| 1 |
|
| 2 |
module active_module |
| 3 |
use w2f__types |
| 4 |
implicit none |
| 5 |
private |
| 6 |
public :: active, saxpy, sax, setderiv, zero_deriv, & |
| 7 |
&convert_p2a_scalar, convert_a2p_scalar, & |
| 8 |
&convert_p2a_vector, convert_a2p_vector, & |
| 9 |
&convert_p2a_matrix, convert_a2p_matrix |
| 10 |
|
| 11 |
|
| 12 |
! |
| 13 |
! active needs to be a sequence type |
| 14 |
! with no initialization |
| 15 |
! |
| 16 |
type active |
| 17 |
sequence |
| 18 |
double precision :: v |
| 19 |
! initialization does not work for active variables |
| 20 |
! inside of common block, such as in boxmodel |
| 21 |
! initialization is required for correct adjoint |
| 22 |
double precision :: d=0.0 |
| 23 |
! double precision :: d |
| 24 |
end type active |
| 25 |
|
| 26 |
interface saxpy |
| 27 |
module procedure saxpy_a_a |
| 28 |
end interface |
| 29 |
|
| 30 |
interface setderiv |
| 31 |
module procedure setderiv_a_a |
| 32 |
end interface |
| 33 |
|
| 34 |
interface zero_deriv |
| 35 |
module procedure zero_deriv_a |
| 36 |
end interface |
| 37 |
|
| 38 |
interface sax |
| 39 |
module procedure sax_d_a_a, sax_i_a_a |
| 40 |
end interface |
| 41 |
|
| 42 |
interface convert_p2a_scalar |
| 43 |
module procedure convert_p2a_scalar_impl |
| 44 |
end interface |
| 45 |
interface convert_a2p_scalar |
| 46 |
module procedure convert_a2p_scalar_impl |
| 47 |
end interface |
| 48 |
interface convert_p2a_vector |
| 49 |
module procedure convert_sp2a_vector_impl |
| 50 |
end interface |
| 51 |
interface convert_a2p_vector |
| 52 |
module procedure convert_a2sp_vector_impl |
| 53 |
end interface |
| 54 |
interface convert_p2a_matrix |
| 55 |
module procedure convert_sp2a_matrix_impl |
| 56 |
end interface |
| 57 |
interface convert_a2p_matrix |
| 58 |
module procedure convert_a2sp_matrix_impl |
| 59 |
end interface |
| 60 |
|
| 61 |
contains |
| 62 |
|
| 63 |
! |
| 64 |
! chain rule saxpy to be used in forward and reverse modes |
| 65 |
! |
| 66 |
|
| 67 |
subroutine saxpy_a_a(a,x,y) |
| 68 |
double precision, intent(in) :: a |
| 69 |
type(active), intent(in) :: x |
| 70 |
type(active), intent(inout) :: y |
| 71 |
|
| 72 |
y%d=y%d+x%d*a |
| 73 |
end subroutine saxpy_a_a |
| 74 |
|
| 75 |
! |
| 76 |
! chain rule saxpy to be used in forward and reverse modes |
| 77 |
! derivative component of y is equal to zero initially |
| 78 |
! note: y needs to be inout as otherwise value component gets |
| 79 |
! zeroed out |
| 80 |
! |
| 81 |
|
| 82 |
subroutine sax_d_a_a(a,x,y) |
| 83 |
double precision, intent(in) :: a |
| 84 |
type(active), intent(in) :: x |
| 85 |
type(active), intent(inout) :: y |
| 86 |
|
| 87 |
y%d=x%d*a |
| 88 |
end subroutine sax_d_a_a |
| 89 |
|
| 90 |
subroutine sax_i_a_a(a,x,y) |
| 91 |
integer(kind=w2f__i8), intent(in) :: a |
| 92 |
type(active), intent(in) :: x |
| 93 |
type(active), intent(inout) :: y |
| 94 |
|
| 95 |
y%d=x%d*a |
