OSEC

Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com
 
Subject: Re: New_AIXV4_Fixes
From: AIX Service Mail Server (aixservaustin.ibm.com)
Date: Tue Nov 07 2000 - 09:08:40 CST

APAR: IY12313 COMPID: 5765D5100 REL: 320
ABSTRACT: INCORRECT RETRANSMITTAL OF PACKET BY KLAPI CAUSING CRASH.

PROBLEM DESCRIPTION:
Given the LAPI state at the time of the crash , the packet
which was retransmitted should not have been retransmitted.
However, it is possible that and error callback had been made
to VSD. If this were done, the LAPI lock would have been
released and another thread could have change the LAPI state
from one where the current packet should be sent to one where
it should not have been sent. Since the packet to resend are
calculated before any error callback is made. This is certainly

PROBLEM SUMMARY:
If a packet acknowlgement is received from a node while
calling an error handler resulting from a timeout to the
same node, an invalid packet will be transmitted.

PROBLEM CONCLUSION:
No packets will be retransmitted to a node immediately after
calling a user error handler as a result of a timeout to
that node. This will prevent erroreous packet transmission.

------

APAR: IY12748 COMPID: 5765D5100 REL: 320
ABSTRACT: MIGRATING NODES CAN GIVE LED A69 HANG IN GET_KEYFILES SCRIPT

PROBLEM DESCRIPTION:
Customer migrates a node to PSSP3.2. They experience a A69
hang during the node migration/pssp_script step. The hang will
eventually time out, and the pssp_script continues on.
Looking at the configfb log in the sysman directory, on the
node, you can see that the c69 step caught a SIGALRM and exited.
The problem occurs during the call of:
/usr/lpp/ssp/bin/get_keyfiles
Which is trying to get the new servtab thru the nodes tty.

LOCAL FIX:
Manually ftp the new servtab for the node after the A69 hang
has timed out.

PROBLEM CONCLUSION:
Problem was due to missing entry in hmacls file . This was
determined to be a missing process (code-wise) in Migration.
 Added code to setup_authent and post_process to update
hmacls w/ SPbgAdm entry.

TEMPORARY FIX:
Add "<cws -hostname> root.SPbgAdm a" (without the quotes) to
the /spdata/sys1/spmon/hmacls file... Stop and re-start
hardmon.

------

APAR: IY12758 COMPID: 5765D6100 REL: 220
ABSTRACT: MULTIPLE MASTER NODE EXCLUSIVE CRASHES THE NEGOTIATO

PROBLEM DESCRIPTION:
Defining multiple machines with master_node_exclusive = True
causes severe negotiator performance problems ans crashes.

PROBLEM SUMMARY:
When master_node_exclusive = true is set for more than one
machine, the negotiator loops and crashes. Jobs are
hung.

PROBLEM CONCLUSION:
When master_node_exclusive = true is set for more than one
machine, jobs will now run properly.

------

APAR: IY12825 COMPID: 5765D5101 REL: 120
ABSTRACT: NODE POWER INDICATOR NOT DISPLAYED ON BACKUP CWS

PROBLEM DESCRIPTION:
With hacws configured, when backup cws takes over, perspectives
doesn't show node power indicator.

LOCAL FIX:
Change the name of primary cws to the secondary cws one in
/spdata/sys1/spmon/hmacls file.

PROBLEM SUMMARY:
In hmrmd, it uses the gethostname instead of virtual
hostname to communicate to hardmon. For HACWS, the
backup HACWS hostname is different with primary HACWS
hostname, so hmrmd cannot communicate to hardmon by
using the backup HACWS hostname and it cause the
perspectives sphardware does not show any node power
indiator.

PROBLEM CONCLUSION:
Changed the code in hmrmd to use getvhostname instead of
gethostname. So the backup HACWS hmrmd can communicate with
hardmon properly.

