Re: What's wrong with alloca() ?

Control Data's internal development language CYBIL had both an alloca type
structure and malloc type allocation within a fixed region.


The PUSH statement allocates storage which was freed on procedure exits.
I believe most implementations merely decrement the stack pointer,
providing a very low cost method of allocating structures whose size
varied at runtime.


CYBIL also had method of declaring a block of memory and allocating and
freeing that area as a distinct heap. Complementing this was the concept
of "relative pointers", basically an offset from a base location (which
was dereferenced using both the pointer and the base). I don't recall the
syntax and may be mistaken on exact semantics since all of this is from
work I did 7 years ago.


The CYBIL facilities of heap allocation within a memory block and relative
pointers allows complex data structures which include pointers to be
written to disk and read back -- just read the block which everything is
allocated within. Relative pointers don't need to be adjusted since they
don't contain absolute addresses. I assume this is what the feature was
used for.


Afraid I don't have any references, however I'm sure the CYBIL language
definition is available from Control Data.
[Relative pointers sound like PL/I OFFSET pointers, a useful feature. -John]
--

Return to the comp.compilers page.
Search the comp.compilers archives again.