FLTBOUNDS coredump, relative to sun performance library

Hi,
I got an core dump, the "truss" log is:
27346: lwp_kill(1209, SIG#0) Err#3 ESRCH
27346: lwp_create(0xFFBEF088, LWP_SUSPENDED, 0xFFBEF084) = 1210
27346: lwp_create() (returning as new lwp ...) = 0
27346: lwp_continue(1210) = 0
27346: lwp_schedctl(SC_STATE|SC_PREEMPT, 0, 0xFDEFBDA4) = 0
27346: sigprocmask(SIG_SETMASK, 0xFF35819C, 0x00000000) = 0
27346: lwp_exit()
27346: lwp_wait(1210, 0xFFBEF24C) = 0
27346: lwp_wait(1210, 0xFFBEF24C) Err#3 ESRCH
27346: lwp_kill(1210, SIG#0) Err#3 ESRCH
27346: lwp_create(0xFFBEF088, LWP_SUSPENDED, 0xFFBEF084) = 1211
27346: lwp_create() (returning as new lwp ...) = 0
27346: lwp_continue(1211) = 0
27346: lwp_schedctl(SC_STATE|SC_PREEMPT, 0, 0xFDEFBDA4) = 0
27346: Incurred fault #6, FLTBOUNDS %pc = 0xFEF7756C
27346: siginfo: SIGSEGV SEGV_MAPERR addr=0xFF194000
27346: Received signal #11, SIGSEGV [default]
27346: siginfo: SIGSEGV SEGV_MAPERR addr=0xFF194000
27346: *** process killed ***
I used "/mnts/cdstools/SunCC-7.0.5/solaris-sparc-2.8/bin/CC test.C -xlic_l=sunperf -g -mt -lthread -R/usr/lib/lwp/" to compile.
I suspect there is a bug in dgemm(one function from Sun Performance Library), since if I comment it out, anything become ok.
Any one have similar problem?
Thanks.
ljg

Can you post a testcase of your code so I can try to replicate the problem?
Also, what is is your per-thread stack size set to? A large (4MB) stacksize is required if libsunperf is called from a user created thread.
See the following for more info if it is indeed a stacksize issue..
http://docs.sun.com/source/816-2463/plug_optimizing.html#pgfId-19550
-Moazam

Similar Messages

Sun Performance Library and OpenMP applications

Hi list,
I have a code that uses BLAS/Lapack libraries. This code requires the BLAS/Lapack to run in serial mode. If and when it runs in parallel this is a disaster. The code runs in parallel with OpenMP directives in SMP systems. I specify the number of threads by setting
OMP_NUM_THREADS
to the number of threads desired. Unfortunately if OMP_NUM_THREADS > 1 then sunperf runs in parallel as well. Is there anyway to tell to the sunperf library to run in serial mode no matter what the value of OMP_NUM_THREADS is?
For example GOTO BLAS library provides the environmental variable GOTO_NUM_THREADS which is dedicated to the threads of the GOTO library. So by setting GOTO_NUM_THREADS = 1, the library always runs in serial mode no matter what the value of OMP_NUM_THREADS is.
How could I achieve the same effect with Sun Performance Library?

Take a look at Perflib's USE_THREADS routine :
man -s 3p use_threads
Doing a call to USE_THREADS with the value of "1", before calling Perflib routines may work for you.
Also, I've noticed applications that have lots of small chunks of parallel work, need the threads to be very responsive. (And not sleep in between work chunks)
A way to make the threads not sleep is with the ${SUNW_MP_THR_IDLE} environment variable. I always set the value to "SPIN" :
% setenv SUNW_MP_THR_IDLE SPIN
You can see other settings for this in the OpenMP User's Guide, page 18 :
http://docs.sun.com/app/docs/doc/819-5270
Hope this helps.