------

APAR: IY12831 COMPID: 5765D5100 REL: 320
ABSTRACT: PTPE UNABLE TO START WITH PTPE.CF FILE AND LARGE NUMBER OF

PROBLEM DESCRIPTION:
Customer has over 9000 stats because of a large # of VSDs. PTPE
starts fine w/o a ptpe.cf file, but will not start on nodes w/
a large # of statistics if the ptpe.cf file is present. Please
note that this customer indicates that this "test" system has
only a small number of VSDs as compared to the production
system that they wish to use it on.

PROBLEM SUMMARY:
PTPE obtains SP specific performance information from the
Event Management "harmld" daemon. If this daemon is not
running when PTPE data collection is started, the PTPE
daemons start up the "harmld" daemon. After the daemon has
started, PTPE instructs the "harmld" daemon to provide all
available SP-specific statistics to the Spmi shared memory.
PTPE then waits for a specific period of time for "harmld"
to confirm this request. In the case described by this
APAR, "harmld" is taking much longer than anticipated to
confirm the request made by PTPE. PTPE reports this as a
failure, but continues, only to find the Spmi shared memory
locked by another process (harmld). Assuming this to be a
deadlock situation, the PTPE daemon shuts down on this node.

PROBLEM CONCLUSION:
The repair is to increase the timeout value used by the PTPE
daemons in waiting for the "harmld" daemon confirmation.
This timeout is static, due to the design of the PTPE
daemon.

------

APAR: IY13172 COMPID: 5765D5100 REL: 320
ABSTRACT: NODE CRASH IN XMFREE1() WHEN MODS ENABLED WITH SOME TB3PCI

PROBLEM DESCRIPTION:
A node crash can occur when MODs are turned on with some TB3PCI
switch adapters after applying IY03548 (there is no code error
with the APAR but it exposed a problem with the loading of
xilinx on some cards). This problem has only been seen on
S80 and S7A servers.

LOCAL FIX:
It is possible to avoid the crash on boot by removing the TB3PCI
card.

PROBLEM SUMMARY:
Some TB3PCI switch cards have a compatibility problem with
APAR IY03548, which can cause crashes on boot when loading
the switch xilinx if MODs is enabled.

PROBLEM CONCLUSION:
The xilinx_file3pci file has been changed to prevent
crashes that can occur loading the xilinx file with
some TB3PCI switch cards.

TEMPORARY FIX:
remove TB3PCI switch card/disable MODs

------

APAR: IY13173 COMPID: 576552900 REL: 220
ABSTRACT: IF_LS KERNEL EXTENSION CAUSES PERF. PROB ON NON-POWER NODE WITH

PROBLEM DESCRIPTION:
Performance problems can occur on non-power nodes (e.g POWERPC),
when transferring large amounts of data over the SP-Switch.
A kernel trace shows a looping of "resume interrupt process" and
"PROGRAM CHECK", which is burning up a lot of CPU. The problem
is that the instruction is a POWER instruction, which must be
emulated.

PROBLEM SUMMARY:
On non-POWER machines (e.g. POWERPC) a performance
problem can occur when the SP switch is under a very
heavy load.

PROBLEM CONCLUSION:
The switch IP kernel extension, if_ls, has been changed
to remove POWER instructions which must be emulated
on non-POWER machines (e.g. POWERPC), as emulating
instructions can cause a performance problem.

------

APAR: IY13176 COMPID: 5765D5100 REL: 320
ABSTRACT: IF_LS KERNEL EXTENSION CAUSES PERF. PROB ON NON-POWER NODE WITH

PROBLEM DESCRIPTION:
Performance problems can occur on non-power nodes (e.g POWERPC),
when transferring large amounts of data over the SP-Switch.
A kernel trace shows a looping of "resume interrupt process" and
"PROGRAM CHECK", which is burning up a lot of CPU. The problem
is that the instruction is a POWER instruction, which must be
emulated.

PROBLEM SUMMARY:
On non-POWER machines (e.g. POWERPC) a performance
problem can occur when the SP switch is under a very
heavy load.

