[email protected]> wrote in message
news:3d5aae20$[email protected]..
>
In setting the executeThreadCount variable for Weblogic 5.10, if thevariable is
too high there will be a lot of context switching going on. What wouldconstitute
a lot of context switching as opposed to what would be a normal orexpected amount?
Our executeThreadCount is set at 100 and we have 4 CPUs. In Perfmon,about 10%
of the threads show 1 to 2 context switches per second.

Dynamically-Launched VIs and Context Switching

I'm working on a project that dynamically calls instances of a reentrant VI, which I assume is a pretty common practice. Everything works pretty well, until the number of calls to this dynamic VI gets pretty large--on the order of 1000 or more--at which point, we begin to see performance degradation. My guess is that we are taking hits due to context switching, since the number of threads far exceeds the number of logical processor cores available.
A little more background:
The dynamic VIs being called effectively run as daemons, each running a while loop and waiting on a dedicated input queue to receive data and save it to disk. All are stopped via a globally shared stop notifier (passed as a ControlValue.Set method argument at launch). Each is waiting on its respective queue with a 1 second timeout so that the stop notifier can be polled. Under normal operating conditions, each one will run at some rate between 0.1Hz and 25Hz (the various rates are a large driving factor for separating them and needing to spawn them dynamically).
So, this leads me to the following questions:
Am I correct that the context switching is the likely culprit in the performance degradation?
If so, is there a fundamental difference in how LabVIEW handles multithreading with dynamic VI calls versus explicitly drawing separate while loops on a block diagram, or dropping multiple instances of a reentrant VI directly on the block diagram?
Is it likely that reducing the number of dynamic clones to equal the number of available processor cores would improve performance? (the scope of each clone would grow, as it would have to maintain the state information that was original distributed across multiple clones)
I realize that this question is pretty vague without concrete examples, but I'm hoping someone (AQ? Ben? Any of you NI gurus?) out there could provide some general insight into what's going on under the hood without needing to get too specific.

TurboPhil wrote:
Each is waiting on its respective queue with a 1 second timeout so that the stop notifier can be polled.
There is one relatively easy fix you can probably make here - set the timeout to -1 and destroy the queues to stop the loop (destroying the queue will output an error from the wait primitive). This should at least stop all the code from running all the time, although I'm still not sure how the threading of the different VIs will play with each other. This might be an issue if the queue is only created in the VI, but I'm assuming it isn't.
Try to take over the world!

Server's context switch rate ?

what does it mean by server's context switch rate

I've never heard the term. "Context switching" refers to things like pausing one thread or application to give time to another.
So I'd guess that "server's context switch rate" refers to how frequently the server in question alternates CPU time from one thread or app to the next.