CBLAS in sun performance library coming with solaris studio 12.2

Hello,
In sunperf library coming with sun studio 12u1 (linux x86) I can use cblas callings using the standard names cblas_xxxx. This interface do not appears in sunperf.h (you can use standard cblas.h) but the objects are in libsunperf.(a|so).
But in solaris studio 12.2 libsunperf cblas_xxxx objects do not exists. Is this normal? Contains libsunperf a standard c blas interface?
Thanks

Hello again,
In this blog post
http://www.mlds-networks.com/index.php/component/option,com_mojo/Itemid,29/p,35/
is explained how to link the ACML (AMD core math library) in order to use the standard CBLAS interface (ACML do not provides a standard CBLAS). I tried it and all runs OK. I did the same for the sun performance library in solstudio 12.2 and the compilation proccess runs ok, but in the testing step all functions fails becaude an incorrect argument in each function.
Exists any way for use the standard CBLAS interface with sun performance library? As I noted in my previous post, version 12.1 of sunstudio libsunperf contains the standard CBLAS interface, but I would like to use version 12.2
Thanks

Unable to link to sun performance library

I am using SS12 with nevada build 70. When trying to link to sun performance library routine from F90 code, I am getting errors. See the code and output below,
PROGRAM DeepWell
     USE SUNPERF
     IMPLICIT NONE
     INTEGER, PARAMETER :: N=5          ! No. of nucleons in consideration
     INTEGER :: i,j,k,ios
     DOUBLE PRECISION,DIMENSION(N,N) :: S,H
     DOUBLE PRECISION,DIMENSION(N) :: eig
     DOUBLE PRECISION :: tmp
     DO i=1,N
          DO j=i,N
               k = i+j
               IF (MOD(k,2) == 0) THEN
                    S(i,j) = 2.0/(k+5.0) - 4.0/(k+3.0) + 2.0/(k+1.0)
                    H(i,j) = 8*(k+2.0*i*j-1)/(k+3.0)/(k+1.0)/(k-1.0)
               ELSE
                    S(i,j) = 0.0
                    H(i,j) = 0.0
               ENDIF
ENDDO
     ENDDO
     call sygy(1,'V','U','N',H,B=S,W=eig)
END PROGRAM DeepWell
Output:
Running "/usr/bin/dmake -f Makefile CONF=Release" in /export/home/mbkumar/Thissjen/C3/This_C3_P1_VC
dmake: defaulting to parallel mode.
See the man page dmake(1) for more information on setting up the .dmakerc file.
bharat1 --> 1 job
/usr/bin/dmake -f nbproject/Makefile-Release.mk SUBPROJECTS= .build-conf
bharat1 --> 1 job
mkdir -p build/Release/Sun12-Solaris-x86
f90 -xtarget=opteron -m64 -dalign -xlic_lib=sunperf -c -O3 -w1 -o build/Release/Sun12-Solaris-x86/VC.o VC.f90
bharat1 --> Job output
mkdir -p build/Release/Sun12-Solaris-x86
f90 -xtarget=opteron -m64 -dalign -xlic_lib=sunperf -c -O3 -w1 -o build/Release/Sun12-Solaris-x86/VC.o VC.f90
(/export/home/mbkumar/Thissjen/C3/This_C3_P1_VC)VC.f90:
call ssygy(1,'V','U','N',H,B=S,W=eig)
^
"VC.f90", Line = 41, Column = 30: ERROR: An actual argument keyword is being used when an explicit interface is not known.
^
"VC.f90", Line = 41, Column = 34: ERROR: An actual argument keyword is being used when an explicit interface is not known.
f90comp: 44 SOURCE LINES
f90comp: 2 ERRORS, 0 WARNINGS, 0 OTHER MESSAGES, 0 ANSI
:(/export/home/mbkumar/Thissjen/C3/This_C3_P1_VC)VC.f90
*** Error code 1
dmake: Fatal error: Command failed for target `build/Release/Sun12-Solaris-x86/VC.o'
Current working directory /export/home/mbkumar/Thissjen/C3/This_C3_P1_VC
*** Error code 1
dmake: Fatal error: Command failed for target `.build-impl'
Build failed. Exit value 1.

please ignore the above question. I called non-existing routine.

Sun Performance library or just a tuned BLAS for Solaris x86?

