初始化大型静态数组时,为什么Fortran 90编译会花费很长时间? [英] Why does Fortran 90 compilation takes very long when large static array are initialized?
问题描述
我需要在Fortran子例程中初始化静态数组
双精度A(56136,8)
像这样:
A(1,1)= 0.999950528145
A(1,2)= 0.99982470274
A(1,3)= 0.999987006187
。
。
。
A(56136,7)= 0.933468163013
A(56136,8)= 0.0668926686049
后者是由另一个程序生成的。
使用ifort 13.0 ifort file.f -O0
进行编译需要很长时间(大约30分钟)。
第一季度:原因是什么,我该如何避免?
我对主程序没有任何处理,该子例程链接到第三方文件。该子例程经常被调用,因此不希望访问文件。
第二季度:是否可以在没有主程序的情况下将初始化放在子例程之外,从而避免每次调用子例程时都进行初始化? / p>
编辑
它是常数。在声明语句中初始化它会像这样吗?
双精度A(56136:8)= reshape(/ *
&#,#,#,#,#,#,#,#,
&#,#,#,#,#,#,#,#,#,
:
& * /,(56136,8))
这不起作用,因为有太多
我用10000个整数进行了测试,使用 DATA
语句。我的数组包含50000个整数,但想查看是否可以以10000为单位分配它们。
程序Console1
隐式无
!变量
整数A(50000)
数据A(1:10000)/& ! 1000行,每个行跟随10个数字:
2、3、5、7、11、13、17、19、23、29和&
31,37,41,43,47,53,59,61,67,71,&
73、79、83、89、97、101、103、107、109、113和&
127、131、137、139、149、151、157、163、167、173和&
179、181、191、193、197、199、211、223、227、229和&
233、239、241、251、257、263、269、271、277、281和&
...
104087,104089,104107,104113,104119,104123,104147,104149,104161,104173&
104179,104183,104207,104231,104233,104239,104243,104281,104287,104297&
104309,104311,104323,104327,104347,104369,104381,104383,104393,104399&
104417,104459,104471,104473,104479,104491,104513,104527,104537,104543&
104549,104551,104561,104579,104593,104597,104623,104639,104651,104659&
104677,104681,104683,104693,104701,104707,104711,104717,104723,104729 /
!将前10000个元素追加到其余数组
A(10001:50000)= [A(1:10000),A(1:10000),A(1:10000),A(1:10000)]
打印*,A(9998:10002)
最终程序Console1
编辑1
这是使用<$ c的方法$ c> DATA 语句和 COMMON
块可在子例程中设置值。请注意,几乎不赞成使用通用块。
子例程A_Fill ()
隐式无
整数:: A(10)
common / vals / A
data A / 1,2,3,4,5,6,7,8 ,9,10 /
结束子程序
程序Console1
隐式无
!变量
整数:: A(10)
common / vals / A
call A_Fill()
print *,A
end程序Console1
编辑2
另一种使用函数将保存的数组复制到辅助工作副本中的解决方案。
函数A_Fill()result(B)
隐式无
整数:: B(10)
整数,保存:: A(10)
数据A / 1,2,3,4,5,6,7,8,9,10 /
B = A
终端功能
程序Console1
隐式无
接口
函数A_Fill()result(B)
隐式无
整数:: B(10)
终端函数
终端接口
!变量
整数:: A(10)
A = A_Fill()
print *,A
最终程序Console1
I need to initialize an static array in a Fortran subroutine
double precision A(56136,8)
like so:
A(1,1)=0.999950528145
A(1,2)=0.99982470274
A(1,3)=0.999987006187
.
.
.
A(56136,7)=0.933468163013
A(56136,8)=0.0668926686049
The latter is generated by another program.
Compilation with ifort 13.0 ifort file.f -O0
takes very long (around 30 minutes).
Q1: What is the reason for this and how can I avoid it?
I have no handle on the main program, the subroutine is linked to third party files. The subroutine is called very often, so file access is not desirable.
Q2: Can I put the initialization outside the subroutine, without having a main program, avoiding the initialization every time the subroutine is called?
Edit
It is constant. Initializing it in the declaration statement would look like this?
double precision A(56136:8)=reshape(/*
& #, #, #, #, #, #, #, #,
& #, #, #, #, #, #, #, #,
:
& */,(56136,8))
This does not work because there are too many newlines.
I did a test with 10000 integers and it compiles in seconds when using a DATA
statement. My array holds 50000 integers but wanted to see if I can assign them in blocks of 10000.
program Console1
implicit none
! Variables
integer A(50000)
data A(1:10000) / & ! 1000 lines of 10 numbers each to follow:
2, 3, 5, 7, 11, 13, 17, 19, 23, 29, &
31, 37, 41, 43, 47, 53, 59, 61, 67, 71, &
73, 79, 83, 89, 97, 101, 103, 107, 109, 113, &
127, 131, 137, 139, 149, 151, 157, 163, 167, 173, &
179, 181, 191, 193, 197, 199, 211, 223, 227, 229, &
233, 239, 241, 251, 257, 263, 269, 271, 277, 281, &
...
104087,104089,104107,104113,104119,104123,104147,104149,104161,104173, &
104179,104183,104207,104231,104233,104239,104243,104281,104287,104297, &
104309,104311,104323,104327,104347,104369,104381,104383,104393,104399, &
104417,104459,104471,104473,104479,104491,104513,104527,104537,104543, &
104549,104551,104561,104579,104593,104597,104623,104639,104651,104659, &
104677,104681,104683,104693,104701,104707,104711,104717,104723,104729 /
! Append the first 10000 elements to the remaining array
A(10001:50000) = [A(1:10000),A(1:10000),A(1:10000),A(1:10000)]
print *, A(9998:10002)
end program Console1
Edit 1
Here is how do use a DATA
statement with a COMMON
block to set the values inside a subroutine. Note that use of common blocks is nearly deprecated.
subroutine A_Fill()
implicit none
integer :: A(10)
common /vals/ A
data A/ 1,2,3,4,5,6,7,8,9,10 /
end subroutine
program Console1
implicit none
! Variables
integer :: A(10)
common /vals/ A
call A_Fill()
print *, A
end program Console1
Edit 2
Another solution that uses a function to copy the saved array into a secondary working copy.
function A_Fill() result(B)
implicit none
integer :: B(10)
integer,save :: A(10)
data A/ 1,2,3,4,5,6,7,8,9,10 /
B = A
end function
program Console1
implicit none
interface
function A_Fill() result(B)
implicit none
integer :: B(10)
end function
end interface
! Variables
integer :: A(10)
A = A_Fill()
print *, A
end program Console1
这篇关于初始化大型静态数组时,为什么Fortran 90编译会花费很长时间?的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!