Problem with behavior of the Thread Context ClassLoader on OC4J 10.1.3.2.0

Hi,
We run into different behavior of the Thread Context ClassLoader on OC4J 10.1.3.1.0 (which is OK) and OC4J 10.1.3.2.0 (here we have problems). It appeared in the Threads directly and indirectly (e.g. through Timer) created by our application. Threads, created by OC4j (e.g. the Thread, where our start-up servlet are running) are OK.
Thread Context ClassLoader (could be received from Thread.currentThread().getContextClassLoader()) on OC4J 10.1.3.1.0 in all threads is the same ClassLoader instance as the current ClassLoader (could be received from getClass().getClassLoader()). Actually the same is true for JBoss, WebLogic and WebSphere.
Thread Context ClassLoader on OC4J 10.1.3.2.0 is NOT the same ClassLoader instance as the current ClassLoader, it is even the instance of the different Class. This ClassLoader does not see classes, available at application level, which causes multiple problems.
Is it possible to configure OC4J 10.1.3.2.0 to behave similar to OC4J 10.1.3.1.0?
We tried to use “–userThreads”, but it did not help.
To investigate this problem we created a simple Web Application, which has Start-up Servlet, creates one thread and one Timer Task. From each point it is checking if Thread Context ClassLoader is the same ClassLoader instance as the current ClassLoader and printing Class Loaders if they are different. In addition it checks if correct parser accessed (should be xerces, which is in the Web App classpath and for which we added a shared library). Below you could see what we received on OC4J 10.1.3.1.0 (which is OK) and OC4J 10.1.3.2.0:
E:\OracleAS_1\j2ee\home>java -Xms128m -Xmx1024m -jar oc4j.jar -userThreads
07/04/30 10:33:00 Oracle Containers for J2EE 10g (10.1.3.1.0) initialized
07/04/30 10:39:31 ------In StartupServlet Init-----
07/04/30 10:39:31 +++++++ ++++++++++++++++++++++++++++++++++++ ++++++++
07/04/30 10:39:31 +++++++ Executing Code in the StrtServlet... ++++++++
07/04/30 10:39:31 +++++++ (this Thread created by App Server) +++++++++
07/04/30 10:39:31 ClassLoaders: fromClass = getClass().getClassLoader() = <TestL
oaders.web.TestLoaders:0.0.0>*******<class oracle.classloader.PolicyClassLoader
07/04/30 10:39:31 fromThread =Thread.currentThread().getContextCla
ssLoader() = <TestLoaders.web.TestLoaders:0.0.0>*******<class oracle.classloader
.PolicyClassLoader
07/04/30 10:39:31 ******* THE SAME CLASSLOADER USED IN BOTH PLACES!!! *******
07/04/30 10:39:31 DocumentBuilderFactory = <org.apache.xerces.jaxp.DocumentBuild
erFactoryImpl@1908ba7>*******
07/04/30 10:39:31 C O R R E C T !!!
07/04/30 10:39:31 -CWD- =E:\OracleAS_1\j2ee\home; uer directory = E:\OracleAS_1\
j2ee\home
07/04/30 10:39:31 Constructing the Timer Task...
07/04/30 10:39:31 Starting some Timer...
07/04/30 10:39:31 Some Timer is started ...
07/04/30 10:39:31 ------All actions done Successfully--------
07/04/30 10:39:41 +++++++ +++++++++++++++++++++++++++++++++++ ++++++++
07/04/30 10:39:41 +++++++ Executing Code in the new Thread... ++++++++
07/04/30 10:39:41 ClassLoaders: fromClass = getClass().getClassLoader() = <TestL
oaders.web.TestLoaders:0.0.0>*******<class oracle.classloader.PolicyClassLoader
07/04/30 10:39:41 fromThread =Thread.currentThread().getContextCla
ssLoader() = <TestLoaders.web.TestLoaders:0.0.0>*******<class oracle.classloader
