Re: Memory safe implementations of C/C++

• To: Corky Cartwright <cork@rice.edu>
• Subject: Re: Memory safe implementations of C/C++
• From: Philip Wadler <wadler@research.avayalabs.com>
• Date: Fri, 04 Jan 2002 13:06:20 -0500
• cc: Philip Wadler <wadler@research.avayalabs.com>, detlefs@east.sun.com, types@cis.upenn.edu
• In-Reply-To: Message from Corky Cartwright <cork@rice.edu> of "Fri, 04 Jan 2002 11:04:01 CST." <15413.57473.492624.89320@frosty.cs.rice.edu>

Corky, thanks for the pointer to Ellis and Detlefs's work.

> I am not an expert on the ANSI Standard for C but I would be
> very surprised if the restrictions required for memory safety are
> compatible with an efficient implementation of the standard.

As I said, until now all "safe" implementations of C have been an
order of magnitude slower than "unsafe" C.  Necula and company are
only a factor of two slower (sometimes even less) because they use a
type system to discover when pointers are used in a provably safe way
that does not require overhead.  They don't quite implement ANSI C,
but it seems that their technique should extend to doing so.

-- P

• Re: Memory safe implementations of C/C++
• From: David Detlefs - Sun Microsystems Labs BOS <david.detlefs@sun.com>

• Prev: Memory safe implementations of C/C++
• Next: Re: Memory safe implementations of C/C++
• Index(es):
  • Main
  • Thread