home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.arch:10492 comp.sources.wanted:4960 comp.unix.programmer:5199
- Newsgroups: comp.arch,comp.sources.wanted,comp.unix.programmer
- Path: sparky!uunet!decwrl!netsys!agate!usenet.ins.cwru.edu!magnus.acs.ohio-state.edu!zaphod.mps.ohio-state.edu!saimiri.primate.wisc.edu!doug.cae.wisc.edu!umn.edu!mmm.serc.3m.com!mmc.mmmg.com!timbuk.cray.com!equalizer!sanjay
- From: sanjay@cray.com (+Sanjay Krishnamurthy)
- Subject: Re: Tools to measure Pointer Usage
- Message-ID: <1992Nov7.000405.3016@equalizer.cray.com>
- Keywords: Pointers, loads, safety
- Organization: Cray Research Superservers Inc., San Diego CA, USA
- References: <1992Nov5.124515.26485@cs.mcgill.ca>
- Date: Sat, 7 Nov 1992 00:04:05 GMT
- Lines: 24
-
- >We are about to undertake a project to study pointer usage
- >in computer programs, and were wondering if anybody has
- >written a tool that will tell you what percentage of loads
- >are, in fact, just getting addresses for pointer dereferences...
-
- In "Proving safety of speculative load instructions at
- compile time," ESOP'92, David Bernstein, Michael Rodeh
- and Mooly Sagiv, there are some results which go way
- beyond what you are asking. They have an answer to the
- question-can you prove it is safe to execute a load
- instruction at a certain point x in the program, at
- compile time. Their method and results are relevant
- to your question because one of two ways in which a load
- becomes safe at point x is when an address is loaded
- into the register used in the load instruction along
- every execution path leading to the point x. The
- second safety criterion is when the loaded register is
- used along every execution path leading from x (in
- the same way Ken Kennedy handled safety issues dealing
- with code motion).
-
- -Sanjay M. Krishnamurthy
-
-
-