PROBLEM CONCLUSION:
The switch IP kernel extension, if_ls, has been changed
to remove POWER instructions which must be emulated
on non-POWER machines (e.g. POWERPC), as emulating
instructions can cause a performance problem.

------

APAR: IY13267 COMPID: 5765B9501 REL: 320
ABSTRACT: MMFS: 6027-696 TOO MANY DISKS ARE UNAVAILABLE MMCHDISK FAILS

PROBLEM DESCRIPTION:
After some disk is marked "down", filesystem cannot be mounted a
gain
with message: MMFS: 6027-696 Too many disks are unavailable.
Then mmchdisk start also gives the same message when it should a
llow
the disk to be started.

PROBLEM SUMMARY:
If only one disk is setup for metadata residency and that
disk is marked down because of a VSD server failure,
mmchdisk will not restore the disk

PROBLEM CONCLUSION:
Handle the situation of the only metadata disk being marked
down.

------

APAR: IY13277 COMPID: 5765D5100 REL: 320
ABSTRACT: IF_LS ATTEMPTS TO SEND PACKETS WITHOUT CORRECTLY CHECKING THE

PROBLEM DESCRIPTION:
In PSSP 3.1.1, AIX 4.3.3.02environment it has been detected
next error:
if_lscan attempt to send packets without correctly checking
the css interface flags.
This problem produces a dump when rebooting the node.

PROBLEM SUMMARY:
A node crash can occur when the switch kernel
extension, if_ls, attempts to send packets
before its data structures are initialized.

PROBLEM CONCLUSION:
The switch if_ls kernel extension has been
changed to
prevent a node crash from occuring when the IP driver
attempts to send before initializing its data
structures.

------

APAR: IY13401 COMPID: 5765D6100 REL: 220
ABSTRACT: LLCTL DUMPS CORE

PROBLEM DESCRIPTION:
llctl -h sys5itoh suspend
llctl -h sys5itoh resume schedd
llctl -h sys5itoh resume startd allclasses
=> then llctl dumps core

PROBLEM SUMMARY:
llctl -g resume startd allclasses core dumps every
time it is executed on customer machine.

PROBLEM CONCLUSION:
It was determined that llctl was having a type
casting problem while parsing the command line
for the allclasses keyword.

------

APAR: IY13439 COMPID: 5765D6100 REL: 220
ABSTRACT: LLEXTSDR: 2512-379, "SDRTURBOGETOBJECTS, RC=26

PROBLEM DESCRIPTION:
When issueing llextSDR to extract SDR data fir the LoadLeveler
Administration file, you get error:
llextSDR: 2512-379 The SDR function "SDRTurboGetObjects"
                   failed with a return code of "26".

PROBLEM SUMMARY:
The LoadLeveler supports an optional class in the SDR for
the first time and the error checking was not set up
properly for it

PROBLEM CONCLUSION:
LoadLeveler error checking now tolerates the absence of
optional fields in the SDR

TEMPORARY FIX:
Redirect standard error to a file and filter file througH
grep -v SDRTurboGetObjects

------

APAR: IY13448 COMPID: 5765D5101 REL: 120
ABSTRACT: GENERIC FIXES FOR RSCT 120

PROBLEM DESCRIPTION:
GENERIC FIXES FOR RSCT 120

------

APAR: IY13449 COMPID: 5765D5100 REL: 320
ABSTRACT: GENERIC FIXES FOR PSSP 320

PROBLEM DESCRIPTION:
GENERIC FIXES FOR PSSP 320

------

APAR: IY13450 COMPID: 5765D9300 REL: 310
ABSTRACT: GENERIC FIXES FOR PPE 310

PROBLEM DESCRIPTION:
GENERIC FIXES FOR PPE 310

------

APAR: IY13451 COMPID: 5765B9501 REL: 320
ABSTRACT: GENERIC FIXES FOR GPFS 320

PROBLEM DESCRIPTION:
GENERIC FIXES FOR GPFS 320

------

APAR: IY13452 COMPID: 5765D6100 REL: 220
ABSTRACT: GENERIC FIXES FOR LOADL 220

PROBLEM DESCRIPTION:
GENERIC FIXES FOR LOADL 220

------

APAR: IY13455 COMPID: 5765D6100 REL: 220
ABSTRACT: UNIX_GROUP NAME IS NOT BEING ACKNOWLEDGED.