.PolicyClassLoader
07/04/30 10:39:41 ******* THE SAME CLASSLOADER USED IN BOTH PLACES!!! *******
07/04/30 10:39:41 DocumentBuilderFactory = <org.apache.xerces.jaxp.DocumentBuild
erFactoryImpl@17e5e56>*******
07/04/30 10:39:41 C O R R E C T !!!
07/04/30 10:40:31 +++++++ +++++++++++++++++++++++++++++++++++ ++++++++
07/04/30 10:40:31 +++++++ Executing the Session Timer Task... ++++++++
07/04/30 10:40:31 ClassLoaders: fromClass = getClass().getClassLoader() = <TestL
oaders.web.TestLoaders:0.0.0>*******<class oracle.classloader.PolicyClassLoader
07/04/30 10:40:31 fromThread =Thread.currentThread().getContextCla
ssLoader() = <TestLoaders.web.TestLoaders:0.0.0>*******<class oracle.classloader
.PolicyClassLoader
07/04/30 10:40:31 ******* THE SAME CLASSLOADER USED IN BOTH PLACES!!! *******
07/04/30 10:40:31 DocumentBuilderFactory = <org.apache.xerces.jaxp.DocumentBuild
erFactoryImpl@1c6988d>*******
07/04/30 10:40:31 C O R R E C T !!!
07/04/30 10:44:01 Shutting down OC4J...
07/04/30 10:44:01
E:\OracleAS_1\j2ee\home>
C:\oc4j\j2ee\home>java -Xms128m -Xmx1024m -classpath .;oc4j.jar;C:\oc4j\j2ee\hom
e\properties -Xdebug -Xnoagent -Djava.compiler=NONE -Xrunjdwp:transport=dt_socke
t,address=3999,server=y,suspend=n -jar oc4j.jar -userThreads
07/05/01 21:14:33 ------In StartupServlet Init-----
07/05/01 21:14:33 +++++++ ++++++++++++++++++++++++++++++++++++ ++++++++
07/05/01 21:14:33 +++++++ Executing Code in the StrtServlet... ++++++++
07/05/01 21:14:33 +++++++ (this Thread created by App Server) +++++++++
07/05/01 21:14:33 ClassLoaders: fromClass = getClass().getClassLoader() = <TestL
oaders.web.TestLoaders:0.0.0>*******<class oracle.classloader.PolicyClassLoader
07/05/01 21:14:33 fromThread =Thread.currentThread().getContextCla
ssLoader() = <TestLoaders.web.TestLoaders:0.0.0>*******<class oracle.classloader
.PolicyClassLoader
07/05/01 21:14:33 ******* THE SAME CLASSLOADER USED IN BOTH PLACES!!! *******
07/05/01 21:14:33 DocumentBuilderFactory = <org.apache.xerces.jaxp.DocumentBuild
erFactoryImpl@723c42>*******
07/05/01 21:14:33 C O R R E C T !!!
07/05/01 21:14:33 -CWD- =C:\oc4j\j2ee\home; uer directory = C:\oc4j\j2ee\home
07/05/01 21:14:33 Constructing the Timer Task...
07/05/01 21:14:33 Starting some Timer...
07/05/01 21:14:33 Some Timer is started ...
07/05/01 21:14:33 ------All actions done Successfully--------
07/05/01 21:14:33 Oracle Containers for J2EE 10g (10.1.3.2.0) initialized
07/05/01 21:14:43 +++++++ +++++++++++++++++++++++++++++++++++ ++++++++
07/05/01 21:14:43 +++++++ Executing Code in the new Thread... ++++++++
07/05/01 21:14:43 ClassLoaders: fromClass = getClass().getClassLoader() = <TestL
oaders.web.TestLoaders:0.0.0>*******<class oracle.classloader.PolicyClassLoader
07/05/01 21:14:43 fromThread =Thread.currentThread().getContextCla
ssLoader() = <[ThreadContextLoader, current context: TestLoaders.web.TestLoaders
:0.0.0]>*******<class com.evermind.server.ApplicationContextClassLoader
07/05/01 21:14:43 ******************************************************
07/05/01 21:14:43 Chain of ClassLoaders from Class
07/05/01 21:14:43 --- TestLoaders.root:0.0.0 --- class oracle.classloader.Polic
yClassLoader
07/05/01 21:14:43 --- default.root:0.0.0 --- class oracle.classloader.PolicyCla
ssLoader
07/05/01 21:14:43 --- system.root:0.0.0 --- class oracle.classloader.PolicyClas
sLoader
