non-constant-bound Loops (Yeong-Sheng Chen)
Fri, 2 Dec 1994 23:59:18 GMT

          From comp.compilers

Related articles
non-constant-bound Loops (1994-12-02)
| List of all articles for this month |

Newsgroups: comp.compilers
From: (Yeong-Sheng Chen)
Keywords: optimize, question, parallel
Organization: National Taiwan University
Date: Fri, 2 Dec 1994 23:59:18 GMT

Hi everybody,

        The following is a loop nest. Does anyone know which machines can
efficiently execute it?

        DoSequentially 200 I=1, N1
          DoAll 100 J=1, f(I)
              loop body;
    100 Continue
    200 Continue

where f(I) is a function of I.

        When max(f(I)) processors are allocated to this program, it can be
executed in maximal speed. However, it is not efficient because some
processors will be idle. So, does any system (compiler, architecture,
OS ...) provides good method to execute this program (Efficiently achieve
Maximal speed).
        Is "dynamically allocate processors to this program" a good way? Which
machine can do it? And, how about the overhead of dynamic allocation?

        Thank you.

Eric Y.S. Chen

Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.