[gmx-users] Re: latest benchmarks (Itanium/Xeon/Opteron)

[Michel Cuendet]

Hi guys,

  I have the opportunity to use a brand new big itanium2 cluster. I 
compiled gromacs with the latest intel compiler and options "-ip -unroll 
-O3". I get only about 60% of the performance of an opteron 2400 on a 
single cpu. Quite disappointing when you know the price ratio of the 
machines.

  The --disable-software-sqrt option doesn't help. I noted also that 
gromacs single precision computes wrong results (box exploding 
systematically), whereas double precision doesn't take more time and 
gives more reasonable output.

  What are the best flags to use for compiling gromacs?

  Maybe the "-ipo" option for the compiler would help, but it gives the 
following error during compilation :

IPO: performing multi-file optimizations
../mdlib/.libs/libmd.a: could not read symbols: Archive has no index; 
run ranlib to add one
 
   Is there any simple way to work this out ?
   
Kind Regards,

                                                            Michel Cuendet


P.S. :  If needed, I could run benchmarks for developpers on either 
opteron+myrinet or itanium2.