TREX/qmckl
1
0
Fork 0
mirror of https://github.com/TREX-CoE/qmckl.git synced 2025-01-03 10:06:09 +01:00
Code Issues Projects Releases Wiki Activity

Fix bug in HPC AOs

Browse Source
This commit is contained in:
Anthony Scemama 2022-02-19 19:24:18 +01:00
parent 73399e24ec
commit d919c53c42
2 changed files with 124 additions and 89 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
configure.ac
View File

@ -222,13 +222,13 @@ AC_ARG_ENABLE(debug, [AS_HELP_STRING([--enable-debug],[compile for debugging])],
if test "$ok" = "yes"; then
if test "$GCC" = "yes"; then
CFLAGS="$CFLAGS \
-Wall -W -Wbad-function-cast -Wcast-qual -Warray-bounds -Wdisabled-optimization \
-g -Wall -W -Wbad-function-cast -Wcast-qual -Warray-bounds -Wdisabled-optimization \
-Wpointer-arith -Wcast-align -Wpedantic -Wextra -Walloc-zero -Werror \
"
fi
if test "$GFC" = "yes"; then
FCFLAGS="$FCFLAGS \
-fcheck=all -Waliasing -Wampersand -Wconversion \
-g -fcheck=all -Waliasing -Wampersand -Wconversion \
-Wsurprising -ffpe-trap=zero,overflow,underflow \
-Wintrinsics-std -Wno-tabs -Wintrinsic-shadow -Wline-truncation \
-Wreal-q-constant -Wuninitialized -fbacktrace -finit-real=nan"