PROBLEM DESCRIPTION:
If "Unix_group" is specified as the user's default group
and the include_groups lines within the class stanza
contains the user's unix group name, the unix group name is
not being acknowledged thus preventing the job from being
submitted.
Sample
 Yoshida: type = user
          default_group = UNIX_GROUP
 large: type = class
        include_groups = staff
Error
 llsubmit: 2512-080 Class "large" is not valid for group "Unix_G
           roup".
 llsubmit: 2512-051 This job has not been submitted to LoadLevel
           er.

PROBLEM SUMMARY:
If "Unix_Group" is specified as the user's default group
and the include_groups lines within the class stanza
contains the user's unix group name, the unix group name is
not being acknowledged thus preventing the job from being
submitted.

PROBLEM CONCLUSION:
The llsubmit command has been changed so that when the unix
group name is listed on either the include_groups or
exclude_groups line within the class stanza, it will be
acknowledged when the user's default group is specified as
"Unix_Group".

------

APAR: IY13459 COMPID: 5765D5100 REL: 320
ABSTRACT: PCP FAILS FOR FILES > 2 GIG

PROBLEM DESCRIPTION:
The pcp script fails when files greater than 2 gig are specified
This is due to using the perl '-r' function with the pssp level
of perl which does not support files > 2 gig.

PROBLEM SUMMARY:
The PERL which we ship does 32bit addressing which
prevents processing files larger than 2 GB in size. As
a result pcp of a file larger than 2 GB would fail when
checking the existence of the file with the -r flag with
the message: Invalid file specified

PROBLEM CONCLUSION:
Modified pcp to use the system ls command instead of the
-r flag to determine the existence of a file. This allows
files larger than 2 GB to be processed.

------

APAR: IY13512 COMPID: 5765D5100 REL: 320
ABSTRACT: 128WAY: VSD.UP1 WAS RUNNING MANY CFGMGR CALLS

PROBLEM DESCRIPTION:
128way: vsd.UP1 was running many cfgmgr calls

PROBLEM SUMMARY:
RVSD recovery can potentially run the "cfgmgr" command
several
times when trying to bring volume groups online. It is only
necessary to run "cfgmgr" once.

PROBLEM CONCLUSION:
RVSD will use a "file lock" to serialize multiple threads
deciding whether to run the "cfgmgr" command. Therefore,
the "cfgmgr" command will only be run once when it is
determined that it is needed.

------

APAR: IY13626 COMPID: 5765B9501 REL: 320
ABSTRACT: MMFSD DIES DUE TO EINVAL TRAP

PROBLEM DESCRIPTION:
mmfsd dies due to einval trap

PROBLEM SUMMARY:
GPFS terminated incorrectly if it received an EINPROGRESS
return code from connect

PROBLEM CONCLUSION:
After receiving EINPROGRESS, issue a select() call to wait
for completion.

------

APAR: IY13627 COMPID: 5765D5100 REL: 320
ABSTRACT: SYSCTLD DAEMON CORE DUMPS IF NO INSTANCE SERPERATOR IS USED IN

PROBLEM DESCRIPTION:
If no instance seperator is used in the customer's /etc/*.acl
files the sysctld daemon core dumps.
The sysctld daemon will also core dump if short REALM
names are used in the /etc/*.acl files

PROBLEM SUMMARY:
If no instance seperator is used in the customer's
/etc/*.acl files or if short REALM names are used,
the sysctld daemon will core dump.

PROBLEM CONCLUSION:
Changed the verification permission logic.

