Assura 3.1.2 filling virtual memory

[Tim]

Hi,

I am running Assura 3.1.2 on IC5141 and can only run 2 or 3 DRC or LVS
runs before the virtual memory is filled up and I have to close and
restart Cadence to do more runs or even do certain commands (browse for
a component). The error message is "Unable to fork() child process...."
Some of the cells are small too, 15-20 components.

I know version 3.1.2 isn't qualified for IC5141 but we are stuck with
these versions due to foundry PDK support.

I tried the "ASSURA_UI_NONMETADATA 1" env variable which didn't seem to
have any affect.

My hardware and op system is Solaris 8 running on Sparc Ultra 10, 1G
physical memory, 1G virtual memory.

Is there anything I can do to help this?

Thanks in Advance,
Tim

 

[Martin Groepl]

Tim wrote:

Hi,

I am running Assura 3.1.2 on IC5141 and can only run 2 or 3 DRC or LVS
runs before the virtual memory is filled up and I have to close and
restart Cadence to do more runs or even do certain commands (browse for
a component). The error message is "Unable to fork() child process...."
Some of the cells are small too, 15-20 components.

I know version 3.1.2 isn't qualified for IC5141 but we are stuck with
these versions due to foundry PDK support.

I tried the "ASSURA_UI_NONMETADATA 1" env variable which didn't seem to
have any affect.

My hardware and op system is Solaris 8 running on Sparc Ultra 10, 1G
physical memory, 1G virtual memory.

Is there anything I can do to help this?

Thanks in Advance,
Tim

Check if more than one process "avxp" is running.
If so, try to kill them after finishing LVS.
pkill avxp
What happens if you use Assura314usr1cdb with your PDK?

Martin

 

[Tim]

Hi Martin,

Thanks for the reply.

I didn't see an avxp process running during an Assura run. I saw
"assura", "aveng", & "avlck". There were 2 avlck's there and 1 stayed
after I closed the run. I tried killing it but it didn't help.

I believe 314 solves this problem with a better use of memory. This
problem wasn't as bad with the previous usr release of 312 but we
needed to use the latest to solve the false "schematic and layout
match" gui problem.

Thanks Again,
Tim