This describes features of Release 5.2 (September 1996) of GEMPACK which were not available in Release 5.1 (April 1994) of GEMPACK. These include:
Some details are given below.
When running GEMSIM or TABLO-generated programs, extra TABLO-like statements (for example, declaring extra sets or asking for extra WRITEs or DISPLAYs) can be included in the Command file. Such statements are carried out as if they were appended to the end of the original TABLO Input file.
Arrays on Header Array files can now contain set and element labelling information, namely the names of the coefficients usually associated with the arrays, the names of the sets over which the arguments of the array run and the names of the elements of the sets involved with the array. TABLO-generated programs and GEMSIM now automatically write this information to any arrays they write to a Header Array file. This information is shown in SEEHAR output and also when SLTOHT produces Header Array output (the default) or when the new option 'SSE' is selected. It is also shown by VIEWHAR (see below).
When GEMPACK runs on DOS machines under Microsoft Windows or Windows95, it can be run under a new GEMPACK Windows interface called WinGEM. Users can carry out modelling tasks using GEMPACK windows. We hope that this will make modelling easier for both new and experienced modellers using these machines. As part of this, a new Windows program VIEWHAR (written by our colleague Mark Horridge) allows windows-type access to examine the data in Header Array files. [You can think of VIEWHAR as a Windows version of SEEHAR.]
New sparse linear equation solving routines MA48 (sometimes a faster alternative to the MA28 routines) are now the default. The new routines are usually significantly faster for multi-regional models (including GTAP) and intertemporal models, though they are not faster for some single-country models (for example, MONASH).
You can now specify in advance what accuracy you require for simulations carried out via TABLO-generated programs or GEMSIM. You specify what accuracy you require (how many significant figures) and what percentage of results you require to be this accurate and the program works out how many steps and how many subintervals are required to achieve this accuracy.
ASSERTION # Check no negative values # (all,c,COM) DVHOUS(c) >= 0 ;