------

APAR: IY13879 COMPID: 5765B8100 REL: 220
ABSTRACT: UNEXPECTED RETURN CODE FROM GET_SERVICEID CALL IN DIRTALK GEOTEL

PROBLEM DESCRIPTION:
Custom server 'forgetting' services exist

PROBLEM SUMMARY:
Custom server 'forgetting' that services
exist

PROBLEM CONCLUSION:
A modification to the header files and
source code to ensure that there were no conflicts between
variable declarations

------

APAR: IY13880 COMPID: 5765B8100 REL: 220
ABSTRACT: RDSET3270 DISABLE DOES NOT DISABLE IF DIRTALK.3270 IS INSTALLED.

PROBLEM DESCRIPTION:
If you set 'RDSET3270 disable' on the command line, and
dirTalk.3270 is installed, it does not disable the 3270 via the
GUI.

PROBLEM SUMMARY:
If dirTalk.3270 installed the 3270 feature
is always enabled.

PROBLEM CONCLUSION:
Corrected dirTalk startup to operate
correctly and not always enable 3270

------

APAR: IY14022 COMPID: 576552900 REL: 220
ABSTRACT: LATEST PSSP 2.2 FIXES AS OF OCTOBER 2000

PROBLEM DESCRIPTION:
This is the lastest PSSP ptf as of October 2000.
Order this apar to get all of the ptfs as of October 2000.

PROBLEM SUMMARY:
This is a packaging apar for PSSP 2.2 fixes
as of October 2000.

PROBLEM CONCLUSION:
This is a packaging apar for PSSP 2.2 fixes
as of October 2000.

------

APAR: IY14067 COMPID: 5765C3403 REL: 430
ABSTRACT: AIX 4.3 SECURITY RELATED UPDATES AS OF OCTOBER 2000

PROBLEM DESCRIPTION:
This APAR delivers security related updates for AIX 4.3
available as of October 2000.
This is a packaging APAR only. It will not appear in the list
of APARs on the SMIT "Update Software by Fix (APAR)" panel, nor
will the 'instfix' command show this APAR as being installed
after the updates delivered by this package are installed.
To install selected updates from this package, use the command:
  smit update_by_fix
To install all updates from this package that apply to installed
filesets on your system, use the command:
  smit update_all

PROBLEM SUMMARY:
Packaging only.

PROBLEM CONCLUSION:
Packaging only.

------

APAR: IY14072 COMPID: 5765C3403 REL: 433
ABSTRACT: LATEST AIX 4.3.3 UPDATES AS OF OCTOBER 2000

PROBLEM DESCRIPTION:
Order this APAR to receive the latest AIX 4.3.3 updates
available as of October 2000. These updates are applicable to
AIX 4.3.3 with the 4330-06 Recommended Maintenance Level, and
should be ordered using service level 433006.
This is not a recommended preventive service package and could
contain unresolved PTFs in error (PEs). This is simply a
packaging of the latest available updates.
This is is packaging APAR only. It will not appear in the list
of APARs on the SMIT "Update Software by Fix (APAR)" panel, nor
will the the 'instfix' command show this APAR as being installed
after the updates delivered by this package are installed.
To install selected updates from this package, use the command:
  smit update_by_fix
To install all updates from this package that apply to installed
filesets on your system, use the command:
  smit update_all

PROBLEM SUMMARY:
Packaging only.

PROBLEM CONCLUSION:
Packaging only.

------

APAR: IY14108 COMPID: 5765D5100 REL: 320
ABSTRACT: LATEST PSSP 3.2.0 FIXES AS OF OCTOBER 2000

PROBLEM DESCRIPTION:
This is the lastest PSSP ptf as of October 2000.
Order this apar to get all of the ptfs as of October 2000.

PROBLEM SUMMARY:
This is a packaging apar for PSSP 3.2.0 fixes
as of October 2000.

PROBLEM CONCLUSION:
This is a packaging apar for PSSP 3.2.0
fixes as of October 2000.

------

APAR: PQ41506 COMPID: 5765C4100 REL: 220
ABSTRACT: BLACS MIN/MAX/SUM SUBROUTINES DO NOT HANDLE CDEST=-1 PROPERLY.