| 96 |
end subroutine sax_i_a_a |
| 97 |
|
| 98 |
! |
| 99 |
! set derivative of y to be equal to derivative of x |
| 100 |
! note: making y inout allows for already existing active |
| 101 |
! variables to become the target of a derivative assignment |
| 102 |
! |
| 103 |
|
| 104 |
subroutine setderiv_a_a(y,x) |
| 105 |
type(active), intent(inout) :: y |
| 106 |
type(active), intent(in) :: x |
| 107 |
|
| 108 |
y%d=x%d |
| 109 |
end subroutine setderiv_a_a |
| 110 |
|
| 111 |
! |
| 112 |
! set derivative components to 0.0 |
| 113 |
! |
| 114 |
subroutine zero_deriv_a(x) |
| 115 |
type(active), intent(out) :: x |
| 116 |
|
| 117 |
x%d=0.0d0 |
| 118 |
end subroutine zero_deriv_a |
| 119 |
|
| 120 |
subroutine convert_a2p_scalar_impl(convertTo, convertFrom) |
| 121 |
double precision, intent(out) :: convertTo |
| 122 |
type(active), intent(in) :: convertFrom |
| 123 |
convertTo=convertFrom%v |
| 124 |
end subroutine |
| 125 |
|
| 126 |
subroutine convert_p2a_scalar_impl(convertTo, convertFrom) |
| 127 |
double precision, intent(in) :: convertFrom |
| 128 |
type(active), intent(out) :: convertTo |
| 129 |
convertTo%v=convertFrom |
| 130 |
end subroutine |
| 131 |
|
| 132 |
subroutine convert_a2sp_vector_impl(convertTo, convertFrom) |
| 133 |
type(active), dimension(:), intent(in) :: convertFrom |
| 134 |
real, dimension(:), intent(out) :: convertTo |
| 135 |
integer i |
| 136 |
do i=lbound(convertFrom,1),ubound(convertFrom,1) |
| 137 |
convertTo(i)=convertFrom(i)%v |
| 138 |
end do |
| 139 |
end subroutine |
| 140 |
|
| 141 |
subroutine convert_sp2a_vector_impl(convertTo, convertFrom) |
| 142 |
real, dimension(:), intent(in) :: convertFrom |
| 143 |
type(active), dimension(:), intent(out) :: convertTo |
| 144 |
integer i |
| 145 |
do i=lbound(convertFrom,1),ubound(convertFrom,1) |
| 146 |
convertTo(i)%v=convertFrom(i) |
| 147 |
end do |
| 148 |
end subroutine |
| 149 |
|
| 150 |
subroutine convert_a2sp_matrix_impl(convertTo, convertFrom) |
| 151 |
type(active), dimension(:,:), intent(in) :: convertFrom |
| 152 |
real, dimension(:,:), intent(out) :: convertTo |
| 153 |
integer i,j |
| 154 |
do i=lbound(convertFrom,1),ubound(convertFrom,1) |
| 155 |
do j=lbound(convertFrom,2),ubound(convertFrom,2) |
| 156 |
convertTo(i,j)=convertFrom(i,j)%v |
| 157 |
end do |
| 158 |
end do |
| 159 |
end subroutine |
| 160 |
|
| 161 |
subroutine convert_sp2a_matrix_impl(convertTo, convertFrom) |
| 162 |
real, dimension(:,:), intent(in) :: convertFrom |
| 163 |
type(active), dimension(:,:), intent(out) :: convertTo |
| 164 |
integer i,j |
| 165 |
do i=lbound(convertFrom,1),ubound(convertFrom,1) |
| 166 |
do j=lbound(convertFrom,2),ubound(convertFrom,2) |
| 167 |
convertTo(i,j)%v=convertFrom(i,j) |
| 168 |
end do |
| 169 |
end do |
| 170 |
end subroutine |
| 171 |
|
| 172 |
end module |
| 173 |
|
| 174 |
|
Parent Directory
| 
Revision Log
| 
Revision Graph