求助！！！！调试程序时发现诡异现象！！！！

mkingx

首先我把大致情况介绍一下，编写了一个程序，包含两个文件main.f90和checklimts.f90，
在ubuntu平台上用gfortran编译成功，调试时，插入print *,'LXX=',LXX语句，观察LXX的值如何变化，
总共插入了两条print *,'LXX=',LXX语句，分别位于循环
do i=1,nptemp
lv(i)=0.0
lm(i)=0.0
enddo

的前边和后边，输出结果如图：图片挂了，字节要求传不上。
从图中可以看出，LXX的值相差很大，不知道是为什么？循环应该没有对LXX进行任何修改。

mkingx

[Fortran] 纯文本查看 复制代码

Subroutine checklimits
  Use param
  Implicit None
  ni1 = ni + 1
  niminx = -0.5*(lxx-llx) - 0.5*dx
  nimaxx = -0.5*dx
  niminz = -0.5*dz
  nimaxz = llz - 0.5*dz
  nominx = llx - 0.5*dx
  nomaxx = 0.5*(lxx+llx) - 0.5*dx
  nominz = -0.5*dz
  nomaxz = llz - 0.5*dz
  npminx = -0.5*dx
  npmaxx = llx - 0.5*dx
  npminz = -0.5*dz
  npmaxz = llz - 0.5*dz
  n2minx = -0.5*dx
  n2maxx = -0.5*dx + 0.5*(lxx-llx)
  n3minx = llx - 0.5*dx - 0.5*(lxx-llx)
  n3maxx = llx - 0.5*dx
  n5minxx = -0.5*dx + 0.5*(lxx-llx)
  n5maxx = llx - 0.5*dx - 0.5*(lxx-llx)
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
  Do k = 1, npmax
    xptemp(k) = 0.0
    zptemp(k) = 0.0
    uptemp(k) = 0.0
    wptemp(k) = 0.0
    ctemp(k) = 0.0
  End Do
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!check前，2,3,5三个区域各自的粒子数目分别为g2，g3，g5
  qq = 0
  no1 = no + 1 !!!!!此处修改过，9月9日
  Do k = no1, np
    If (xp(k)>=n2minx .And. xp(k)<=n2maxx) Then
      qq = qq + 1
    End If
  End Do
  g2 = qq
  qq = 0
  Do k = no1, np
    If (xp(k)>=n3minx .And. xp(k)<=n3maxx) Then
      qq = qq + 1
    End If
  End Do
  g3 = qq
  qq = 0
  Do k = no1, np
    If (xp(k)>=n5minxx .And. xp(k)<=n5maxx) Then
      qq = qq + 1
    End If
  End Do
  g5 = qq
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!更正出口粒子数no，更正流体粒子数np
  notemp = no + g3 - r
  nptemp = notemp + r + g5 + g3
  n21 = notemp + 1
  n22 = notemp + r
  n51 = notemp + r + 1
  n32 = notemp + r + g5 + g3
  n41 = ni + 1
  n42 = ni + g3
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!lv和lm的初始化
  Print *, 'LXX=', lxx
  Do i = 1, nptemp
    lv(i) = 0.0
    lm(i) = 0.0
  End Do
  Print *, 'LXX=', lxx
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!重新铺设2区域粒子,用临时转换标志temp表示!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
  Do k = n21, n22
    xptemp(k) = xp(k-g3-r) + 0.5*(lxx-llx) !!!!!!!!!!!!!!!!!
    zptemp(k) = zp(k-g3-r)
    uptemp(k) = up(k-g3-r)
    wptemp(k) = wp(k-g3-r)
    ctemp(k) = c(k-g3-r)
  End Do
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!重新铺设4区域粒子!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
  b = n41
  Do k = no1, np
    If (xp(k)>=n3minx .And. xp(k)<=n3maxx) Then
      lv(b) = k
      xpp(k) = xp(k) + 0.5*(lxx-llx)
      zpp(k) = zp(k)
      upp(k) = up(k)
      wpp(k) = wp(k)
      cpp(k) = c(k)
      b = b + 1
    End If
  End Do
  Do k = n41, n42
    xptemp(k) = xpp(lv(k))
    zptemp(k) = zpp(lv(k))
    uptemp(k) = upp(lv(k))
    wptemp(k) = wpp(lv(k))
    ctemp(k) = cpp(lv(k))
  End Do
!!!!!!!!!!!!!!!!!!!!!!!!!!!!
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!重新铺设53区域粒子!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
  b = n51
  Do k = no1, np
    If (xp(k)>=n5minxx .And. xp(k)<=n3maxx) Then
      lm(b) = k
      xpp(k) = xp(k)
      zpp(k) = zp(k)
      upp(k) = up(k)
      wpp(k) = wp(k)
      cpp(k) = c(k)
      b = b + 1
    End If
  End Do
  Do k = n51, n32
    xptemp(k) = xpp(lm(k))
    zptemp(k) = zpp(lm(k))
    uptemp(k) = upp(lm(k))
    wptemp(k) = wpp(lm(k))
    ctemp(k) = cpp(lm(k))
  End Do
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
  no = notemp
  np = nptemp
  Do k = ni1, np
    xp(k) = xptemp(k)
    zp(k) = zptemp(k)
    up(k) = uptemp(k)
    wp(k) = wptemp(k)
    c(k) = ctemp(k)
  End Do
End Subroutine checklimits

