Skip to content
GitLab
Projects
Groups
Snippets
Help
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in
Toggle navigation
elpa
Project overview
Project overview
Details
Activity
Releases
Repository
Repository
Files
Commits
Branches
Tags
Contributors
Graph
Compare
Issues
11
Issues
11
List
Boards
Labels
Service Desk
Milestones
Merge Requests
0
Merge Requests
0
Operations
Operations
Incidents
Environments
Analytics
Analytics
Repository
Value Stream
Wiki
Wiki
Members
Members
Collapse sidebar
Close sidebar
Activity
Graph
Create a new issue
Commits
Issue Boards
Open sidebar
elpa
elpa
Commits
a077abc8
Commit
a077abc8
authored
Nov 24, 2017
by
Andreas Marek
Browse files
Options
Browse Files
Download
Email Patches
Plain Diff
Add some doxygen documentation for the autotuning routines
parent
0be5ab47
Changes
3
Pipelines
1
Hide whitespace changes
Inline
Side-by-side
Showing
3 changed files
with
101 additions
and
15 deletions
+101
-15
src/elpa.F90
src/elpa.F90
+7
-1
src/elpa_api.F90
src/elpa_api.F90
+40
-13
src/elpa_impl.F90
src/elpa_impl.F90
+54
-1
No files found.
src/elpa.F90
View file @
a077abc8
...
...
@@ -208,16 +208,22 @@ module elpa
obj
=>
elpa_impl_allocate
()
end
function
!> \brief function to deallocate an ELPA instance
!> Parameters
!> \details
!> \param obj class(elpa_t), pointer : pointer to object to be destroyed and deallocated
!> \param obj class(elpa_t), pointer : pointer to
the ELPA
object to be destroyed and deallocated
subroutine
elpa_deallocate
(
obj
)
class
(
elpa_t
),
pointer
::
obj
call
obj
%
destroy
()
deallocate
(
obj
)
end
subroutine
!> \brief function to deallocate an ELPA autotune instance
!> Parameters
!> \details
!> \param obj class(elpa_autotune_t), pointer : pointer to then autotune object to be destroyed and deallocated
subroutine
elpa_autotune_deallocate
(
obj
)
class
(
elpa_autotune_t
),
pointer
::
obj
call
obj
%
destroy
()
...
...
src/elpa_api.F90
View file @
a077abc8
...
...
@@ -96,10 +96,10 @@ module elpa_api
procedure
(
elpa_can_set_i
),
deferred
,
public
::
can_set
!< method to check whether key/value can be set
! Timer
procedure
(
elpa_get_time_i
),
deferred
,
public
::
get_time
procedure
(
elpa_print_times_i
),
deferred
,
public
::
print_times
procedure
(
elpa_timer_start_i
),
deferred
,
public
::
timer_start
procedure
(
elpa_timer_stop_i
),
deferred
,
public
::
timer_stop
procedure
(
elpa_get_time_i
),
deferred
,
public
::
get_time
!< method to get the times from the timer object
procedure
(
elpa_print_times_i
),
deferred
,
public
::
print_times
!< method to print the timings tree
procedure
(
elpa_timer_start_i
),
deferred
,
public
::
timer_start
!< method to start a time measurement
procedure
(
elpa_timer_stop_i
),
deferred
,
public
::
timer_stop
!< method to stop a time measurement
! Actual math routines
generic
,
public
::
eigenvectors
=>
&
!< method eigenvectors for solving the full eigenvalue problem
...
...
@@ -132,14 +132,14 @@ module elpa_api
elpa_invert_trm_dc
,
&
elpa_invert_trm_fc
generic
,
public
::
solve_tridiagonal
=>
&
!< method to solve the eigenvalue problem for a tridiagonal
elpa_solve_tridiagonal_d
,
&
!< matrix
generic
,
public
::
solve_tridiagonal
=>
&
!< method to solve the eigenvalue problem for a tridiagonal
elpa_solve_tridiagonal_d
,
&
!< matrix
elpa_solve_tridiagonal_f
! Auto-tune
procedure
(
elpa_autotune_setup_i
),
deferred
,
public
::
autotune_setup
procedure
(
elpa_autotune_step_i
),
deferred
,
public
::
autotune_step
procedure
(
elpa_autotune_set_best_i
),
deferred
,
public
::
autotune_set_best
procedure
(
elpa_autotune_setup_i
),
deferred
,
public
::
autotune_setup
!< method to prepare the ELPA autotuning
procedure
(
elpa_autotune_step_i
),
deferred
,
public
::
autotune_step
!< method to do an autotuning step
procedure
(
elpa_autotune_set_best_i
),
deferred
,
public
::
autotune_set_best
!< method to set the best options
!> \brief These method have to be public, in order to be overrideable in the extension types
procedure
(
elpa_set_integer_i
),
deferred
,
public
::
elpa_set_integer
...
