Re: MG benchmark

From: Kathy Yelick (yelick_at_EECS_dot_Berkeley_dot_EDU)
Date: Sat Jul 11 2009 - 07:32:53 PDT

  • Next message: Michael D Lewis: "remove"
    Shan,
    
    It looks like MG is written only to work with static compilation mode,  
    meaning you need to tell it the thread number at compile-time.  This  
    is in the CS267 lecture notes, although the specific flags aren't  
    there.  Here's a link that has them:
    
    http://www.psc.edu/general/software/packages/upc/
    
    Kathy
    
    On Jul 10, 2009, at 11:39 AM, Hongzhang Shan wrote:
    
    > Hi,
    >
    > I tried to compile NAS MG benchmark ( I downloaded from n2003 cvs  
    > repository) on Franklin.
    > But I got the following compiling problems.
    >
    > nid00163 NPB2.4/MG> make CLASS=C NP=4
    > make[1]: Entering directory `/u0/h/hshan/upc/NPB2.4/sys'
    > make[1]: Nothing to be done for `all'.
    > make[1]: Leaving directory `/u0/h/hshan/upc/NPB2.4/sys'
    > ../sys/setparams mg C 4
    > upcc -c -I../common  file_output.c
    > upcc: error during UPC-to-C translation (sgiupc stage):
    > In file included from file_output.c:4:
    > /u0/h/hshan/upc/NPB2.4/MG/globals.h:56: In the dynamic translation  
    > environment, THREADS must appear exactly once in declarations of  
    > shared arrays with definite block size.  Offending variable: sh_a
    > /u0/h/hshan/upc/NPB2.4/MG/globals.h:56: In the dynamic translation  
    > environment, THREADS must appear exactly once in declarations of  
    > shared arrays with definite block size.  Offending variable: sh_c
    > /u0/h/hshan/upc/NPB2.4/MG/globals.h:57: In the dynamic translation  
    > environment, THREADS must appear exactly once in declarations of  
    > shared arrays with definite block size.  Offending variable:  
    > sh_debug_vec
    > /u0/h/hshan/upc/NPB2.4/MG/globals.h:63: In the dynamic translation  
    > environment, THREADS must appear exactly once in declarations of  
    > shared arrays with definite block size.  Offending variable: jg
    > make: *** [file_output.o] Error 255
    > nid00163 NPB2.4/MG>
    >
    > Any idea how to solve it ?
    >
    > Thanks!
    >
    > shan
    

  • Next message: Michael D Lewis: "remove"