checklimits.f90  的源代码

mkingx

main.f90的源代码

[Fortran] 纯文本查看 复制代码

Module param
  Implicit None
  Integer :: i, j, no1, np, kind_p2, lz, ncz, lx, ncx, lz2, nct, ii, jj, qq, no, g2, g3, g5, notemp, nptemp, r, nfile
  Integer :: n21, n22, n51, n32, n41, n42, ni, k, b, icell, kcell, nn, nnx, nnz, nbfd, nnn, nbfu, jj_start, time1
  Integer, Parameter :: npmax = 15000, kindp = 2, nct_max = 3000, nplink_max = 120, b1max = 100, b2max = 9940
  Integer :: lv(b1max), lm(b2max), iflag(npmax), ibox(nct_max, 2, nplink_max), nc(nct_max, kindp)
  Double Precision :: ax(npmax), az(npmax), ac(npmax), udot(npmax), wdot(npmax)
  Double Precision :: cdot(npmax), xp(npmax), zp(npmax), up(npmax)
  Double Precision :: wp(npmax), c(npmax), mass(npmax), xpp(b2max), zpp(b2max), upp(b2max), wpp(b2max)
  Double Precision :: xptemp(npmax), zptemp(npmax), uptemp(npmax), wptemp(npmax), ctemp(npmax)
  Double Precision :: cpp(b2max), xdot(npmax), zdot(npmax), xo(npmax), zo(npmax), uo(npmax), wo(npmax), co(npmax)
  Double Precision :: etap, d, drx, drz, rr2, fourh2, dux, duz, dcij, rr, hsml, tdwdr, eta, rho, niminx, nimaxx, niminz, nimaxz, nominx
  Double Precision :: nomaxx, nominz, nomaxz, npminx, npmaxx, npminz, npmaxz, n2minx, n2maxx, n3minx, n3maxx, n5minxx, n5maxx
  Double Precision :: xmin_ini, zmin_ini, xmax_ini, zmax_ini, one_over_2h, one_over_h, tw, factor, pi, g, time, dt, dt2, tmax
  Double Precision :: ncall11, ncall12, ncall13, h2
  Integer :: j1, j2, kind_p1, ini_kind_p2, lx2, ni1, n_start, n_end, kind_p
  Double Precision :: q
  Double Precision :: llx, llz, dx, dz, lxx, lzz, ddx, ddz
End Module param
Program sph
  Use param
  Implicit None
  ni = 4036
  ni1 = 4037
  llx = 2.5E-2
  llz = 2.5E-4
  lxx = 2.5E-2 + 4*2.5E-5
  lzz = 2.5E-4 + 4*2.5E-5
  dx = 2.5E-5
  dz = 2.5E-5
  no = 4056
  np = 14056
  r = 20
  k = 0
  Do i = 1, 14
    Do j = 1, 1004
      k = k + 1
      xp(nn) = (j-1)*dx - 0.5*(lxx-llx)
      zp(nn) = (i-1)*dz - 0.5*(lzz-llz)
      up(nn) = 0.0
      wp(nn) = 0.0
      c(nn) = 0.0
    End Do
  End Do
  Call checklimits
End Program sph

fcode

1.图片可以压缩后上传。2MB的要求应该挺宽裕。
2.你的程序里有很多的越界。建议开启 -fcheck 开关编译。
3.比如，主程序里的 nn 没有值。
4.比如，子程序里，nptemp 的值远远大于 lv 的上限 b1max=100
第四个问题可能是造成错误的根源。

mkingx

问题解决了，就是第四个的问题。真心感受到fcode的强大，纠结了一晚上的问题，就这样轻松的解决了。
我还有一点好奇的是，这个程序在ubuntu编译gfortran main.f90 chencklimits.f90 -o tes,  是编译成功了的，
是不是开启了-fcheck，编译就不会通过并且说明数组越界？是不是在windows上的ivf编译就会提示数组越界？

楚香饭

编译应该会通过，越界是属于运行时错误。
ivf 默认开启越界检查，而 gfortran 默认不开启。这只是默认行为。但是不得不说，ivf 的运行时错误提示，比gfortran清晰。

mkingx

fcode 发表于 2014-9-10 23:13
1.图片可以压缩后上传。2MB的要求应该挺宽裕。
2.你的程序里有很多的越界。建议开启 -fcheck 开关编译。
3. ...

-fcheck怎么开启？？

楚香饭

在命令行里加入就可以了。

如果你用的 ide，则要看什么 ide ，进行什么操作了