|Pointer alias analysis? email@example.com (R Seshadri) (1994-02-05)|
|Re:Pointer Alias Analysis firstname.lastname@example.org (1994-02-08)|
|From:||email@example.com (Rakesh GHIYA)|
|Keywords:||optimize, analysis, bibliography, FTP|
|Date:||Tue, 8 Feb 1994 07:09:26 GMT|
firstname.lastname@example.org (R Seshadri) writes.....
>I am doing aliasing analysis on language C and Fortran 90 as part of my
>M.E thesis. Most of the aliasing analysis I have seen does not take care
>of pointers to array locations.
>For example, Aliasing output doesn't distinguish between the below
> int *a,b;
> a = &b;
> a = &b;
> If the range of arrays can be properly incorporated in the
>aliasing analysis then probably this analysis can be done. Has anybody
>done any work in this regard?
The interprocedural points-to analysis implemented in the McCAT C compiler
at McGill by Maryam Emami, distinguishes between the assignments:
a = &b and a = &b[i] where i is not statically known.
In the first case, the information is 'a definitely points-to b' while in
the second case it is 'a possibly points-to b', as pointer 'a' is not
properly aligned with the array 'b'. If pointer arithmetic is performed on
'a' , it is assumed that its target still remains within the bounds of
array 'b', but 'a' is now considered to be possibly pointing-to 'b'.
Our compiler uses the notion of abstract stack to do points-to analysis
for stack-allocated data structures, and we have two abstract stack
locations for each array: array_first to represent the first array element
and array_middle, to represent the rest of the array.
For further details, you can have a look at:
'Context-Sensitive Interprocedural Points-to Analysis in the Presence of
Function Pointers. Maryam Emami, Rakesh Ghiya and Laurie J Hendren'. ACAPS
Technical Memo 54.
It is available via anon ftp from
wally.cs.mcgill.ca:pub/doc/memos/memo54.ps.gz . It would also appear in
the Proceedings of SIGPLAN PLDI '94 Conference.
You can also look at Maryam's Master's thesis on points-to analysis:
Internet address of wally:126.96.36.199
Hope this helps.
School of Computer Science
Return to the
Search the comp.compilers archives again.