Hi,
I know the sun performance library is not available on Solaris x86, but is
there at least a tuned blas library available?
thanks,
tim

The compilation line you list is correct. One way to see if there are outstanding unsatisfied references is to use the ldd command on the shared library that you created:
% CC -dalign -xarch=v9a -G -o LA.so $(OBJECTS) -library=sunperf -staticlib=sunperf
% ldd LA.so
I assume that one or more of the .o files in the $(OBJECTS) list makes calls to Perflib routines. Does your application (the code that is linking to the LA.so shared library) also make calls to Perflib routines? If so, these will not have been satisfied in the previous creation of the LA.so shared library.

How to Use the Sun Fortran Library

I wish to use the Sun Fortran Library of functions, but I am having trouble compiling my program. I would appreciate some help on selecting the proper compiler options, code modifications, etc.
My OS is Fedora 10 on a 64-bit system. I use the latest Sun Studio Express. I have used the Sun f95 compiler for quite some time as well, but this is my first use of one of the functions from the Sun Fortran Library.
Here is my first attempt after doing some unsuccessful research of my own on this issue:
PROGRAM test
IMPLICIT NONE
REAL*8 x, y
REAL*8, EXTERNAL :: d_lgamma
x=2.0
y=d_lgamma(x)
y=EXP(y)
WRITE(*,*) x, y
END PROGRAM test
Here's how I tried to compile:
f95 -lm test.f95test.o: In function `MAIN_':
test.f95:(.text+0x91): undefined reference to `d_lgamma_'
Obviously I am not doing this correctly. Is it possible that the Sun compiler and libraries have not been installed correctly on my system? How can I test if this is case?
Thank you for your help with this matter.

The routine d_lgamma_ is not supplied as part of Sun's Fortran product.
Sun's Fortran team does not maintain the library that contains that
routine. The team that does support the library has not yet ported it to
Linux. Different teams have different priorities.
In this case, there is a simple workaround. The C routine
#include <math.h>
double
d_lgamma_(double *x)
return (lgamma(*x));
implements the routine d_lgamma_.
Bob Corbett

ECC 6.0 SPRO has several notes related to BI performance setting

On my ECC 6.0 instance there are several nodes related to BI performance. After reading through the notes attached to the node and the reference SAP notes, I am perplexed by the setting and what is considered required or good to have.
Is there a note or some kind of document that could help provide a technical basis for the parameters setting.
Under Performance Setting-->
-Quantity Conversion: Set Buffer Size
-Maintain Runtime Parameters of DataStore Objects
-Settings for InfoSets
-Settings for Database Interfaces
-Settings for Database Interfaces (Oracle)
-Parameters for Aggregates
-Global Cache Settings
-Settings for Analysis Processes
Anyone have any thoughts on changing these parameters to better performance betweer or ECC and BW instances.
Thanks
Weyland Yutani

Hi Weyland,
a good starting point for BI Performance is the following SDN site:
Performance Reporting in BW [original link is broken]
See also these useful SAP Notes:
892513 - Consulting: Performance: Loading data, no of pkg, r
409641 - Examples of packet size dependency on ROIDOCPRMS
417307 - Extractor package size: Collective note for applications
857998 - Number range buffering for DIM IDs and SIDs
Rgds,
Colum

