1
0
mirror of https://github.com/TREX-CoE/qmckl.git synced 2024-08-25 22:51:47 +02:00

Started adding the pedagogical kernels for the HAVE_DOC builds.

This commit is contained in:
Francois Coppens 2023-02-02 17:04:34 +01:00
parent 2e45927e04
commit d3aebe52ff

View File

@ -32,7 +32,6 @@ range(2, 22)
#+end_src
* Naïve Sherman-Morrison
** ~qmckl_sherman_morrison_naive~
:PROPERTIES:
:Name: qmckl_sherman_morrison_naive
@ -90,6 +89,31 @@ range(2, 22)
* ~breakdown~ is a small number such that $0 < breakdown << 1$
* ~Slater_inv~ is allocated with $Dim \times Dim$ elements
*** Fortran source
#+begin_src f90 :tangle (eval f)
integer function qmckl_sherman_morrison_naive_doc_f &
(context, Dim, N_updates, Updates, Updates_index, breakdown, Slater_inv, determinant) &
bind(C)
use, intrinsic :: iso_c_binding
implicit none
integer (c_int64_t) , intent(in) , value :: context
integer (c_int64_t) , intent(in) , value :: Dim
integer (c_int64_t) , intent(in) , value :: N_updates
real (c_double ) , intent(in) :: Updates(N_updates*Dim)
integer (c_int64_t) , intent(in) :: Updates_index(N_updates)
real (c_double ) , intent(in) , value :: breakdown
real (c_double ) , intent(inout) :: Slater_inv(Dim*Dim)
real (c_double ) , intent(inout) :: determinant
write(*,*) "Function 'qmckl_sherman_morrison_naive_doc_f' does nothing for now..."
end function qmckl_sherman_morrison_naive_doc_f
#+end_src
*** C header
#+CALL: generate_c_header(table=qmckl_sherman_morrison_naive_args,rettyp=get_value("CRetType"),fname=get_value("Name"))
@ -108,9 +132,19 @@ range(2, 22)
double* determinant);
#+end_src
#+begin_src c :tangle (eval h_func) :comments org
qmckl_exit_code qmckl_sherman_morrison_naive_doc(
const qmckl_context context,
const uint64_t Dim,
const uint64_t N_updates,
const double* Updates,
const uint64_t* Updates_index,
const double breakdown,
double* Slater_inv,
double* determinant);
#+end_src
*** C source
#+begin_src c :tangle (eval c) :comments org
#include <stdbool.h>
@ -173,9 +207,9 @@ qmckl_exit_code qmckl_sherman_morrison_naive_hpc(
const int cui = Updates_index[l] - 1;
double den = 1.0f + C[cui];
if (fabs(den) < breakdown) {
if (fabs(den) < breakdown)
return QMCKL_FAILURE;
}
double iden = 1.0f / den;
// Update det(A)
@ -204,10 +238,6 @@ qmckl_exit_code qmckl_sherman_morrison_naive_hpc(
}
#+end_src
#+NAME:naive_template_code
#+begin_src c
static inline qmckl_exit_code qmckl_sherman_morrison_naive_{Dim}(
@ -281,10 +311,6 @@ static inline qmckl_exit_code qmckl_sherman_morrison_naive_{Dim}(
}
#+end_src
#+NAME:naive_kernel_generator
#+begin_src python :noweb yes :exports none
text="""
@ -298,10 +324,6 @@ for Dim in <<kernel_generator_range>>:
return '\n'.join(result)
#+end_src
#+NAME:naive_switch-case_generator
#+begin_src python :noweb yes :exports none
text="""
@ -322,10 +344,6 @@ for Dim in <<kernel_generator_range>>:
return '\n'.join(result)
#+end_src
#+begin_src c :tangle (eval c) :comments org :noweb yes
<<naive_kernel_generator()>>
@ -346,6 +364,7 @@ qmckl_exit_code qmckl_sherman_morrison_naive(const qmckl_context context,
NULL);
}
#ifdef HAVE_HPC
if (LDS == (1+(Dim-1)/SIMD_LENGTH)*SIMD_LENGTH) { // Most cases
switch (Dim) {
<<naive_switch-case_generator()>>
@ -363,58 +382,30 @@ qmckl_exit_code qmckl_sherman_morrison_naive(const qmckl_context context,
determinant);
}
#else
return qmckl_sherman_morrison_naive_doc(context,
Dim,
N_updates,
Updates,
Updates_index,
breakdown,
Slater_inv,
determinant);
#endif
return QMCKL_FAILURE;
}
#+end_src
*** Performance
This function performs best when there is only 1 rank-1 update in the update cycle. It is not useful to
use Sherman-Morrison with update splitting for these cycles since splitting can never resolve a situation
where applying the update causes singular behaviour.