...
@@ -177,6 +177,8 @@ module elpa_api
procedure
(
elpa_solve_tridiagonal_f_i
),
deferred
,
public
::
elpa_solve_tridiagonal_f
end
type
elpa_t
!> \brief Abstract definition of the elpa_autotunet type
type
,
abstract
::
elpa_autotune_t
private
contains
...
...
@@ -200,7 +202,7 @@ module elpa_api
end
interface
!> \brief abstract definition of setup method
!> \brief abstract definition of
the ELPA
setup method
!> Parameters
!> \details
!> \param self class(elpa_t): the ELPA object
...
...
@@ -215,6 +217,13 @@ module elpa_api
end
interface
!> \brief abstract definition of the autotune setup method
!> Parameters
!> \details
!> \param self class(elpa_t): the ELPA object, which should be tuned
!> \param level integer: the level of "thoroughness" of the tuning steps
!> \param domain integer: domain (real/complex) which should be tuned
!> \result tune_state class(elpa_autotune_t): the autotuning object
abstract
interface
function
elpa_autotune_setup_i
(
self
,
level
,
domain
)
result
(
tune_state
)
import
elpa_t
,
elpa_autotune_t
...
...
@@ -226,6 +235,12 @@ module elpa_api
end
interface
!> \brief abstract definition of the autotune step method
!> Parameters
!> \details
!> \param self class(elpa_t): the ELPA object, which should be tuned
!> \param tune_state class(elpa_autotune_t): the autotuning object
!> \param unfinished logical: state whether tuning is unfinished or not
abstract
interface
function
elpa_autotune_step_i
(
self
,
tune_state
)
result
(
unfinished
)
import
elpa_t
,
elpa_autotune_t
...
...
@@ -236,7 +251,13 @@ module elpa_api
end
function
end
interface
!> \brief abstract definition of the autotune set_best method
!> Parameters
!> \details
!> \param self class(elpa_t): the ELPA object, which should be tuned
!> \param tune_state class(elpa_autotune_t): the autotuning object
!> Sets the best combination of ELPA options
abstract
interface
subroutine
elpa_autotune_set_best_i
(
self
,
tune_state
)
import
elpa_t
,
elpa_autotune_t
...
...
@@ -1214,7 +1235,10 @@ module elpa_api
end
subroutine
end
interface
!> \brief abstract definition of interface to print the autotuning state
!> Parameters
!> \param self class(elpa_autotune_t): the ELPA autotune object
abstract
interface
subroutine
elpa_autotune_print_i
(
self
)
import
elpa_autotune_t
...
...
@@ -1223,7 +1247,10 @@ module elpa_api
end
subroutine
end
interface
!> \brief abstract definition of interface to destroy the autotuning state
!> Parameters
!> \param self class(elpa_autotune_t): the ELPA autotune object
abstract
interface
subroutine
elpa_autotune_destroy_i
(
self
)
import
elpa_autotune_t
...
...
src/elpa_impl.F90
View file @
a077abc8
...
...
@@ -373,6 +373,12 @@ module elpa_impl
end
function
!> \brief function to convert a value to an human readable string
!> Parameters
!> \param self class(elpa_impl_t) the allocated ELPA object
!> \param option_name string: the name of the options, whose value should be converted
!> \param error integer: errpr code
!> \result string string: the humanreadable string
function
elpa_value_to_string
(
self
,
option_name
,
error
)
result
(
string
)
class
(
elpa_impl_t
),
intent
(
in
)
::
self
character
(
kind
=
c_char
,
len
=*
),
intent
(
in
)
::
option_name
...
...
@@ -448,8 +454,13 @@ module elpa_impl
call
c_f_pointer
(
name_p
,
name
)
call
elpa_get_double
(
self
,
name
,
value
,
error
)
end
subroutine
!> \brief function to associate a pointer with an integer value
!> Parameters
!> \param self class(elpa_impl_t) the allocated ELPA object
!> \param name string: the name of the entry
!> \result value integer, pointer: the value for the entry
function
elpa_associate_int
(
self
,
name
)
result
(
value
)
class
(
elpa_impl_t
)
::
self
character
(
*
),
intent
(
in
)
::
name
...
...
@@ -465,6 +476,13 @@ module elpa_impl
end
function
!> \brief function to querry the timing information at a certain level
!> Parameters
!> \param self class(elpa_impl_t) the allocated ELPA object
!> \param name1 .. name6 string: the string identifier for the timer region.
