home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.parallel
- Path: sparky!uunet!gatech!hubcap!fpst
- From: karent@quip.eecs.umich.edu (Karen A. Tomko)
- Subject: Scientific programs with conflicting memory strides
- Message-ID: <1992Sep15.172707.12520@hubcap.clemson.edu>
- Apparently-To: comp-parallel@archive.cis.ohio-state.edu
- Sender: news@eecs.umich.edu (Mr. News)
- Organization: University of Michigan EECS Dept., Ann Arbor, MI
- Date: Tue, 15 Sep 1992 17:02:08 GMT
- Approved: parallel@hubcap.clemson.edu
- Lines: 42
-
- *WANTED*
-
- Scientific Fortran Programs to optimize
-
- I am looking for programs with conflicting memory
- strides to verify an automatic restructuring tool.
-
- For example:
-
- do i = 1, N/k
- ... = A(i) ... A(k*i)
- enddo
-
- Where the array A has stride 1 and stride k (small integer)
- accesses within the same loop nest.
-
-
- or
-
- real A2(3,N), A1(3*N)
- equivalence (A2(1,1), A1(1))
- .
- .
- .
-
- do i = 1, N
- ... = A2(2,i) ... A1(i)
- enddo
-
- Were the data of arrays A2 and A1 is accessed with a stride of
- 3 and a stride of 1.
-
- If you have programs which have loops similar to the ones above and
- can give me an electronic copy of your source file please send email.
-
- I have been working on techniques for optimizing these types of accesses
- for shared memory multiprocessors and am looking for some 'real' examples.
-
- Karen Tomko
- karent@eecs.umich.edu
- University of Michigan
-
-