PROBLEM DESCRIPTION:
For Parallel ESSL BLACS codes xGAMX2D, xGAMN2D and xGSUM2D where
x is any precision(C,D,I,S,Z), CDEST=-1 causes an error to be
reported by MPI unless RDEST=-1.

PROBLEM SUMMARY:
The BLACS standard was incorrectly interpreted.
As a result, the MIN/MAX/SUM codes only left the results on all
processors when RDEST=-1. Results should be left on all
processors when RDEST=-1 or CDEST=-1. When CDEST=-1 and RDEST
was not -1, then an MPI error message was issued.

PROBLEM CONCLUSION:
All precisions of the BLACS MIN/MAX/SUM
codes were corrected to interpret RDEST=-1 or CDEST=-1
correctly.

TEMPORARY FIX:
User may code their call such that RDEST=-1 and
CDEST=-1.

------

APAR: PQ41508 COMPID: 5765C4100 REL: 220
ABSTRACT: PBLAS FAIL TO ISSUE REASON FOR TERMINATING WHEN INTERNAL STORAGE

PROBLEM DESCRIPTION:
Applications using Parallel ESSL may terminate without issuing
a 400 error when an internal allocation fails in the PBLAS.

PROBLEM SUMMARY:
Applications using Parallel ESSL may terminate
without issuing a 400 error message when an internal allocation
fails in a call to any PBLAS subroutine.

PROBLEM CONCLUSION:
PBUFFGET(and internal code) was corrected
to properly handle the allocation failure.

------

APAR: PQ41575 COMPID: 564806201 REL: 110
ABSTRACT: SYNCHRONIZE ACIF MAINTENANCE AND INCLUDES MINOR CCM CHANGES

PROBLEM DESCRIPTION:
Synchronize ACIF version maintenance and include minor CCM
changes.CHANGES

PROBLEM SUMMARY:
****************************************************************
* USERS AFFECTED: All ACIF users. *
****************************************************************
* PROBLEM DESCRIPTION: PQ41575 synchronizes internal files *
* used to build the ACIF product and *
* incorporates several minor changes to *
* the CCM code. CCM is common to several *
* products and the changes provided here *
* do not address any known ACIF defects. *
* The changes are being incorporated into *
* ACIF for common code consistency. *
****************************************************************
* RECOMMENDATION: Apply the applicable PTF. *
****************************************************************
PQ41575 synchronizes internal files used to build the ACIF
product and incorporates several minor changes to the CCM
code. CCM is common to several products and the changes provided
here do not address any known ACIF defects. The changes are
being incorporated into ACIF for common code consistency.

PROBLEM CONCLUSION:
PQ41575 is a maintenance roll-up.

------

APAR: PQ42484 COMPID: 5765C4200 REL: 312
ABSTRACT: POTENTIAL WRONG ANSWERS IN OUTPUT ONLY VECTORS FOR LEVEL 1 BLAS

PROBLEM DESCRIPTION:
ESSL Level 1 BLAS subroutines with output only vectors may
produce incorrect results if the output vector has non-numeric
data on input. Included in this are the matrix add and subtract
routines.

LOCAL FIX:
Initialize the output only vector or matrix to have valid
numeric data before calling the ESSL subroutines.

PROBLEM SUMMARY:
Level 1 and matrix add/subtract routines
which have output only vectors/matrices may have incorrect
results if the data in the output only vector/matrix was
non-numeric(INF or NANQ) on input.

PROBLEM CONCLUSION:
Code was corrected to not reference the
output only arrays.

------

APAR: PQ42485 COMPID: 5765C4200 REL: 312
ABSTRACT: GENERAL MATRIX FACTOR MAY CORE DUMP WHEN USING 64-BIT, SMP AND

PROBLEM DESCRIPTION:
When using ESSL general matrix factor subroutines in 64-bit
mode with the ESSL SMP library and AIX Malloc Debug turned on,
users may experience an unexpected core dump.

