Our dead code elimination phase is a simple code transformation example and in fact all the needed resources already exist in PIPS and we do not produce some new resources . However, if we try to add some new semantical analyses, some new languages features, etc. some new resources to store these information is necessary. This new resource creation process is explained in the PIPS development environment guide [8].
To avoid conflicts with already defined data structures in the PIPS sources, a resource resource is named internally DBR_RESOURCE with big caps.
Since we do not use dynamic linking and UNIX nlist() for
portability reasons, adding a new resource with the related methods
require a recompilation of PIPS . But since adding a new resource
does not occur very often and implies that a new phase is also added
or at least modified to use it and that it requires also a
recompilation, it is not a big deal.
The new resources must be added in the
$PIPS_DIR/Validation/Libs/pipsdbm library that is the core of PIPSDBM .
The methods to be used for loading, unloading, freeing and verifying
are added in the method_map[] array of the file
method_io.c. The file module.c must be modified to
reflect the unloading order of the resource hierarchy. For file
resources (used for example for prettyprinters ), the file
disk.c need also to be updated. After this modification, the
PIPSDBM library must be installed and PIPS recompiled (with
make-pips -r).
Further explanation are found in the development environment
guide [8].