March, 1994 CIDER1b.1 Installation Notes ------------------------- Follow the instructions for installing SPICE3F2 found in the spice/common directory of this distribution. The instructions are in the readme file. Once you have installed SPICE, you should be somewhat familiar with the installation process. The same scripts are used to build the CIDER libraries of the distribution. Configure the conf/defaults file to correspond to your system. (Diffs between the original and modified SPICE defaults files and the original SPICE and CIDER files should help in modifying the CIDER default file.) You will also need to provide system-specific exception files (e.g. a 'mips' file), just as in SPICE. Build the CIDER source from cider/common using: util/build Install the CIDER executable and help file using: util/build install By default the CIDER executable is placed in cider//bin, and the help file is placed in the SPICE help directory. If you are confident that the installation will take place without error, you can both build and install using only the second command above. The parallel version of CIDER is built by appending the "parallel" configuration file to the above commands, i.e.: util/build install parallel On some systems, the "parallel" file conflicts with the "system" file. To compile for parallel, you will have to manually merge the two to create a specific "system-parallel" combination file before starting the build process. The serial version of CIDER has been successfully compiled, installed and run on the following systems: Ultrix 4, MIPS OSF 1.3, DEC Alpha The parallel version of CIDER has been compiled and run on a network of Ultrix 4/ MIPS workstations. A previous version, similar but not identical to the current version, has also been run successfully on a network of SUN4 machines, including some SparcStation 10's. In the past, serial-version CIDER has been run successfully on these systems: SunOS 4, Sun4 HP-UX 8.0, 9000/700 AIX V3, RS/6000 Unix SVR3, IPSC/860 node (must be installed manually.) A good rule of thumb is that if SPICE3 can be ported to the machine, then CIDER can be ported as well with little additional difficulty. However, there are two major exceptions to this: 1. SPICE3 runs on PC's and Macintoshes. A port to an MSDOS PC is probably infeasible due to memory limitations. Porting to a Windows PC or a Macintosh may be possible, but the level of difficulty of doing this is unknown. 2. Compilers that demand the ANSI C function prototypes will not compile properly. Prototypes for most of the CIDER functions would need to be generated. Running CIDER ------------- The serial version of CIDER has the same set of command-line arguments as SPICE3. The parallel version must be invoked specially however. A script has been written that automatically invokes the parallel version using the "parallel" command from the TCGMSG distribution. The syntax is either: qcider -p or: qcider In the first case, thq "qcider" script creates a process-group (".p") file containing num-nodes machines to run on. A list of available machines, one-per-line, must either be available in the current working directory in a file called "cpuservers" or in a location pointed to by the CPUSERVERS environment variable. Qcider will scan the available machines to identify those with the lightest loads, and then set up a ".p" file using those nodes. The file is called circuit-filename.suffix.p (the extension is removed). On subsequent calls to qcider this file can be reused by omitting the -p option to qcider. However, care should be taken to avoid using old .p files that correspond to machines that have become more active in the interim. Qcider will automatically store the output in "circuit-filename.suffix.raw", diagnostic output from all the machines in "circuit-filename.suffix.sum", and error messages in "circuit-filename.suffix.err". The parallel version is intended for batch use only, and the behavior if interactive use is attempted is undefined. The qcider script avoids this problem by always running in batch mode.