From: Hans Boldt
I suppose my point is that the need for wider pointers is just one of the
disadvantages of single-level store.
You argue that wide pointers add overhead, consume more memory, and negatively impact performance. But Frank Soltis asserts that SLS significantly improves performance in multi-user, multi-tasking settings.
Take the idea of multiple users running the same application, but each instance of the application in a separate process. There may be many jobs running, but never more than one instance of application code and data in memory under SLS, which IIRC greatly reduces the conventional overhead of each process allocating its own space for code and data, and the conventional overhead of switching between jobs.
That may not appeal particularly to those who may be deploying applications under a single J2EE application server. It may not appeal to those who may be deploying applications on one machine and a DBMS on another.
But that REALLY appeals to those of us who use native languages and interfaces that take advantage of that architecture.