Native performance library not loading in weblogic 9.1. help

I'm trying to review an application and this is my first exposure to weblogic9.1. I've noticed that the native library failed to load in the logs but even after adding the right path, i keep seeing the same error message.
This surely affect the performance of the application running under this evironment i'm reviewing. Can someone else locate the resolve this problem?
####<Jun 27, 2007 3:40:13 PM EDT> <Info> <Server> <lydmtltst06> <AdminServer> <[STANDBY] ExecuteThread: '1' for queue: 'weblogic.kernel.Default (self-tuning)'> <<WLS Kernel>> <> <> <1182973213868> <BEA-002609> <Channel Service initialized.>
####<Jun 27, 2007 3:40:13 PM EDT> <Error> <Socket> <lydmtltst06> <AdminServer> <[ACTIVE] ExecuteThread: '0' for queue: 'weblogic.kernel.Default (self-tuning)'> <<WLS Kernel>> <> <> <1182973213916> <BEA-000438> <Unable to load performance pack. Using Java I/O instead. Please ensure that a native performance library is in: '/opt/java/1.5.0/jre/lib/sparc/server:/opt/java/1.5.0/jre/lib/sparc:/opt/java/1.5.0/jre/../lib/sparc::/usr/dt/lib:/usr/lib:/usr/ccs/lib:/usr/openwin/lib:/usr/ucb/lib:/usr/bin:/usr/ucb:/etc:/opt/bea/weblogic91/domains/lineage/server/native/solaris/sparc:/opt/bea/weblogic91/domains/lineage/server/native/solaris/sparc/oci920_8:/usr/lib'
>
####<Jun 27, 2007 3:40:13 PM EDT> <Info> <Socket> <lydmtltst06> <AdminServer> <[ACTIVE] ExecuteThread: '0' for queue: 'weblogic.kernel.Default (self-tuning)'> <<WLS Kernel>> <> <> <1182973213932> <BEA-000447> <Native IO Disabled. Using Java IO.>
####<Jun 27, 2007 3:40:14 PM EDT> <Info> <IIOP> <lydmtltst06> <AdminServer> <[ACTIVE] ExecuteThread: '0' for queue: 'weblogic.kernel.Default (self-tuning)'> <<WLS Kernel>> <> <> <1182973214947> <BEA-002014> <IIOP subsystem enabled.>
####<Jun 27, 2007 3:40:22 PM EDT> <Debug> <Deployment> <lydmtltst06> <AdminServer> <[STANDBY] ExecuteThread: '1' for queue: 'weblogic.kernel.Default (self-tuning)'> <<WLS Kernel>> <> <> <1182973222439> <000000> <Add DeploymentEventListener weblogic.security.service.DeploymentListener@6cc2a4>
####<Jun 27, 2007 3:40:22 PM EDT> <Debug> <Deployment> <lydmtltst06> <AdminServer> <[STANDBY] ExecuteThread: '1' for queue: 'weblogic.kernel.Default (self-tuning)'> <<WLS Kernel>> <> <> <1182973222483> <000000> <Add VetoableDeploymentListener: weblogic.security.service.DeploymentListener@6cc2a4>
The hardware used is a sunfire machine running solaris 9 (32).

Stating the obvious, but it is in the mbeantypes folder isn't it?
Did you let the upgrade wizard automatically upgrade the security provider? I used the domain upgrade wizard which converted my security provider, once I'd done that, I just took that new file (not sure what it changed) and I've been using it without problems.
Pete

Question related to project performance management

Hi,
We have implemented peoject management and project accounting module with Grants module for one of the semi public sector client. Now we are planning to enable project performance repporting.
I assume below things
1. If Grant is implemented, Project budgeting doesn't work
And
2. project performance reporting based on project accounting,project budgeting and project management modules.
If anyone of you have done similar requirement , Please let us know the approach your approach.
Thanks in advance
Praveen Javvaji

Duplicate post -- question related to project performance management

Question related to project performance reporting

Hi,
We have implemented peoject management and project accounting module with Grants module for one of the semi public sector client. Now we are planning to enable project performance repporting.
I assume below things
1. If Grant is implemented, Project budgeting doesn't work
And
2. project performance reporting based on project accounting,project budgeting and project management modules.
If anyone of you have done similar requirement , Please let us know the approach your approach.
Thanks in advance
Praveen Javvaji

Duplicate post -- question related to project performance management

Best performance library set up?

I was wondering if anyone can inform on what gives the best possible performance in how Aperture library is set up.
I prefer not having the library bloated on the main drive(OS) so is Aperture going to perform best if the - Library manages the RAW files and running on an external drive?
OR
- Library on main drive but "referencing" files?
OR
- Library on main drive and managing my RAW files?
If anyone has any informative insight on this I'd appreciate it
Thanks!

