added timing facilities for LpAndItsTangent

trunk/CPFEM.f90

@@ -652,6 +652,7 @@
  logical failed
  integer(pInt) cp_en, ip, grain
  integer(pInt) iInner,dummy, i,j,k,l,m,n
+ integer(8) tick,tock,tickrate,maxticks
  real(pReal) dt, Temperature, det, p_hydro, leapfrog,maxleap
  real(pReal), dimension(6) :: Tstar_v
  real(pReal), dimension(9,9) :: dLp,dTdLp,dRdLp,invdRdLp,eye2
@@ -701,7 +702,12 @@ Inner: do              ! inner iteration: Lp
    Tstar_v = 0.5_pReal*math_mul66x6(C_66,math_mandel33to6(math_mul33x33(BT,AB)-math_I3))
    p_hydro=(Tstar_v(1)+Tstar_v(2)+Tstar_v(3))/3.0_pReal
    forall(i=1:3) Tstar_v(i) = Tstar_v(i)-p_hydro                ! subtract hydrostatic pressure
+   call system_clock(count=tick,count_rate=tickrate,count_max=maxticks)
    call constitutive_LpAndItsTangent(Lp,dLp, Tstar_v,Temperature,grain,ip,cp_en)
+   call system_clock(count=tock,count_rate=tickrate,count_max=maxticks)
+   debug_cumLpCalls = debug_cumLpCalls + 1_pInt
+   debug_cumLpTicks  = debug_cumLpTicks + tock-tick
+   if (tock < tick) debug_cumLpTicks  = debug_cumLpTicks + maxticks
    Rinner = Lpguess - Lp                                        ! update current residuum
 
    if (.not.(any(Rinner/=Rinner)) .and. &                       ! exclude any NaN in residuum

trunk/debug.f90

@@ -9,6 +9,8 @@
  integer(pInt), dimension(nCutback+1) :: debug_cutbackDistribution = 0_pInt
  integer(pInt), dimension(nInner) :: debug_InnerLoopDistribution = 0_pInt
  integer(pInt), dimension(nOuter) :: debug_OuterLoopDistribution = 0_pInt
+ integer(8)    :: debug_cumLpTicks = 0_pInt
+ integer(pInt) :: debug_cumLpCalls = 0_pInt
  logical :: debugger = .false.
  logical :: distribution_init = .false.
 
@@ -23,9 +25,19 @@
  use prec
  implicit none
 
- integer(pInt) i
+ integer(pInt) i,integral
+ integer(8) tickrate
 
+ write(6,*)
  write(6,*) 'DEBUG Info'
+ write(6,*)
+ write(6,'(a33,x,i9)')	'total calls to LpAndItsTangent  :',debug_cumLpCalls
+ if (debug_cumLpCalls > 0_pInt) then
+   call system_clock(count_rate=tickrate)
+   write(6,'(a33,x,f10.6)') 'avg CPU time/microsecs per call :',dble(debug_cumLpTicks)/tickrate/1.0e-6_pReal/debug_cumLpCalls
+   write(6,'(a33,x,i12)')   'total CPU ticks                 :',debug_cumLpTicks
+ endif
+ write(6,*)
  write(6,*)	'distribution_cutback :'
  do i=0,nCutback
    if (debug_cutbackDistribution(i+1) /= 0) write(6,*) i,debug_cutbackDistribution(i+1)
@@ -33,18 +45,26 @@
  write(6,*) 'total',sum(debug_cutbackDistribution)
  write(6,*)
  
+ integral = 0_pInt
  write(6,*)	'distribution_InnerLoop :'
  do i=1,nInner
-   if (debug_InnerLoopDistribution(i) /= 0) write(6,*) i,debug_InnerLoopDistribution(i)
+   if (debug_InnerLoopDistribution(i) /= 0) then
+     integral = integral + i*debug_InnerLoopDistribution(i)
+     write(6,*) i,debug_InnerLoopDistribution(i)
+   endif
  enddo
- write(6,*) 'total',sum(debug_InnerLoopDistribution)
+ write(6,*) 'total',sum(debug_InnerLoopDistribution),integral
  write(6,*)
  
+ integral = 0_pInt
  write(6,*)	'distribution_OuterLoop :'
  do i=1,nOuter
-   if (debug_OuterLoopDistribution(i) /= 0) write(6,*) i,debug_OuterLoopDistribution(i)
+   if (debug_OuterLoopDistribution(i) /= 0) then
+     integral = integral + i*debug_OuterLoopDistribution(i)
+     write(6,*) i,debug_OuterLoopDistribution(i)
+   endif
  enddo
- write(6,*) 'total',sum(debug_OuterLoopDistribution)
+ write(6,*) 'total',sum(debug_OuterLoopDistribution),integral
  write(6,*)
 
  END SUBROUTINE

trunk/mpie_cpfem_marc2007r1.f90

@@ -167,6 +167,8 @@
      debug_cutbackDistribution   = 0_pInt       ! initialize debugging data
      debug_InnerLoopDistribution = 0_pInt
      debug_OuterLoopDistribution = 0_pInt
+     debug_cumLpTicks = 0
+     debug_cumLpCalls = 0_pInt
    endif
  endif
  if (cptim > theTime .or. theInc /= inc) then                                   ! reached convergence

trunk/mpie_cpfem_marc2008r1.f90

@@ -143,8 +143,7 @@
  !     ** End of generated type statements **
  !
  dimension e(*),de(*),t(*),dt(*),g(*),d(ngens,*),s(*), n(2),coord(ncrd,*),disp(ndeg,*),matus(2),dispt(ndeg,*),ffn(itel,*),&
-           frotn(itel,*),strechn(itel),eigvn(itel,*),ffn1(itel,*),frotn1(itel,*),strechn1(itel),eigvn1(itel,*),kcus(2),&
-           lclass(2)
+           frotn(itel,*),strechn(itel),eigvn(itel,*),ffn1(itel,*),frotn1(itel,*),strechn1(itel),eigvn1(itel,*),kcus(2)
 
 ! Marc common blocks are in fixed format so they have to be reformated to free format (f90)
 ! Beware of changes in newer Marc versions
@@ -168,6 +167,8 @@
      debug_cutbackDistribution   = 0_pInt       ! initialize debugging data
      debug_InnerLoopDistribution = 0_pInt
      debug_OuterLoopDistribution = 0_pInt
+     debug_cumLpTicks = 0
+     debug_cumLpCalls = 0_pInt
    endif
  endif
  if (cptim > theTime .or. theInc /= inc) then                                   ! reached convergence