[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Help-glpk] Re: Help on umalloc -VICTORY!!!

From: Zimmermann Karel
Subject: [Help-glpk] Re: Help on umalloc -VICTORY!!!
Date: Thu, 15 Jul 2004 13:19:53 +0400

Andrew Makhorin wrote:

>>First, the result of the script is :
>>sizeof(int)    = 4
>>sizeof(void *) = 8
>>INT_MAX        = 2147483647
>>So the INT_MAX  is equal to 2Go with "gcc -m64" ? is ti normal ?
>Umalloc/ucalloc used internally by glpk routines prevents from
>allocating more than INT_MAX bytes of core although glpk is able to
>work correctly with 64-bit addresses.
>To fix the problem please replace the file glplib2.c in glpk 4.4
>distribution by the one I'm attaching and recompile/reinstall the
>package as usual (do not forget to specify the option `-m64'; for
>details please see the file INSTALL in the distribution). I removed
>checks from umalloc/ucalloc/ufree, so glpsol will use all core available
>on your machine.
>>About running the gplsol and parameters, Karel Zimmermann will reply
>>to you today
>>my student has told me that the matrix of our problem is rather 
>>dense: from about 1300 variables, 1000 to 1300 are typically present in 
>>each constraint.
>>The program is just running well xith about 20000 constraints, she meets 
>>the problems  with 41000 constraints.
>In fact, the constraint matrix is very dense. However, I believe that
>glpsol with modified alloc/free routines is able to solve the instance
>with 41,000 constraints.
>Could you please report me and to <address@hidden> about your
>expirience on using glpk to solve huge lp problems in 64-bit mode?
>Your information would be extremely valueable for other glpk users.
>Best regards,
>Andrew Makhorin
Dear Andrew,
                      after the modifications (made by Christophe Caron) 
that you have indicated, it seems that it works!! (it's rather long, but 
my student has told me that before it never lasted so long time in the 
machine). She is not here today, I hope that she will confirmate the 
success tomorrow.
We use your package in the "threading" -a method that shoul predict the 
secondary structure af the proteins from their sequence.

Many thanks for your kind help. Sincerely
Karel Z.

M.Karel Zimmermann      (address@hidden)     
INRA - unité MIG (Mathématique, Informatique & Génome)                  
78352 Jouy-en-Josas cedex , France                    
Tel : +33-(0)1 34 65 28 93  Fax : +33-(0)1 34 65 29 01
                                     | |         
                                  |U-1077 \_ _ _ _    
                          ===]¤   |MIG-INRA \_|_|_/   
           \ §  -------  --- --- ---           ---------        /
          ~%%%%%~%~  ~~~~~~~~~~~~~     ~~     ~~~~   ~~~~~~~~ ~~%%%%%~%~

reply via email to

[Prev in Thread] Current Thread [Next in Thread]