一个数组运算的问题

pasuka

andy8496 发表于 2017-8-6 18:30
这个是实际应用的需要。

症结即在于此呀，没有看出来?

andy8496

pasuka 发表于 2017-8-6 21:28
症结即在于此呀，没有看出来?

真看不出来。我试验用的那段代码，咋就能按照我的理解运行呢？索引数组也是有重复元素的：
program main
implicit none
   integer :: ia(5),ib(9)
   ia = [1,5,5,5,2]
   ib = 0
   ib(ia) = ib(ia) + [1,1,1,1,1]
   write(*,*) ib
   stop
end program

chiangtp

a many-one array section must NOT Assign To NOR Read In
Compiler may not Check at Runtime owing to the Performance Cost

andy8496

好吧，总算看到一点依据了。虽然心中不服，比如：五楼的大哥的方法为什么又可行？
还有两点疑问，还要继续麻烦大家指点：
chiangtp附件中提到的，
①a many-one array section must NOT Assign To NOR Read In Compiler may not Check at Runtime owing to the Performance Cost
VS+IVF中，Debug中如何打开这个检查？
②the use of vector subscripting is very inefficient and should not be used unless absolutely necessary (although easy to coding)
这句话该如何理解？数据片段只是写起来方便，但是运行效率低？

pasuka

andy8496 发表于 2017-8-6 21:40
真看不出来。我试验用的那段代码，咋就能按照我的理解运行呢？索引数组也是有重复元素的：
program main
  ...

左端赋值包含重复元素的切片这样的写法本身就是不正确的，起码IBM XL和Intel Fortran的用户手册都明确强调这点
传送门1： http://w3.pppl.gov/~hammett/comp/f90tut/f90.tut5.html
Note that a vector subscript with duplicate values cannot appear on the left-hand side of an assignment as it would be ambiguous. Thus,
       b( (/ 1, 7, 3, 7 /) ) = (/ 1, 2, 3, 4 /)
is illegal.
传送门2： https://www.ibm.com/support/know ... ectorsubscript.html
An array section with a vector subscript in which two or more elements of the vector subscript have the same value is called a many-one section. Such a section must not:
Appear on the left side of the equal sign in an assignment statement传送门3： https://software.intel.com/en-us/node/678553

An array section with a vector subscript that has two or more elements with the same value is called a many-one array section. For example:

  REAL A(3, 3), B(4)  INTEGER K(4)  ! Vector K has repeated values  K = (/3, 1, 1, 2/)  ! Sets all elements of A to 5.0  A = 5.0  B = A(3, K)

The array section A(3,K) consists of the elements:

  A(3, 3)  A(3, 1)  A(3, 1)  A(3, 2)

A many-one section must not appear on the left of the equal sign in an assignment statement, or as an input item in a READ statement.

chiangtp

1. 15F, posuka見多識廣, 佩服
2. whole array的assignment(多個要做的規則), 定義上是in-any-order+parallel(如果能), runtime時 "五楼的大哥的方法", "Debug打开检查" 都是compiler/compile-time-options/... dependent
3. "vector subscripting is very inefficient "應可理解(?), 我沒有理想的實證

chiangtp

抱歉, pasuka誤為posuka

pasuka

再补充一些内容
Fortran 2003 handbook的第6章，第184页如是说：

A many-one section must not appear on the left of the equal sign in an assignment statement or as an input item in a READ statement. The reason is that the result will depend on the order of evaluation of the subscripts, which is not specified by the language. The results would not be predictable and the program containing such a statement would not be portable.

也即这个属于Fortran规范的盲点，lz有心的话，不妨邮件咨询RAL的J K Reid爵士或者I Duff
btw，无需见多识广，反倒是科学上网很重要

andy8496

多谢各位了！
现在补救的办法就是依赖编译器检查了，不知道可行否？在哪里设置？

“
2. whole array的assignment(多個要做的規則), 定義上是in-any-order+parallel(如果能), runtime時 "五楼的大哥的方法", "Debug打开检查" 都是compiler/compile-time-options/... dependent
”
这个设置我没找到在哪儿……

pasuka

andy8496 发表于 2017-8-7 14:52
多谢各位了！
现在补救的办法就是依赖编译器检查了，不知道可行否？在哪里设置？

直接改回方法1不行吗？
在ia index数组包含重复元素且不能剔除的前提下，IBM XL和Intel Fortran编译器手册均明确对方法2亮红牌，这下总得遵守规则被罚下场吧？
仍然不服气的话，建议研读g95或gfortran前端代码，自行实现需要的功能。g95的代码值得看看，既能加深Fortran的了解，也能助益C语言进阶