Svadev mailing list

[John Criswell <criswell AT illinois.edu>]

On 4/1/12 5:47 AM, Baozeng wrote: Hello all,
 There is a project: Improve the performance of the run-time checks. one of  the open projects of SAFECode(http://sva.cs.illinois.edu/projects.html)
 It suggests that  the methods that SoftBound or Baggy Bounds Checking are faster. I take a look at the source tree of SAFECode and find that SoftBound and Baggy Bound sub dir  are contained in the SAFECode library. So is it mean that  they have been integrated in SAFECode?
Initial work on Baggy Bounds checking has been done, but it is not finished.  The alignment/padding transform needs to be updated to LLVM 3.0, and the run-time code needs to be thoroughly tested.

Santosh from University of Pennsylvania integrated SoftBound and CETS into the SAFECode compiler.  The implementation should be robust and complete, but I haven't tested it personally.


If yes, then why there is an openproject about this?  thanks.

They're both listed because neither BBC nor SoftBound are used to enforce SAFECode's memory safety guarantees.  SoftBound does not provide sound points-to analysis results unless CETS is enabled, and CETS adds significant overhead.  BBC does not provide sound points-to analysis or tight array bounds checking but could if it were modified to do so.

In other words, the project isn't so much about adding BBC or SoftBound to the SAFECode compiler as yet-another-option.  It is more about using the insights from the BBC and SoftBound research to make SAFECode do what it does faster.

-- John T.


--
     Best Regards,
                                                                 Baozeng Ding
                                                                 OSTG,NFS,ISCAS

_______________________________________________
svadev mailing list
svadev AT cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/svadev