16.5.1 General concept of process continuation

Let’s think of a process in execution as having: a table of descriptors, a code memory segment, a heap memory segment and a stack memory segment, machinery continuations general. The Instruction Pointer CPU register (IP) keeps track of the next instruction to execute in the code segment; the Stack Pointer CPU register (SP) keeps track of the current location on the stack; dynamically allocated data goes in the heap segment.

|---|---|---| descriptors table
  |   |   |
  |   |   |      IP
  |   |   v      v
  |   |   |------+---------------| code segment
  |   v
  |   |----------------------| heap segment
  v
  |----------+-----------| stack segment
             ^
             SP

One way to implement continuations is to push the current IP on the stack segment, duplicate the stack segment itself, finally save somewhere the SP; later we can reinstall the saved stack segment, reset the SP and perform a ret assembly instruction that will pop the saved IP and resume the execution; machinery continuations general. Of course we must make sure that the resources referenced by the saved stack segment are not released while the continuation still exists.

|---|---|---| descriptors table
  |   |   |
  |   |   |     IP IP1
  |   |   v      v v
  |   |   |------+-+--------------| code segment
  |   v
  |   |----------------------| heap segment
  |
  v           IP1
  |----------+--+---------| stack segment
             ^
             SP

              IP1
  |----------+--+---------| duplicated stack segment
             ^
             SP

Stack segment duplication is not the way Vicare implements continuations; but the actual implementation is not conceptually different.