Most large models need to be condensed. That is, substitutions, backsolves and/or omissions need to be done when TABLO is run.
Did you forget to condense the model? When a model needs to be condensed, there is usually a Stored-input (STI) file which contains a standard list of omissions, substitutions and backsolves. You should use this STI file when running TABLO.
TABLO-generated programs and GEMSIM report detailed information about the amount of memory they need in various places during their run. In particular, see the end of the LOG file from any run for memory information.
With Release 9 and later, you have the option of adding the statement
share_memory = yes ;
to your Command file. Then GEMSIM and TABLO-generated programs attempt to release memory which is not needed for the current part of the calculation. This usually reduces the maximum memory requirement. See here for more details.
If you include the above statement and are using GEMPACK 9 or earlier, we recommend that you set the Environment variable FLIB_DVT_BUFFER equal to 0 [zero].
If there is not enough memory after you have condensed the model, maybe you need to condense the model further.
(a) Perhaps you can omit other exogenous variables which you are not going to shock.
(b) Perhaps you can make more substitutions or backsolves.
(c) Perhaps you need to convert some backsolves to substitutions.
See here for more about condensation.
If you are considering purchasing more memory, you should take note of possible memory limits.
Go back to
GEMPACK FAQ Page