!> at the moment 6 nested levels can be queried
!> \result s double: the timer metric for the region. Might be seconds,
!> or any other supported metric
function
elpa_get_time
(
self
,
name1
,
name2
,
name3
,
name4
,
name5
,
name6
)
result
(
s
)
class
(
elpa_impl_t
),
intent
(
in
)
::
self
! this is clunky, but what can you do..
...
...
@@ -479,6 +497,11 @@ module elpa_impl
end
function
!> \brief function to print the timing tree below at a certain level
!> Parameters
!> \param self class(elpa_impl_t) the allocated ELPA object
!> \param name1 .. name6 string: the string identifier for the timer region.
!> at the moment 4 nested levels can be specified
subroutine
elpa_print_times
(
self
,
name1
,
name2
,
name3
,
name4
)
class
(
elpa_impl_t
),
intent
(
in
)
::
self
character
(
len
=*
),
intent
(
in
),
optional
::
name1
,
name2
,
name3
,
name4
...
...
@@ -488,6 +511,10 @@ module elpa_impl
end
subroutine
!> \brief function to start the timing of a code region
!> Parameters
!> \param self class(elpa_impl_t) the allocated ELPA object
!> \param name string: a chosen identifier name for the code region
subroutine
elpa_timer_start
(
self
,
name
)
class
(
elpa_impl_t
),
intent
(
inout
)
::
self
character
(
len
=*
),
intent
(
in
)
::
name
...
...
@@ -497,6 +524,10 @@ module elpa_impl
end
subroutine
!> \brief function to stop the timing of a code region
!> Parameters
!> \param self class(elpa_impl_t) the allocated ELPA object
!> \param name string: identifier name for the code region to stop
subroutine
elpa_timer_stop
(
self
,
name
)
class
(
elpa_impl_t
),
intent
(
inout
)
::
self
character
(
len
=*
),
intent
(
in
)
::
name
...
...
@@ -1349,6 +1380,8 @@ module elpa_impl
call
elpa_hermitian_multiply_f
(
self
,
uplo_a
,
uplo_c
,
ncb
,
a
,
b
,
nrows_b
,
&
ncols_b
,
c
,
nrows_c
,
ncols_c
,
error
)
end
subroutine
!> \brief elpa_hermitian_multiply_dc: class method to perform C : = A**H * B for double complex matrices
!> where A is a square matrix (self%na,self%na) which is optionally upper or lower triangular
!> B is a (self%na,ncb) matrix
...
...
@@ -1435,6 +1468,7 @@ module elpa_impl
ncols_b
,
c
,
nrows_c
,
ncols_c
,
error
)
end
subroutine
!> \brief elpa_hermitian_multiply_fc: class method to perform C : = A**H * B for float complex matrices
!> where A is a square matrix (self%na,self%na) which is optionally upper or lower triangular
!> B is a (self%na,ncb) matrix
...
...
@@ -2069,6 +2103,9 @@ module elpa_impl
end
subroutine
!> \brief function to destroy an elpa object
!> Parameters
!> \param self class(elpa_impl_t) the allocated ELPA object
subroutine
elpa_destroy
(
self
)
#ifdef WITH_MPI
integer
::
mpi_comm_rows
,
mpi_comm_cols
,
mpierr
...
...
@@ -2091,6 +2128,13 @@ module elpa_impl
end
subroutine
!> \brief function to setup the ELPA autotuning and create the autotune object
!> Parameters
!> \param self class(elpa_impl_t) the allocated ELPA object
!> \param level integer: the "thoroughness" of the planed autotuning
!> \param domain integer: the domain (real/complex) which should be tuned
!> \result tune_state class(elpa_autotune_t): the created autotuning object
function
elpa_autotune_setup
(
self
,
level
,
domain
)
result
(
tune_state
)
class
(
elpa_impl_t
),
intent
(
inout
),
target
::
self
integer
,
intent
(
in
)
::
level
,
domain
...
...
@@ -2112,6 +2156,11 @@ module elpa_impl
end
function
!> \brief function to do an autotunig step
!> Parameters
!> \param self class(elpa_impl_t) the allocated ELPA object
!> \param tune_state class(elpa_autotune_t): the autotuning object
!> \result unfinished logical: describes the state of the autotuning (completed/uncompleted)
function
elpa_autotune_step
(
self
,
tune_state
)
result
(
unfinished
)
implicit
none
class
(
elpa_impl_t
),
intent
(
inout
)
::
self
...
...
@@ -2151,6 +2200,10 @@ module elpa_impl
end
function
!> \brief function to set the up-to-know best options of the autotuning
!> Parameters
!> \param self class(elpa_impl_t) the allocated ELPA object
!> \param tune_state class(elpa_autotune_t): the autotuning object
subroutine
elpa_autotune_set_best
(
self
,
tune_state
)
implicit
none
class
(
elpa_impl_t
),
intent
(
inout
)
::
self
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment