2GB File Size Limitation with "External" Simulator ?

[JimE]

We are running an external simulator (Eldo) with Cadence Analog
Artist (I know, that's the old name !) version 5.0.33 and we
have a problem that the simulations all stop when the generated
results file (.wdb) reaches 2GB.

There are no error messages generated in any log I can find,
the simulation just stops and says it has completed, but it
hasn't, the results file just reached 2GB !

The OS it is running on is the 32 bit version of RHEL3, and the
simulator on its own, from the commandline with the same
netlist on the same machine, is able to pass 2GB with no
problem, so it does not seem to be a "simple" file system
issue !

Any ideas ?


JimE

 

[Andrew Beckett]

On Wed, 14 Dec 2005 08:36:49 -0600, JimE <gusr19@mot.com> wrote:

We are running an external simulator (Eldo) with Cadence Analog
Artist (I know, that's the old name !) version 5.0.33 and we
have a problem that the simulations all stop when the generated
results file (.wdb) reaches 2GB.

There are no error messages generated in any log I can find,
the simulation just stops and says it has completed, but it
hasn't, the results file just reached 2GB !

The OS it is running on is the 32 bit version of RHEL3, and the
simulator on its own, from the commandline with the same
netlist on the same machine, is able to pass 2GB with no
problem, so it does not seem to be a "simple" file system
issue !

Any ideas ?


JimE

If it's producing "wdb" (which is not a Cadence format), I can't
see how this has anything to do with the fact that it is run from ADE.

Have you reported it to Mentor?

Andrew.

 

[JimE]

Andrew Beckett wrote:

On Wed, 14 Dec 2005 08:36:49 -0600, JimE <gusr19@mot.com> wrote:


We are running an external simulator (Eldo) with Cadence Analog
Artist (I know, that's the old name !) version 5.0.33 and we
have a problem that the simulations all stop when the generated
results file (.wdb) reaches 2GB.

There are no error messages generated in any log I can find,
the simulation just stops and says it has completed, but it
hasn't, the results file just reached 2GB !

The OS it is running on is the 32 bit version of RHEL3, and the
simulator on its own, from the commandline with the same
netlist on the same machine, is able to pass 2GB with no
problem, so it does not seem to be a "simple" file system
issue !

Any ideas ?


JimE


If it's producing "wdb" (which is not a Cadence format), I can't
see how this has anything to do with the fact that it is run from ADE.

Have you reported it to Mentor?

Andrew.

Eldo is producing both a .wdb and a .psf compatible file called
timeSweep and the failing simulation dies when this second file
reaches 2GB. If Eldo is run to create just the .wdb it "happily"
goes beyond 2GB. Information from another simulator team says
that they have to partition .psf data into sub-2GB chunks for
Cadence to accept it, but that may not be related !

It is just very confusing and frustrating not to be able to
pinpoint a cause for what seems to be a problem other people
should have come across !


JimE

 

[Andrew Beckett]

On Thu, 15 Dec 2005 06:50:01 -0600, JimE <gusr19@mot.com> wrote:

Andrew Beckett wrote:
On Wed, 14 Dec 2005 08:36:49 -0600, JimE <gusr19@mot.com> wrote:


We are running an external simulator (Eldo) with Cadence Analog
Artist (I know, that's the old name !) version 5.0.33 and we
have a problem that the simulations all stop when the generated
results file (.wdb) reaches 2GB.

There are no error messages generated in any log I can find,
the simulation just stops and says it has completed, but it
hasn't, the results file just reached 2GB !

The OS it is running on is the 32 bit version of RHEL3, and the
simulator on its own, from the commandline with the same
netlist on the same machine, is able to pass 2GB with no
problem, so it does not seem to be a "simple" file system
issue !

Any ideas ?


JimE


If it's producing "wdb" (which is not a Cadence format), I can't
see how this has anything to do with the fact that it is run from ADE.

Have you reported it to Mentor?

Andrew.

Eldo is producing both a .wdb and a .psf compatible file called
timeSweep and the failing simulation dies when this second file
reaches 2GB. If Eldo is run to create just the .wdb it "happily"
goes beyond 2GB. Information from another simulator team says
that they have to partition .psf data into sub-2GB chunks for
Cadence to accept it, but that may not be related !

It is just very confusing and frustrating not to be able to
pinpoint a cause for what seems to be a problem other people
should have come across !


JimE

Jim,

It is possible for psf to be written (for transient type analyses at least)
where it breaks up the psf into 2G chunks (which is a limit on many OS
versions; it was done this way due to large file support not being available
on all supported OS at the time). However, I'm not 100% sure whether that
is available in the 3rd party psf API - but anyway, this should be reported to
Mentor - and they can request this from Cadence.

I suspect what's happening is that when the psf limit is reached, the error
causes the wdb file to stop being written too.

Regards,

Andrew.

 

[JimE]

Andrew Beckett wrote:

On Thu, 15 Dec 2005 06:50:01 -0600, JimE <gusr19@mot.com> wrote:


Andrew Beckett wrote:

On Wed, 14 Dec 2005 08:36:49 -0600, JimE <gusr19@mot.com> wrote:



We are running an external simulator (Eldo) with Cadence Analog
Artist (I know, that's the old name !) version 5.0.33 and we
have a problem that the simulations all stop when the generated
results file (.wdb) reaches 2GB.

There are no error messages generated in any log I can find,
the simulation just stops and says it has completed, but it
hasn't, the results file just reached 2GB !

The OS it is running on is the 32 bit version of RHEL3, and the
simulator on its own, from the commandline with the same
netlist on the same machine, is able to pass 2GB with no
problem, so it does not seem to be a "simple" file system
issue !

Any ideas ?


JimE


If it's producing "wdb" (which is not a Cadence format), I can't
see how this has anything to do with the fact that it is run from ADE.

Have you reported it to Mentor?

Andrew.

Eldo is producing both a .wdb and a .psf compatible file called
timeSweep and the failing simulation dies when this second file
reaches 2GB. If Eldo is run to create just the .wdb it "happily"
goes beyond 2GB. Information from another simulator team says
that they have to partition .psf data into sub-2GB chunks for
Cadence to accept it, but that may not be related !

It is just very confusing and frustrating not to be able to
pinpoint a cause for what seems to be a problem other people
should have come across !


JimE


Jim,

It is possible for psf to be written (for transient type analyses at least)
where it breaks up the psf into 2G chunks (which is a limit on many OS
versions; it was done this way due to large file support not being available
on all supported OS at the time). However, I'm not 100% sure whether that
is available in the 3rd party psf API - but anyway, this should be reported to
Mentor - and they can request this from Cadence.

I suspect what's happening is that when the psf limit is reached, the error
causes the wdb file to stop being written too.

Regards,

Andrew.

We confirmed it, the psf output file is reaching 2GB and then it
kills the simulator. Mentor now says this is a problem and they
will look into fixing it. I already told them about the splitting
of the file into <2GB chunks, now it is up to them to work out
what they have to do with Cadence. I have seen the splitting
for non-Cadence simulation engines, so they will release how
to do it !

Thanks

JimE