A two-pass compilation method is used where indicated:
   +PSFDO indicates PathScale feedback
            PASS1: -fb_create fbdata  
            PASS2: -fb_opt    fbdata
   +ACML is the AMD Core Math Library V2.0         
  
 Compilers:
    C:       pathcc  (PathScale C) unless otherwise noted
    Fortran: pathf90 (PathScale f90) unless otherwise noted
 If other compilers are used, they are indicated as:
    gcc:   Gnu C
    pgf90: PGI Fortran
 
 Floating Point base tuning:
  Fortran: pgf90 -fastsse  -Mipa=fast -Msmart
  C:       pathcc -Ofast -WOPT:mem_opnds=on +PSFDO

 Floating Point peak tuning:
 168.wupwise:  pgf90 -fastsse -Mipa=fast,inline -Msmart
 171.swim:     -Ofast -OPT:ro=3 -LNO:fusion=2:prefetch=2
 172.mgrid:    -O3 -OPT:Ofast 
               -LNO:fusion=2:blocking=off:ou_max=5:sclrze=off:prefetch=2 
               -OPT:unroll_times=8:unroll_size=256:ro=3 
               -CG:gcm=off:cflow=off
 173.applu:    -O3 -ipa 
               -LNO:fusion=2:interchange=OFF:blocking=OFF:ou_prod_max=10
               :ou_max=5:prefetch=2 -OPT:IEEE_arith=1:ro=3:unroll_size=0 
               -TENV:X=4 -WOPT:mem_opnds=on:retype_expr=on:val=0 -CG:local_fwd_sched=on
 177.mesa:     -O2 -ipa -OPT:Ofast -fno-math-errno -CG:local_fwd_sched=on +PSFDO
 178.galgel:   pgf90 -fastsse -Mipa=fast -mp +ACML
               RM_SOURCES=lapak.f90 ONESTEP
 179.art:      -O3 -OPT:Ofast -fno-math-errno -m32 +PSFDO
 183.equake:   gcc -DSPEC_CPU2000_LP64 -O3 -funroll-all-loops -ffast-math
               -finline-limit=2000 ONESTEP
 187.facerec:  basepeak=true
 188.ammp:     -O3 -OPT:alias=disjoint:unroll_times=8:Ofast:ro=3 
               -fno-math-errno -TENV:X=4 +PSFDO
 189.lucas:  pgf90 -fastsse -Mipa=fast,inline -Msmart 
 191.fma3d:  basepeak=true
 200.sixtrack: basepeak=true
 301.apsi:     -Ofast -TENV:X=4 -LNO:fusion=2:prefetch=0:blocking=off 
               -IPA:linear=on:plimit=525
 
 Portability:
 178.galgel:     -Mfixed    
   
 Notes:
      BIOS build A5S1, default setting was used.