LOCAL FIX:
Do not use AIX Malloc Debug when using ESSL general matrix
factor subroutines in 64-bit and SMP mode.

PROBLEM SUMMARY:
General matrix factor routines may core dump
in 64-bit SMP mode with Malloc Debug enabled.

PROBLEM CONCLUSION:
Problem was due to an incorrect assumption
regarding compiler option -qnostrict. Code updated to use
-qstrict instead.

------

APAR: PQ42486 COMPID: 5765C4200 REL: 312
ABSTRACT: UPDATE ESSL HEADER FILE TO SUPPORT VISUAL AGE C AND C++ V5

PROBLEM DESCRIPTION:
C and C++ programs using the ESSL header file(essl.h) will not
compiler with the Visual Age C and C++ V5 compilers.

PROBLEM SUMMARY:
ESSL header file(essl.h) needed changes to
support Visual Age C and C++ Version 5 compilers.

PROBLEM CONCLUSION:
ESSL header file updated to work with
Visual Age C and C++ Version 5 compilers.

------

APAR: PQ42488 COMPID: 5765C4200 REL: 312
ABSTRACT: ZAXPY DOES NOT EXPLOIT MULTIPLE THREADS

PROBLEM DESCRIPTION:
ESSL subroutine ZAXPY was not enabled to take advantage of
multiple threads when linking with the ESSL SMP library.

PROBLEM SUMMARY:
ZAXPY did not scale in SMP mode.

PROBLEM CONCLUSION:
ZAXPY was not enabled to use multiple
threads internally.

------

APAR: PQ42490 COMPID: 5765C4200 REL: 312
ABSTRACT: SPTNQ/DPTNQ MAY PRODUCE INCORRECT RESULTS FOR N>4

PROBLEM DESCRIPTION:
ESSL quadrature subroutines SPTNQ and DPTNQ may produce wrong
answers for N>4 due to an internal variable which was not
initialized.

PROBLEM SUMMARY:
SPTNQ and DPTNQ may produce wrong answers for N>4 due to
an uninitialized internal variable.

PROBLEM CONCLUSION:
Codes updated to initialize internal variable.

------

APAR: PQ42492 COMPID: 5765C4200 REL: 312
ABSTRACT: FFT PERFORMANCE ENHANCEMENTS WITH SMP LIBRARY

PROBLEM DESCRIPTION:
ESSL FFT subroutines were not scaling well in SMP mode.
Internally, the number of threads to use was being incorrectly
calculated.

PROBLEM SUMMARY:
FFT subroutines do not scale properly in SMP
mode.

PROBLEM CONCLUSION:
FFT subroutines needed updating to properly
choose the number of threads in SMP mode.

------

APAR: PQ42493 COMPID: 5765C4200 REL: 312
ABSTRACT: ESSL ERROR MONITOR ISSUED MESSAGES WHEN PRINTING WAS SUPPRESSED

PROBLEM DESCRIPTION:
ESSL will print error messages that should have been suppressed
when the error count exceeded 32767.

PROBLEM SUMMARY:
ESSL was printing error messages when error
message printing was suppressed.

PROBLEM CONCLUSION:
The problem was that after 32767 messages,
the count of errors became negative which allowed the messages
to print. The error count is now capped at 256.

------

APAR: PQ42495 COMPID: 5765C4200 REL: 312
ABSTRACT: ADD NEW IOPT VALUES FOR SPPF AND DPPF TO LEAVE OUTPUT IN

PROBLEM DESCRIPTION:
In ESSL 312, a new output format for SPPF and DPPF was
introduced. A customer requested the ability to have the output
remain in lower-packed storage mode.

PROBLEM SUMMARY:
ESSL 312 introduced a new output format in
SPPF and DPPF. A customer requested an option to have the
output in the original lower-packed storage mode.

PROBLEM CONCLUSION:
IOPT=10 and IOPT=11 were added to SPPF and
DPPF. These correspond to IOPT=0 and IOPT=1 except that the
output will be left in lower-packed storage mode when using
IOPT=10 or IOPT=11.

------