Hi Zirkel - I'd be happy to answer your question, but I need some clarification on what you mean by "what is the best way to keep this file updated" - Are you talking about the library?
If so, then you don't want to do this by manually putting new files on the external HD and opening aperture, instead you want to open aperture 1st and use the import dialog, let aperture import them and copy them to your masters drive.
If so, it's very easy when importing new files, you just change the settings in the import dialog -
Change the "store files in" dropdown to choose a "master folder" where you want to put all your raws. Then use the "subfolder" dropdown to create a Folder hierarchy automatically for you.
I should note that when I responded earlier, I wasn't thinking of "external HD" as a portable firewire drive - Depending on how convenient this is, I might just put a fast 500GB laptop drive in, and use the internal drive to store the RAW files (But I would still store them as referenced over managed, for the ease of maintenance in the future).

Event ID 2003 The configuration information of the performance library "perf-MSSQLSERVER-sqlctr11.1.3000.0.dll" for the "MSSQLSERVER" service does not match the trusted performance library information stored in the registry.

I keep getting the above error on all my SQL 2012 deployments (standard and enterprise) on all Windows Server 2012 Standard machines. I have already tried the following commands in administrator mode to resolve without success:
lodctr /T:perf-MSSQLSERVER-sqlctr11.1.3000.0.dll
lodctr /T:MSSQLSERVERlodctr /R:perf-MSSQLSERVER-sqlctr11.1.3000.0.dll
lodctr /R
Any other suggestions?
Diane

Hi,
Add the service accounts to the local server’s “Performance Monitor Users” group.
Configure Windows Service Accounts and Permissions
http://msdn.microsoft.com/en-IN/library/ms143504.aspx#VA_Desc

O-R Relations in Sun J2EE SERVER/OC4j

The O-R Mappings in Sun J2EE SERVER using Deployer Tool
Not working ...
I badly need help....

Raj -
You can use an old style (pfd 1 I think) of the ejb2.0 cmp/cmr with OC4J/Orion. The container will manage the relationships for you. I use it quite frequently in fact, though it is not advertised too heavily in the documentation. Once the ejb2.0 spec is incorporated fully into OC4J - you will have to make the odd change here-and-there.
For instance, create your abstract implementation classes -
Dept and Emp...
public abstract class DeptEJB implements EnityBean
//abstract getters/setters
public abstract void setDeptName(String deptName);
public abstract String getDeptName();
(etc)
// abstract 1-* relationship methods
public void addEmp(Emp emp) throws RemoteException,FinderException,NamingException,CreateException
this.getEmps().add(emp);
public abstract void setEmps(Collection emps);
public abstract Collection getEmps();
Create your abstract implementation class for Emp as well. While this relationship is certainly bi-directional (i.e. known to both) I am only doing a uni-directional relationship in this case.
After you have done this, you will need to put the relationship into the ejb-jar.xml file....You will need to create an ejb2.0 style descriptor.
For instance....
<?xml version = '1.0' encoding = 'windows-1252'?>
<!DOCTYPE ejb-jar PUBLIC "-//Sun Microsystems, Inc.//DTD Enterprise JavaBeans 2.0//EN" "http://java.sun.com/dtd/ejb-jar_2_0.dtd">
<ejb-jar>
<enterprise-beans>
<entity>
<description>Dept Entity Bean</description>
<display-name>Department</display-name>
<ejb-name>Dept</ejb-name>
<home>mypackage.DeptHome</home>
<remote>mypackage.Dept</remote>
<ejb-class>mypackage.impl.DeptEJB</ejb-class>
<persistence-type>Container</persistence-type>
<prim-key-class>java.lang.String</prim-key-class>
<abstract-schema-name>Dept</abstract-schema-name>
<cmp-version>2.X</cmp-version>
<reentrant>False</reentrant>
... Put all of your cmp fields as usual...
... In that same list (different than current ejb2.0)...
... place your cmr field ...
<cmp-field>
<field-name>emps</field-name>
</cmp-field>
</entity>

