problem about Fortran 77 in unix - 未名空间MITBBS历史存档

国际科技财经博客移民网络热点娱乐民生时事公众号

Redian新闻

>未名空间

>Computation - 科学计算

problem about Fortran 77 in unix

problem about Fortran 77 in unix# Computation - 科学计算

w*d2003-10-08 07:10

1 楼

Anybody knows how to write a Fortran 77 code runnable under
compaq fortran and sun-workstation?
Our dept bought two compaq machines. They are oveloaded. So I tried
to run my code under a sun-workstation. It gives me "bus error".
It really puzzles me. Thanks in advance.

n*c2003-10-08 07:10

2 楼

Oh, man, did you compile your source code on the sun machine?
Fortran does not seem to have many troubles working in different
platforms, if the compilcation has passed.

【在 w**d 的大作中提到】

: Anybody knows how to write a Fortran 77 code runnable under
: compaq fortran and sun-workstation?
: Our dept bought two compaq machines. They are oveloaded. So I tried
: to run my code under a sun-workstation. It gives me "bus error".
: It really puzzles me. Thanks in advance.

w*d2003-10-08 07:10

3 楼

I did recompile all my codes. Only a couple of warnings about
nonused local variables. I dont think it is the problem.
Thanks anyway.

【在 n*c 的大作中提到】

: Oh, man, did you compile your source code on the sun machine?
: Fortran does not seem to have many troubles working in different
: platforms, if the compilcation has passed.

f*r2003-10-08 07:10

4 楼

You'd better check your array size. It sounds like an out-of-boundary error.

【在 w**d 的大作中提到】

: I did recompile all my codes. Only a couple of warnings about
: nonused local variables. I dont think it is the problem.
: Thanks anyway.

w*d2003-10-08 07:10

5 楼

The array is small. The codes only take around 20M memory.
I found out what caused the problem, but don't know why.
Whenever I declare an array like a(0:node) in a subroutine
with node being passed from other subroutines or main, there will
be bus error.
But if I declare a(node) or a(0:node) with node being given in the
current subroutine, it works fine.

【在 f****r 的大作中提到】

: You'd better check your array size. It sounds like an out-of-boundary error.

w*d2003-10-08 07:10

6 楼

I did the test with the code written a few years ago. It gives the same probelm
. But it worked before. So weird.

【在 w**d 的大作中提到】

: The array is small. The codes only take around 20M memory.
: I found out what caused the problem, but don't know why.
: Whenever I declare an array like a(0:node) in a subroutine
: with node being passed from other subroutines or main, there will
: be bus error.
: But if I declare a(node) or a(0:node) with node being given in the
: current subroutine, it works fine.

w*d2003-10-08 07:10

7 楼

Any comments?

【在 w**d 的大作中提到】

: I did the test with the code written a few years ago. It gives the same probelm
: . But it worked before. So weird.

x*y2003-10-08 07:10

8 楼

It's weird. I also have array declared as a(0:node) and passed around several
subroutines. It works fine on sun machine.

probelm

【在 w**d 的大作中提到】

: Any comments?

w*d2003-10-08 07:10

9 楼

it IS weird. A few years ago, there was no such problem. I don't know if
they change anything on those machines.

【在 x*y 的大作中提到】

: It's weird. I also have array declared as a(0:node) and passed around several
: subroutines. It works fine on sun machine.
:
: probelm

l*g2003-10-08 07:10

10 楼

"If a program gets a bus error (SIGBUS), it usually has some problems with
misaligned data."
可能是caller传递的数组的类型和subroutine中的数组a的类型不一致引起的。
比如说前者是byte，a是word.如果a(i)的地址没对齐成奇数就有bus error。
这大概是自动数组a(0:node)和a(1:node)结果不同的原因。
去年为什么可以？也许今年f77升级，default不一样了；
或者有其它影响complier行为的原因。参考：
Bus Error--Finding the Line Number
http://docs.sun.com/db/doc/802-2997/6i6u2n7rt?a=view#07.Debugging-27

several

【在 w**d 的大作中提到】

: it IS weird. A few years ago, there was no such problem. I don't know if
: they change anything on those machines.

w*d2003-10-08 07:10

11 楼

I asked the same problem on a Fortran newsgroup. Somebody suspected it might
be caused by a compiler bug. I don't know what that means.
That guy also thinks I used F90 compiler. He said the automatic array is
illegal in F77.
Then another guy suggests me to do as follows:
program main
real*8 a(0:10)
call sub(a)
end
subroutine sub(b)
real*8 b(*)
b(0) = 7
end
It is just like those subroutines in Lapack. But there are lots of
work arrays in my code, I don't know how realistic this way is.

【在 l**g 的大作中提到】

: "If a program gets a bus error (SIGBUS), it usually has some problems with
: misaligned data."
: 可能是caller传递的数组的类型和subroutine中的数组a的类型不一致引起的。
: 比如说前者是byte，a是word.如果a(i)的地址没对齐成奇数就有bus error。
: 这大概是自动数组a(0:node)和a(1:node)结果不同的原因。
: 去年为什么可以？也许今年f77升级，default不一样了；
: 或者有其它影响complier行为的原因。参考：
: Bus Error--Finding the Line Number
: http://docs.sun.com/db/doc/802-2997/6i6u2n7rt?a=view#07.Debugging-27
:

l*g2003-10-08 07:10

12 楼

1。这里应该有错：
real*8 b(*) == real*8 b(1:)
所以这里 b(0) 是 bound violation，a(0)的值并不会改变。
2。我觉得你的程序是不是不自觉地引用指针，引起内存访问出错。比如
subroutine sub(b1,N)
real*8 b1(0:N)
b(0) = 7
end
subroutine sub(N)
real*8 b2(0:N)
b(0) = 7
end
其中b1是双精度数组；但fortran把b2解释为local指针数组，指向double.
前者是对的，后者就会出错。
命令g77/f77执行的可能是f90 或 f95

if

【在 w**d 的大作中提到】

: I asked the same problem on a Fortran newsgroup. Somebody suspected it might
: be caused by a compiler bug. I don't know what that means.
: That guy also thinks I used F90 compiler. He said the automatic array is
: illegal in F77.
: Then another guy suggests me to do as follows:
: program main
: real*8 a(0:10)
: call sub(a)
: end
: subroutine sub(b)

w*d2003-10-08 07:10

13 楼

Some guy suggested this. But I did not try it.
The problem is that the same code worked before. And it also runs well
on Compaq machines. Right now, I am busy doing other stuff. Later, I will
try to figure it out.
This is possible.

【在 l**g 的大作中提到】

: 1。这里应该有错：
: real*8 b(*) == real*8 b(1:)
: 所以这里 b(0) 是 bound violation，a(0)的值并不会改变。
: 2。我觉得你的程序是不是不自觉地引用指针，引起内存访问出错。比如
: subroutine sub(b1,N)
: real*8 b1(0:N)
: b(0) = 7
: end
: subroutine sub(N)
: real*8 b2(0:N)