Fix Fortran interface for FillPatch for face variables (#3541)

## Summary

The stride was wrong when copying MultiFab pointers from a Fortran array
to a Vector of Array of MultiFab pointers.

This also fixes cases when the user passes t_new followed by t_old,
instead of the other order. The issue was then teps was negative. This
was a minor issue, because the FillPatch function in C++ does not care
about the order.

## Additional background

#327 (comment)

## Checklist

The proposed changes:
- [x] fix a bug or incorrect behavior in AMReX
- [ ] add new capabilities to AMReX
- [ ] changes answers in the test suite to more than roundoff level
- [ ] are likely to significantly affect the results of downstream AMReX
users
- [ ] include documentation in the code and/or rst files, if appropriate

Src/F_Interfaces/AmrCore/AMReX_fillpatch_fi.cpp

@@ -147,11 +147,11 @@ extern "C"
         Vector<Array<MultiFab*, AMREX_SPACEDIM> > va_fmf(nf);
         for (int i = 0; i < nc; ++i) {
             for (int d = 0; d < AMREX_SPACEDIM; ++d)
-                { va_cmf[i][d] = cmf[i+d*AMREX_SPACEDIM]; }
+                { va_cmf[i][d] = cmf[i*AMREX_SPACEDIM+d]; }
         }
         for (int i = 0; i < nf; ++i) {
             for (int d = 0; d < AMREX_SPACEDIM; ++d)
-                { va_fmf[i][d] = fmf[i+d*AMREX_SPACEDIM]; }
+                { va_fmf[i][d] = fmf[i*AMREX_SPACEDIM+d]; }
         }
 
         Array<FPhysBC, AMREX_SPACEDIM> cbc{ AMREX_D_DECL( FPhysBC(cfill[0], cgeom),

Src/F_Interfaces/AmrCore/AMReX_fillpatch_mod.F90

@@ -97,12 +97,12 @@ subroutine amrex_fillpatch_single (mf, told, mfold, tnew, mfnew, geom, fill_phys
     type(c_ptr) :: smf(2)
     integer :: ns
 
-    teps = 1.e-4_amrex_real * (tnew - told)
-    if (abs(time-tnew) .lt. teps) then
+    teps = 1.e-4_amrex_real * abs(tnew - told)
+    if (abs(time-tnew) .le. teps) then
        ns = 1
        smf  (1) = mfnew%p
        stime(1) =  tnew
-    else if (abs(time-told) .lt. teps) then
+    else if (abs(time-told) .le. teps) then
        ns = 1
        smf  (1) = mfold%p
        stime(1) =  told
@@ -142,12 +142,12 @@ subroutine amrex_fillpatch_two (mf, told_c, mfold_c, tnew_c, mfnew_c, geom_c, fi
     integer :: ncrse, nfine, i
 
     ! coarse level
-    teps = 1.e-4_amrex_real * (tnew_c - told_c)
-    if (abs(time-tnew_c) .lt. teps) then
+    teps = 1.e-4_amrex_real * abs(tnew_c - told_c)
+    if (abs(time-tnew_c) .le. teps) then
        ncrse= 1
        c_mf  (1) = mfnew_c%p
        c_time(1) =  tnew_c
-    else if (abs(time-told_c) .lt. teps) then
+    else if (abs(time-told_c) .le. teps) then
        ncrse= 1
        c_mf  (1) = mfold_c%p
        c_time(1) =  told_c
@@ -160,12 +160,12 @@ subroutine amrex_fillpatch_two (mf, told_c, mfold_c, tnew_c, mfnew_c, geom_c, fi
     end if
 
     ! fine level
-    teps = 1.e-4_amrex_real * (tnew_f - told_f)
-    if (abs(time-tnew_f) .lt. teps) then
+    teps = 1.e-4_amrex_real * abs(tnew_f - told_f)
+    if (abs(time-tnew_f) .le. teps) then
        nfine= 1
        f_mf  (1) = mfnew_f%p
        f_time(1) =  tnew_f
-    else if (abs(time-told_f) .lt. teps) then
+    else if (abs(time-told_f) .le. teps) then
        nfine= 1
        f_mf  (1) = mfold_f%p
        f_time(1) =  told_f
@@ -256,14 +256,14 @@ subroutine amrex_fillpatch_two_faces(mf, told_c, mfold_c, tnew_c, mfnew_c, geom_
     end do
 
     ! coarse level
-    teps = 1.e-4_amrex_real * (tnew_c - told_c)
-    if (abs(time-tnew_c) .lt. teps) then
+    teps = 1.e-4_amrex_real * abs(tnew_c - told_c)
+    if (abs(time-tnew_c) .le. teps) then
        ncrse= 1
        c_time(1) =  tnew_c
        do dim = 1, amrex_spacedim
           c_mf(dim) = mfnew_c(dim)%p
        end do
-    else if (abs(time-told_c) .lt. teps) then
+    else if (abs(time-told_c) .le. teps) then
        ncrse= 1
        c_time(1) =  told_c
        do dim = 1, amrex_spacedim
@@ -280,14 +280,14 @@ subroutine amrex_fillpatch_two_faces(mf, told_c, mfold_c, tnew_c, mfnew_c, geom_
     end if
 
     ! fine level
-    teps = 1.e-4_amrex_real * (tnew_f - told_f)
-    if (abs(time-tnew_f) .lt. teps) then
+    teps = 1.e-4_amrex_real * abs(tnew_f - told_f)
+    if (abs(time-tnew_f) .le. teps) then
        nfine= 1
        f_time(1) =  tnew_f
        do dim = 1, amrex_spacedim
           f_mf(dim) = mfnew_f(dim)%p
        enddo
-    else if (abs(time-told_f) .lt. teps) then
+    else if (abs(time-told_f) .le. teps) then
        nfine= 1
        f_time(1) =  told_f
        do dim = 1, amrex_spacedim
@@ -353,10 +353,10 @@ subroutine amrex_fillcoarsepatch (mf, told_c, mfold_c, tnew_c, mfnew_c, &
     integer :: i
 
     ! coarse level
-    teps = 1.e-4_amrex_real * (tnew_c - told_c)
-    if (abs(time-tnew_c) .lt. teps) then
+    teps = 1.e-4_amrex_real * abs(tnew_c - told_c)
+    if (abs(time-tnew_c) .le. teps) then
        c_mf = mfnew_c%p
-    else if (abs(time-told_c) .lt. teps) then
+    else if (abs(time-told_c) .le. teps) then
        c_mf = mfold_c%p
     else
        call amrex_abort("amrex_fillcoarsepatch: how did this happen?")