6.4 Improvements To Make

• Allow users to set limits for malloc() storage.
• Prefix and make more uniform all C function, variable, and constant names. Provide a file full of #define's to provide backward compatability.
• lgcd() needs to generate at most one bignum, but currently generates more.
• divide() could use shifts instead of multiply and divide when scaling.
• Currently, dumping an executable does not preserve ports. When loading a dumped executable, disk files could be reopened to the same file and position as they had when the executable was dumped.
• Copying all of the stack is wasteful of storage. Any time a call-with-current-continuation is called the stack could be re-rooted with a frame which calls the contin just created. This in combination with checking stack depth could also be used to allow stacks deeper than 64K on the IBM PC.
• In the quest for speed, there has been some discussion about a "Forth" style Scheme interpreter.

  Provided there is still type code space available in SCM, if we devote some of the IMCAR codes to "inlined" operations, we should get a significant performance boost. What is eliminated is the having to look up a GLOC or ILOC and then dispatch on the subr type. The IMCAR operation would be dispatched to directly. Another way to view this is that we make available special form versions of CAR, CDR, etc. Since the actual operation code is localized in the interpreter, it is much easier than uncompilation and then recompilation to handle (trace car); For instance a switch gets set which tells the interpreter to instead always look up the values of the associated symbols.

• Scott Schwartz <schwartz@galapagos.cse.psu.edu> suggests: One way to tidy up the dynamic loading stuff would be to grab the code from perl5.

• VMS Dynamic Linking: Finishing the job.