Re: .f90 runs slower than .f

[Steve Kargl <sgk at troutmask dot apl dot washington dot edu>]

On Fri, Aug 03, 2012 at 11:41:07AM +0100, Rachel Dowdall wrote:

> Flat profile:
> 
> Each sample counts as 0.01 seconds.
>   %   cumulative   self              self     total           
>  time   seconds   seconds    calls   s/call   s/call  name    
>  74.18      2.27     2.27      960     0.00     0.00  psq_
>   5.56      2.44     0.17       20     0.01     0.13  __class_nrqcdpropagator_MOD_evolve
>   5.56      2.61     0.17      960     0.00     0.00  __class_nrqcdpropagator_MOD_kineticterm
>   4.25      2.74     0.13       10     0.01     0.02  __class_gaugefield_MOD_advancegaugefield
>   2.61      2.82     0.08     1442     0.00     0.00  __fouriertransf_MOD_fftw
>   1.96      2.88     0.06   184320     0.00     0.00  __class_gaugefield_MOD_calc_plaquette
>   1.96      2.94     0.06       10     0.01     0.01  __nrqcdmeson_MOD_meson1s0

> 
> Each sample counts as 0.01 seconds.
>   %   cumulative   self              self     total           
>  time   seconds   seconds    calls   s/call   s/call  name    
>  90.84      8.43     8.43      960     0.01     0.01  psq_
>   2.48      8.66     0.23      960     0.00     0.01  __class_nrqcdpropagator_MOD_kineticterm
>   1.51      8.80     0.14       10     0.01     0.03  __class_gaugefield_MOD_advancegaugefield
>   1.40      8.93     0.13       20     0.01     0.44  __class_nrqcdpropagator_MOD_evolve
>   0.75      9.00     0.07   184320     0.00     0.00  __class_gaugefield_MOD_calc_plaquette
>   0.75      9.07     0.07       10     0.01     0.01  __nrqcdmeson_MOD_meson1s0
>   0.54      9.12     0.05     1442     0.00     0.00  __fouriertransf_MOD_fftw


I can't explain the difference :(.

Compiling psq.f and psq.f90 with gfortran 4.6.4, 4.7.2, and 4.8.0
using '-O3 -S', the resulting assembly language is identical.  I
guess I'll need to download your tar archive and play with the
the complete application.


-- 
Steve