<entity>
<description>Emp Entity Bean</description>
<display-name>Employee</display-name>
<ejb-name>Empejb-name>
<home>mypackage.EmpHome</home>
<remote>mypackage.Emp</remote>
<ejb-class>mypackage.impl.EmpEJB</ejb-class>
<persistence-type>Container</persistence-type>
<prim-key-class>java.lang.String</prim-key-class>
<abstract-schema-name>Emp</abstract-schema-name>
<cmp-version>2.X</cmp-version>
<reentrant>False</reentrant>
... Put all of your cmp fields as usual...
<cmp-field>
<field-name>some_field</field-name>
</cmp-field>
... etc ...
</entity>
... The rest of your enterprise bean information
</enterprise-beans>
..... Next comes the relationships section ...
<<<<<NOTE - this differs somewhat from the final EJB2.0 spec
<<<<and you will need to make any required changes once it
<<<becomes final in OC4J
<relationships>
<ejb-relation>
     <ejb-relation-name>dept-emps</ejb-relation-name>
     <ejb-relationship-role>
     <ejb-relationship-role-name>
          dept-has-emps
     </ejb-relationship-role-name>
          <multiplicity>one</multiplicity>
          <role-source>
          <ejb-name>Dept</ejb-name>
          </role-source>
          <cmr-field>
          <cmr-field-name>emps</cmr-field-name>
          <cmr-field-type>java.util.Collection</cmr-field-type>
          </cmr-field>
     </ejb-relationship-role>
     <ejb-relationship-role>
     <ejb-relationship-role-name>
          emps-belong-to-dept
     </ejb-relationship-role-name>
     <multiplicity>many</multiplicity>
     <role-source>
     <ejb-name>Emp</ejb-name>
     </role-source>
     </ejb-relationship-role>
</ejb-relation>
</relationships>
... Finish off your deployment descriptor as usual
When deployed successfully, OC4J will create your entity tables and your relationship tables (build off of the primary ids of you entity beans). You can add and remove employees from your dept.
Okay - that is a quick way to look at it from ejb2.0 (as it exists today in OC4J) container managed relationships. As stated, this will change somewhat when the spec is finalized in OC4J (they are quiet about when that will happen).
If you want to do container managed relationships in ejb1.1 - go to the Orion support site www.atlassian.com and go to this link:
http://kb.atlassian.com/content/orionsupport/articles/complex-or.html
Hope I have helped somewhat and haven't completely confused the issue!
Cheers
Ray
Thanx for the information...
I am Using J2EE RI Server ...
can i have an example Code like Dept-Emp with 1-* Relation Ship
i am trying the same...
Plz clarify...
* Shouls We Create the Relation ship fields and relation
in the DataBase itself
OR
The Container Will Manage for us
Cheers
Raj

Sun C Library's equivalent of __GLIBC__

If GNU's C library is used, __GLIBC__ isdefined. Is there an equivalent macro for Sun's C Library? What is the official name for Sun's libc - Solaris C Library? Sun Studio C Library? Can someone point me to its documentation?

Some parts of libc (both Sun libc and glibc) are simple utility
functions where any normal implementation could be used in theory.
Other parts of the library (both glibc and Sun libc) are very interwoven with
the operating system and the runtime linker. Because of this it's
normal for all Solaris programs (even portable ones, even Linux-designed
software) to use the implementation of libc that comes built-in to Solaris,
when they are running on Solaris. As far as I know, there is no easy #define
to detect that you are using Sun's libc. You could use the combination
of: 1) running on Solaris AND 2) not using GLIBC. Even though I would
not expect to find any/many programs using glibc on Solaris, it's better
to check for both flags.
--chris

Sun L25 library error 2401

Recently, on my SUN L25 I began receiving an error 2401 message, but my library works (i succesfully backuped on it) and nothing shown in Prism Card Event History
http://img44.imageshack.us/img44/7743/0908210935.jpg
I tried to google its description, but without success. Who else has faced this, or at least can explain what this error code mean?
I would be grateful for any assistance.

Hi tolik2525
I`ve the same problem with a storedge L25. I make the following tests:
- Fw test. Firmware upgrade to drives, library and PRISM card.
- Boot library disconnected of host.
- Boot with one drive alternatively.
- Boot with one PSU alternatively.
- Change SCSI terminators.
And a long etc... After all the problem persists.
All sugesttions are wellcome.
Thanks all.
Edited by: zariweya on Dec 1, 2009 4:50 AM

FLTBOUNDS coredump, relative to sun performance library

Similar Messages

Maybe you are looking for