** Fortran interface :noexport:
:PROPERTIES:
:Name: qmckl_sherman_morrison_naive
:CRetType: qmckl_exit_code
:FRetType: qmckl_exit_code
:END:
#+CALL: generate_f_interface(table=qmckl_sherman_morrison_naive_args,rettyp=get_value("FRetType"),fname=get_value("Name"))
#+RESULTS:
#+begin_src f90 :tangle (eval fh_func) :comments org :exports none
interface
integer(c_int32_t) function qmckl_sherman_morrison_naive &
(context, LDS, Dim, N_updates, Updates, Updates_index, breakdown, Slater_inv, determinant) &
bind(C)
use, intrinsic :: iso_c_binding
import
implicit none
integer (c_int64_t) , intent(in) , value :: context
integer (c_int64_t) , intent(in) , value :: LDS
integer (c_int64_t) , intent(in) , value :: Dim
integer (c_int64_t) , intent(in) , value :: N_updates
real (c_double ) , intent(in) :: Updates(N_updates*Dim)
integer (c_int64_t) , intent(in) :: Updates_index(N_updates)
real (c_double ) , intent(in) , value :: breakdown
real (c_double ) , intent(inout) :: Slater_inv(LDS*Dim)
real (c_double ) , intent(inout) :: determinant
end function qmckl_sherman_morrison_naive
end interface
#+end_src
*** Test :noexport:
The tests for the kernels are executed on datasets that are extracted from a run of QMC=Chem on Benzene (21 spin-up/21 spin down electrons) using 329 unique alpha determinants. The tests are run such that the kernels reject the computed inverse whenever the computed intermediate determinants or denominators are smaller than 1e-3. This is the default value in QMC=Chem. The tests will return QMCKL_SUCCESS whenever all the elements of the final matrix $R=S.S^-1 - 1$ are smaller than the given tolerance value of 1e-3, and will return QMCKL_FAILURE if the values are larger than this tolerance value.
The tests for the kernels are executed on datasets that are extracted from a run of
QMC=Chem on Benzene (21 spin-up/21 spin down electrons) using 329 unique alpha determinants.
The tests are run such that the kernels reject the computed inverse whenever the computed
intermediate determinants or denominators are smaller than 1e-3. This is the default value in
QMC=Chem. The tests will return QMCKL_SUCCESS whenever all the elements of the final matrix
$R=S.S^-1 - 1$ are smaller than the given tolerance value of 1e-3, and will return
QMCKL_FAILURE if the values are larger than this tolerance value.
#+begin_src c :tangle (eval c_test)
const uint64_t Dim = 21;
@ -466,8 +457,99 @@ for (unsigned int i = 0; i < Dim; i++) {
assert(rc == QMCKL_SUCCESS);
#+end_src
* Woodbury 2x2
** Performance
This function performs best when there is only 1 rank-1 update in the update cycle. It is
not useful to use Sherman-Morrison with update splitting for these cycles since splitting
can never resolve a situation where applying the update causes singular behaviour.
** C interface
#+begin_src f90 :tangle (eval f) :comments org :exports none
interface
integer(c_int32_t) function qmckl_sherman_morrison_naive_doc &
(context, Dim, N_updates, Updates, Updates_index, breakdown, Slater_inv, determinant) &
bind(C)
use, intrinsic :: iso_c_binding
import
implicit none
integer (c_int64_t) , intent(in) , value :: context
integer (c_int64_t) , intent(in) , value :: Dim
integer (c_int64_t) , intent(in) , value :: N_updates
real (c_double ) , intent(in) :: Updates(N_updates*Dim)
integer (c_int64_t) , intent(in) :: Updates_index(N_updates)
real (c_double ) , intent(in) , value :: breakdown
real (c_double ) , intent(inout) :: Slater_inv(Dim*Dim)
real (c_double ) , intent(inout) :: determinant
integer(c_int32_t), external :: qmckl_sherman_morrison_naive_doc_f
info = qmckl_sherman_morrison_naive_doc_f &
(context, Dim, N_updates, Updates, &
Updates_index, breakdown, Slater_inv, determinant)
end function qmckl_sherman_morrison_naive_doc
end interface
#+end_src
** Fortran interface :noexport:
:PROPERTIES:
:Name: qmckl_sherman_morrison_naive
:CRetType: qmckl_exit_code
:FRetType: qmckl_exit_code
:END:
#+CALL: generate_f_interface(table=qmckl_sherman_morrison_naive_args,rettyp=get_value("FRetType"),fname=get_value("Name"))
#+RESULTS:
#+begin_src f90 :tangle (eval fh_func) :comments org :exports none
interface
integer(c_int32_t) function qmckl_sherman_morrison_naive &
(context, LDS, Dim, N_updates, Updates, Updates_index, breakdown, Slater_inv, determinant) &
bind(C)
use, intrinsic :: iso_c_binding
import
implicit none
integer (c_int64_t) , intent(in) , value :: context
integer (c_int64_t) , intent(in) , value :: LDS
integer (c_int64_t) , intent(in) , value :: Dim
integer (c_int64_t) , intent(in) , value :: N_updates
real (c_double ) , intent(in) :: Updates(N_updates*Dim)
integer (c_int64_t) , intent(in) :: Updates_index(N_updates)
real (c_double ) , intent(in) , value :: breakdown
real (c_double ) , intent(inout) :: Slater_inv(LDS*Dim)
real (c_double ) , intent(inout) :: determinant
end function qmckl_sherman_morrison_naive
end interface
#+end_src
#+begin_src f90 :tangle (eval fh_func) :comments org :exports none
interface
integer(c_int32_t) function qmckl_sherman_morrison_naive_doc &
(context, Dim, N_updates, Updates, Updates_index, breakdown, Slater_inv, determinant) &
bind(C)
use, intrinsic :: iso_c_binding
import
implicit none
integer (c_int64_t) , intent(in) , value :: context
integer (c_int64_t) , intent(in) , value :: Dim
integer (c_int64_t) , intent(in) , value :: N_updates
real (c_double ) , intent(in) :: Updates(N_updates*Dim)
integer (c_int64_t) , intent(in) :: Updates_index(N_updates)
real (c_double ) , intent(in) , value :: breakdown
real (c_double ) , intent(inout) :: Slater_inv(Dim*Dim)
real (c_double ) , intent(inout) :: determinant
end function qmckl_sherman_morrison_naive_doc
end interface
#+end_src
* Woodbury 2x2
** ~qmckl_woodbury_2x2~
:PROPERTIES:
:Name: qmckl_woodbury_2x2
@ -790,7 +872,6 @@ qmckl_exit_code qmckl_woodbury_2x2(const qmckl_context context,
*** Performance
This function is most efficient when used in cases where there are only 2 rank-1 updates and
@ -860,7 +941,6 @@ assert(rc == QMCKL_SUCCESS);
#+end_src
* Woodbury 3x3
** ~qmckl_woodbury_3x3~
:PROPERTIES:
:Name: qmckl_woodbury_3x3
@ -1209,7 +1289,6 @@ qmckl_exit_code qmckl_woodbury_3x3(const qmckl_context context,
#+end_src
*** Performance...
This function is most efficient when used in cases where there are only 3 rank-1 updates and
@ -1279,7 +1358,6 @@ assert(rc == QMCKL_SUCCESS);
#+end_src
* Sherman-Morrison with update splitting
** ~qmckl_sherman_morrison_splitting~
:PROPERTIES:
:Name: qmckl_sherman_morrison_splitting
@ -1314,7 +1392,6 @@ assert(rc == QMCKL_SUCCESS);
If the determinant of the Slater-matrix is passed, it will be updated to the determinant resulting
from applying the updates to the original matrix.
*** Requirements
* ~context~ is not ~QMCKL_NULL_CONTEXT~
@ -1454,7 +1531,6 @@ assert(rc == QMCKL_SUCCESS);
#+end_src
* Woodbury 3x3 and 2x2 with Sherman-Morrison and update splitting
** ~qmckl_sherman_morrison_smw32s~
:PROPERTIES:
:Name: qmckl_sherman_morrison_smw32s