07/05/01 21:14:43 --- oc4j:10.1.3 --- class oracle.classloader.PolicyClassLoade
r
07/05/01 21:14:43 --- api:1.4.0 --- class oracle.classloader.PolicyClassLoader
07/05/01 21:14:43 --- jre.extension:0.0.0 --- class oracle.classloader.PolicyCl
assLoader
07/05/01 21:14:43 --- jre.bootstrap:1.4.2_08 --- class oracle.classloader.Polic
yClassLoader
07/05/01 21:14:43 --- null ---
07/05/01 21:14:43 ********** End of Chain of ClassLoaders **************
07/05/01 21:14:43 ******************************************************
07/05/01 21:14:43 Chain of ClassLoaders from Thread
07/05/01 21:14:43 --- oc4j:10.1.3 --- class oracle.classloader.PolicyClassLoade
r
07/05/01 21:14:43 --- api:1.4.0 --- class oracle.classloader.PolicyClassLoader
07/05/01 21:14:43 --- jre.extension:0.0.0 --- class oracle.classloader.PolicyCl
assLoader
07/05/01 21:14:43 --- jre.bootstrap:1.4.2_08 --- class oracle.classloader.Polic
yClassLoader
07/05/01 21:14:43 --- null ---
07/05/01 21:14:43 ********** End of Chain of ClassLoaders **************
07/05/01 21:14:43 DocumentBuilderFactory = <oracle.xml.jaxp.JXDocumentBuilderFac
tory@74a138>*******
07/05/01 21:14:43 WRONG !!! Should be org.apache.xerces.jaxp.DocumentBuilderFact
oryImpl
07/05/01 21:15:33 +++++++ +++++++++++++++++++++++++++++++++++ ++++++++
07/05/01 21:15:33 +++++++ Executing the Session Timer Task... ++++++++
07/05/01 21:15:33 ClassLoaders: fromClass = getClass().getClassLoader() = <TestL
oaders.web.TestLoaders:0.0.0>*******<class oracle.classloader.PolicyClassLoader
07/05/01 21:15:33 fromThread =Thread.currentThread().getContextCla
ssLoader() = <[ThreadContextLoader, current context: TestLoaders.web.TestLoaders
:0.0.0]>*******<class com.evermind.server.ApplicationContextClassLoader
07/05/01 21:15:33 **************************************************
07/05/01 21:15:33 Chain of ClassLoaders from Class
07/05/01 21:15:33 --- TestLoaders.root:0.0.0 --- class oracle.classloader.Polic
yClassLoader
07/05/01 21:15:33 --- default.root:0.0.0 --- class oracle.classloader.PolicyCla
ssLoader
07/05/01 21:15:33 --- system.root:0.0.0 --- class oracle.classloader.PolicyClas
sLoader
07/05/01 21:15:33 --- oc4j:10.1.3 --- class oracle.classloader.PolicyClassLoade
r
07/05/01 21:15:33 --- api:1.4.0 --- class oracle.classloader.PolicyClassLoader
07/05/01 21:15:33 --- jre.extension:0.0.0 --- class oracle.classloader.PolicyCl
assLoader
07/05/01 21:15:33 --- jre.bootstrap:1.4.2_08 --- class oracle.classloader.Polic
yClassLoader
07/05/01 21:15:33 --- null ---
07/05/01 21:15:33 ********** End of Chain of ClassLoaders **********
07/05/01 21:15:33 **************************************************
07/05/01 21:15:33 Chain of ClassLoaders from Thread
07/05/01 21:15:33 --- oc4j:10.1.3 --- class oracle.classloader.PolicyClassLoade
r
07/05/01 21:15:33 --- api:1.4.0 --- class oracle.classloader.PolicyClassLoader
07/05/01 21:15:33 --- jre.extension:0.0.0 --- class oracle.classloader.PolicyCl
assLoader
07/05/01 21:15:33 --- jre.bootstrap:1.4.2_08 --- class oracle.classloader.Polic
yClassLoader
07/05/01 21:15:33 --- null ---
07/05/01 21:15:33 ********** End of Chain of ClassLoaders **********
07/05/01 21:15:33 DocumentBuilderFactory = <oracle.xml.jaxp.JXDocumentBuilderFac
tory@1d12d43>*******
07/05/01 21:15:33 WRONG !!! Should be org.apache.xerces.jaxp.DocumentBuilderFact
oryImpl
Any help will be greatly appreciated.
Thanks

Sorry for confusion, version of OC4J does not really meter, but jdk does. It happened that in the case where everything OK (OC4J 10.1.3.1.0) I accidentally used jdk 1.5. Now I tried both (OC4J 10.1.3.1.0 and OC4J 10.1.3.2.0) with j2sdk1.4.2_14 (besides I tried j2sdk1.4.2_09 and _08) and j2sdk1.5.0_06.
With JDK 1.5 everything is OK and with JDK 1.4 we have a problem, described in the previous posting. I believe now I should reformulate my initial question:
Is it possible to configure OC4J 10.1.3 running under j2sdk1.4.2_14 to force Thread Context ClassLoader to behave similar to OC4J 10.1.3 under j2sdk1.5.0_06? If yes, how we could do it?
Thanks

Decode Vs Case: context switching?

So I was told recently that among other reasons, CASE is "better" than Decode in SQL statements because Decode context switches to PL/SQL to perform the checks.
I can't find anything in the documentation to support this.
this site here:
http://www.dba-oracle.com/oracle_news/2005_11_23_case_decode_machinations.htm
mentions that one of the disadvantages of decode is that it's post-retrieval, but it also seems to mention that so is CASE.
anyone have any idea where someone may have got the "context switching" idea from?

have often wondered why you would use CASE in PL/SQL when it has IF THEN control structures. Yes, you could, however readability would suffer. But what is more important CASE has a form where expression is evaluated only once:
SQL> SET SERVEROUTPUT ON
SQL> DECLARE
2      X NUMBER;
3 BEGIN
4      PKG1.CNT := 2;
5      X := CASE PKG1.F1
6             WHEN 1 THEN 1
7             WHEN 2 THEN 2
8             WHEN 3 THEN 3
9           END;
10      DBMS_OUTPUT.PUT_LINE('X = ' || X);
11      DBMS_OUTPUT.PUT_LINE('PKG1.CNT = ' || PKG1.CNT);
12 END;
13 /
Call to PKG1.F1
X = 3
PKG1.CNT = 3
PL/SQL procedure successfully completed.
SQL> DECLARE
2      X NUMBER;
3 BEGIN
4      PKG1.CNT := 2;
5      IF PKG1.F1 = 1
6        THEN X := 1;
7      ELSIF PKG1.F1 = 2
8        THEN X := 2;
9      ELSIF PKG1.F1 = 3
10        THEN X := 3;
11      END IF;
12      DBMS_OUTPUT.PUT_LINE('X = ' || X);
13      DBMS_OUTPUT.PUT_LINE('PKG1.CNT = ' || PKG1.CNT);
14 END;
15 /
Call to PKG1.F1
Call to PKG1.F1
Call to PKG1.F1
X =
PKG1.CNT = 5
PL/SQL procedure successfully completed.
SQL> In such case you would have to introduce a temp variable:
SQL> CREATE OR REPLACE
2 PACKAGE PKG1
3 IS
4 CNT NUMBER;
5 FUNCTION F1 RETURN NUMBER;
6 END;
7 /
Package created.
SQL> CREATE OR REPLACE
2 PACKAGE BODY PKG1
3 IS
4 FUNCTION F1 RETURN NUMBER
5 IS
6 BEGIN
7 DBMS_OUTPUT.PUT_LINE('Call to PKG1.F1');
8 CNT := CNT + 1;
9 RETURN CNT;
10 END;
11 END;
12 /
Package body created.
SQL> SET SERVEROUTPUT ON
SQL> DECLARE
2      X NUMBER;
3 BEGIN
4      PKG1.CNT := 2;
5      X := CASE PKG1.F1
6             WHEN 1 THEN 1
7             WHEN 2 THEN 2
8             WHEN 3 THEN 3
9           END;
10      DBMS_OUTPUT.PUT_LINE('X = ' || X);
11      DBMS_OUTPUT.PUT_LINE('PKG1.CNT = ' || PKG1.CNT);
12 END;
13 /
Call to PKG1.F1
X = 3
PKG1.CNT = 3
PL/SQL procedure successfully completed.
SQL> DECLARE
2      X NUMBER;
3      TMP NUMBER;
4 BEGIN
5      PKG1.CNT := 2;
6      TMP := PKG1.F1;
7      IF TMP = 1
8        THEN X := 1;
9      ELSIF TMP = 2
10        THEN X := 2;
11      ELSIF TMP = 3
12        THEN X := 3;
13      END IF;
14      DBMS_OUTPUT.PUT_LINE('X = ' || X);
15      DBMS_OUTPUT.PUT_LINE('PKG1.CNT = ' || PKG1.CNT);
16 END;
17 /
Call to PKG1.F1
X = 3
PKG1.CNT = 3
PL/SQL procedure successfully completed.
SQL> SY.

Reg : Context-switching for built-in functions -

Hi Experts,
Asking this question just out of curiosity to know the internal concepts.
In a SQL query often we use the in-built Oracle functions like LOWER, UPPER, etc.
In this case, does context-switch happen?
Will I be able to look into the code of these functions after logging into SYS schema as SYSDBA?
FYI - I've Oracle XE 11.2 installed in my home pc (currently in office, so don't have access to it).
Help much appreciated!
Thanks,
Ranit

ranit B wrote:
Hi Experts,
Asking this question just out of curiosity to know the internal concepts.
In a SQL query often we use the in-built Oracle functions like LOWER, UPPER, etc.
In this case, does context-switch happen?No, because many of these functions are compiled at low level (C language) into the SQL and the PL/SQL engines, so each has their own 'copy' (in theory) to execute without having to context switch to the other engine.
Will I be able to look into the code of these functions after logging into SYS schema as SYSDBA?No, they are written in C and compiled into the engines.
In terms of the supplied packages (rather than built in functions), many of those are wrapped by oracle so you can only see the public interface, not the actual body code.

Overhead of SQL to PL/SQL context switch using an inline function

Hi,
We have a bit of sql in a third party application that uses an inline pl/sql function to do some security checks.
These security checks are redundant in our system - we don't use the functionality so the result is always true, but the function is always called for each line of output, which is over a thousand for a lot of records.
The function itself is fairly lightweight in our environment - the tables it uses are empty so each iteration of the function is quite quick (about .1 of a second per query in total, vs 12-15 seconds for the 'main' query). What I was wondering if there is any way of measuring the overhead of just doing the function calls.
If I do a trace of the session I see the timings and cost of the 'main' sql query, and the breakdown of the 2 sql statements that have been called in the function (with over 1000 executions each) but is there any way to measure how much of the time to execute the main query is spent doing the context switch?
Regards,
Carl

You could knock up some example to show the timings and measure it...
The following shows an example using context switching from PL/SQL to SQL and back in a loop, which gives an idea of the performance difference...
SQL> ed
Wrote file afiedt.buf
1 declare
2 v_sysdate DATE;
3 begin
4 v_sysdate := SYSDATE;
5 INSERT INTO mytable SELECT rownum FROM DUAL CONNECT BY ROWNUM <= 1000000;
6 DBMS_OUTPUT.PUT_LINE('Single Transaction: Time Taken: '||ROUND(((SYSDATE-v_sysdate)*(24*60*60)),0));
7 EXECUTE IMMEDIATE 'TRUNCATE TABLE mytable';
8 v_sysdate := SYSDATE;
9 FOR i IN 1..1000000
10 LOOP
11 INSERT INTO mytable (x) VALUES (i);
12 END LOOP;
13 DBMS_OUTPUT.PUT_LINE('Multi Transaction: Time Taken: '||ROUND(((SYSDATE-v_sysdate)*(24*60*60)),0));
14 EXECUTE IMMEDIATE 'TRUNCATE TABLE mytable';
15* end;
SQL> /
Single Transaction: Time Taken: 1
Multi Transaction: Time Taken: 37
PL/SQL procedure successfully completed.
SQL>Likewise you could time a query with X number of rows calling a PL/SQL function and not calling a PL/SQL function to see the difference. The more rows you do, the better idea you'll get of the difference.
;)

Question of the day: In which Thread context does the main method run ?

let's say, you create a JFrame in the main method:
public static void main(String[] args)
new MyJFrame();
let's suppose, you construct the whole JFrame
inside the constructor, which contains a lot
of other Swing components, like JTree's and more.
AND the JFrame calls setVisible(true) in its
constructor.
-> According to my knowledge, this won't work
properly - it will function, BUT you will get one
or two quite big exceptions (which usually start
with a nullpointer exception, produced by the
PLAF, but the whole stacktrace shows, that the
the exception was thrown AND catched inside
the Swing objects.
Reason for this would be, that all operations
after the "setVisible(true)" statement must be
carried out in the eventdispatch thread, according
to Swing rules.
Now, as you can see by running the source below,
the main() method always is processed in special thread,
NOT the event dispatch thread - which produces the above behaviour.
My Question:
Wouldn't it have been easier to let the JVM
process the main() methods right in the
event dispatch thread ?
Or:
Why is the JVM designed to process the
main() method in a special thread and not
the Swing event dispatch thread ?
opinions?
Regards
JPlaz / SnowRaver
package Test1;
import java.awt.*;
import java.awt.event.*;
import javax.swing.*;
public class Frame1 extends JFrame
public Frame1( final String title, int position )
this.setSize(400,300);
this.setLocation(position,position);
this.setVisible(true);
if( SwingUtilities.isEventDispatchThread() )
System.out.println("Constructor of "+title+" is processed IN the event dispatch thread.");
else
System.out.println("Constructor of "+title+" is processed in ANOTHER than the event dispatch thread.");
public static void main(String[] args)
// First creation just from the startup thread context :
new Frame1("first Frame", 60);
// Now we let the second creation be done in the
// eventdispatch thread context for sure :
SwingUtilities.invokeLater( new Runnable()
public void run()
new Frame1("second Frame",120);

Matt,
it's been so long, I don't have the full answer in my head.
This page on Adobe.com has an example that shows Flash 4
mouse stuff (I'm Mr
Vague today) when you scroll down. Your timelinme reference
will be
something like
TellTarget(_Root, gotoAndPlay(50))
Hopefully from that vagueness you can locate the right
answer.
Steve
Adobe Community Expert: Authorware, Flash Mobile and Devices
My blog -
http://stevehoward.blogspot.com/
Authorware tips -
http://www.tomorrows-key.com

FORALL context switching .. how it works ?

hi guys,
in the asktom link over here
http://asktom.oracle.com/pls/asktom/f?p=100:11:0::::P11_QUESTION_ID:17483288166654#17514276298239 
it was said that

forall i in 1 .. z_tab.count
insert
does this:
a) gather up inpus to insert (the entire z_tab)
b) perform context switch from PLSQL to SQL
c) execute insert N-times
d) perform context switch back from SQL to PLSQL

my question is does FORALL statement loops ?
does it do this ?
Example 1
loop 1
gather data to insert
loop 2
gather data to insert
loop 3
gather data to insert
loop finish
context switch to sql engine
perform insert 1 by 1
or
Example 2
loop 1 or no loop at all
gather all data required for insert
loop finish
context switch to sql engine
perform insert 1 by 1
my guess is example 1 is the correct answer
Advices gurus ?
Regards,
Noob

At what level are you asking the question?
In the context of PL/SQL, there is no loop. But if you pull back the layers and look at the intermediate language the procedure is compiled into, how the PL/SQL VM happens to implement those intermediate language instructions, how Oracle's C code happens to implement the VM, etc. it wouldn't shock me if there was some sort of loop-like construct in at least some level in some version of Oracle in some situation. Particularly depending on what you want to count as a loop at that level (somewhere in the SQL engine's C code, for example, Oracle might well have a loop when you're doing a full table scan, though it's probably not particularly useful to talk about a full table scan being in a loop)
In addition, why are you asking the question? I cannot envision a functionality or performance difference between the two approaches, so it doesn't seem like something that would have any influence on how you use a particular PL/SQL construct.
Justin

Tab context switch and validators problem?

Hi! We use JDeveloper 11.1.2.4.0.
Our application is using Dynamic Tabs UI Shell Template for dynamic tabs. The problem is, when I open a page with validation in one tab (for example: user edit form), and then try to switch context by clicking other open tab (for example: group edit form), it opens the right content (group edit form), but the validation messages for that prevoius tab are popping up (for user edit form). How can I avoid this behaviour? When I switch context, I don't want other validators to execute.
Regards, Marko

In this case check Decompiling ADF Binaries: How to Skip Validation? and Andrejus Baranovskis's Blog: Skip Validation for ADF Required Tabs
Timo

How to avoid responsibility context switcher?

Hi all.
i developed an OAF page called from JSP. i can navigate from jsp to OAF page successfully.
and i set pageContext.changeResponsibility() in processRequest(). i checked the information in 'About this page' and Responsibility was set on pagecontext .
but if i open the OAF page at first time, following message and 'Switch Responsibility' is occurred.
"The current responsibility context has been switched to: Application Developer".
i want to avoid this message and poplist on my page. how to avoid them?
Please let me know.

Pl see if the solution in MOS Doc 356814.1 (The Current Responsibility Context Has Been Switched To: Manager Self-Service) can help
HTH
Srini

Thread context switching.

Similar Messages

Maybe you are looking for