209
org/qmckl_ao.org
View File

@ -3154,7 +3154,7 @@ assert(0 == test_qmckl_ao_gaussian_vgl(context));
const double* coord,
const double* nucl_coord,
const double* expo,
double* const primitive_vgl );
double* const primitive_vgl );
#+end_src
@ -3493,7 +3493,7 @@ for (j=0 ; j<point_num ; ++j) {
const double* nucl_coord,
const double* expo,
const double* coef_normalized,
double* const shell_vgl );
double* const shell_vgl );
#+end_src
#+begin_src f90 :comments org :tangle (eval f) :noweb yes
@ -4149,7 +4149,7 @@ assert(0 == test_qmckl_ao_power(context));
int32_t* const L,
const int64_t ldl,
double* const VGL,
const int64_t ldv );
const int64_t ldv );
#+end_src
#+begin_src c :tangle (eval c) :comments org
@ -4169,7 +4169,7 @@ qmckl_ao_polynomial_vgl (const qmckl_context context,
return qmckl_ao_polynomial_vgl_doc (context, X, R, lmax, n, L, ldl, VGL, ldv);
#else
return qmckl_ao_polynomial_vgl_doc (context, X, R, lmax, n, L, ldl, VGL, ldv);
#endif
#endif
}
#+end_src
@ -4394,7 +4394,7 @@ end function qmckl_ao_polynomial_vgl_doc_f
int32_t* const L,
const int64_t ldl,
double* const VGL,
const int64_t ldv );
const int64_t ldv );
#+end_src
#+CALL: generate_c_header(table=qmckl_ao_polynomial_vgl_args,rettyp=get_value("CRetType"),fname="qmckl_ao_polynomial_transp_vgl_doc")
@ -4410,7 +4410,7 @@ end function qmckl_ao_polynomial_vgl_doc_f
int32_t* const L,
const int64_t ldl,
double* const VGL,
const int64_t ldv );
const int64_t ldv );
#+end_src
#+CALL: generate_c_header(table=qmckl_ao_polynomial_vgl_args,rettyp=get_value("CRetType"),fname="qmckl_ao_polynomial_transp_vgl_hpc")
@ -4426,7 +4426,7 @@ end function qmckl_ao_polynomial_vgl_doc_f
int32_t* const L,
const int64_t ldl,
double* const VGL,
const int64_t ldv );
const int64_t ldv );
#+end_src
#+begin_src c :tangle (eval c) :comments org
@ -4445,7 +4445,7 @@ qmckl_ao_polynomial_transp_vgl (const qmckl_context context,
return qmckl_ao_polynomial_transp_vgl_hpc (context, X, R, lmax, n, L, ldl, VGL, ldv);
#else
return qmckl_ao_polynomial_transp_vgl_doc (context, X, R, lmax, n, L, ldl, VGL, ldv);
#endif
#endif
}
#+end_src
@ -4523,7 +4523,7 @@ integer function qmckl_ao_polynomial_transp_vgl_doc_f (context, &
VGL(4,4) = 1.d0
n=4
else
else
VGL(1,1) = 1.d0
VGL(1,2:5) = 0.d0
l(1:3,1) = 0
@ -4605,7 +4605,7 @@ qmckl_ao_polynomial_transp_vgl_hpc (const qmckl_context context,
VGL[0 ] = 1.0;
VGL[ldv ] = 0.0;
VGL[ldv<<1 ] = 0.0;
VGL[(ldv<<2)-1] = 0.0;
VGL[(ldv<<1)+ldv] = 0.0;
VGL[ldv<<2 ] = 0.0;
m=1;
break;
@ -4619,10 +4619,10 @@ qmckl_ao_polynomial_transp_vgl_hpc (const qmckl_context context,
const int32_t ltmp[12] = {0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1};
for (int i=0 ; i<12 ; ++i)
L[i] = ltmp[i];
} else {
int32_t* const l[4] = {L, L+ldl, L+(ldl<<1), L+3*ldl};
int32_t* const l[4] = {L, L+ldl, L+(ldl<<1), L+ldl+(ldl<<1)};
l[0][0] = 0; l[0][1] = 0; l[0][2] = 0;
l[1][0] = 1; l[1][1] = 0; l[1][2] = 0;
l[2][0] = 0; l[2][1] = 1; l[2][2] = 0;
@ -4646,7 +4646,7 @@ qmckl_ao_polynomial_transp_vgl_hpc (const qmckl_context context,
double* const vgl1 = VGL;
double* const vgl2 = VGL + ldv;
double* const vgl3 = VGL + (ldv << 1);
double* const vgl4 = VGL + 3*ldv;
double* const vgl4 = VGL + ldv + (ldv << 1);
double* const vgl5 = VGL + (ldv << 2);
for (int32_t k=0 ; k<4 ; ++k) {
@ -4670,7 +4670,7 @@ qmckl_ao_polynomial_transp_vgl_hpc (const qmckl_context context,
{
if (ldv < 10) return QMCKL_INVALID_ARG_9;
if (ldl == 3) {
const int32_t ltmp[30] = {0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1,
const int32_t ltmp[30] = {0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1,
2, 0, 0, 1, 1, 0, 1, 0, 1, 0, 2, 0,
0, 1, 1, 0, 0, 2};
for (int i=0 ; i<30 ; ++i)
@ -4681,11 +4681,11 @@ qmckl_ao_polynomial_transp_vgl_hpc (const qmckl_context context,
for (int32_t i=0 ; i<10 ; ++i) {
l[i] = L + i*ldl;
}
l[0][0] = 0; l[0][1] = 0; l[0][2] = 0;
l[1][0] = 1; l[1][1] = 0; l[1][2] = 0;
l[2][0] = 0; l[2][1] = 1; l[2][2] = 0;
l[3][0] = 0; l[3][1] = 0; l[3][2] = 1;
l[3][0] = 0; l[3][1] = 0; l[3][2] = 1;
l[4][0] = 2; l[4][1] = 0; l[4][2] = 0;
l[5][0] = 1; l[5][1] = 1; l[5][2] = 0;
l[6][0] = 1; l[6][1] = 0; l[6][2] = 1;
@ -4694,8 +4694,8 @@ qmckl_ao_polynomial_transp_vgl_hpc (const qmckl_context context,
l[9][0] = 0; l[9][1] = 0; l[9][2] = 2;
}
const double Y[3] = { X[0]-R[0],
X[1]-R[1],
const double Y[3] = { X[0]-R[0],
X[1]-R[1],
X[2]-R[2] };
if (ldv == 50) {
@ -4743,50 +4743,45 @@ qmckl_ao_polynomial_transp_vgl_hpc (const qmckl_context context,
vgl5[3] = 0.0; vgl5[4] = 2.0; vgl5[5] = 0.0;
vgl5[6] = 0.0; vgl5[7] = 2.0; vgl5[8] = 0.0;
vgl5[9] = 2.0;
}
}
m=10;
break;
}
/*
case 3:
{
if (ldv < 20) return QMCKL_INVALID_ARG_9;
}
,*/
default:
default:
{
const int32_t size_max = (lmax+1)*(lmax+2)*(lmax+3)/6;
if (ldv < size_max) return QMCKL_INVALID_ARG_9;
double* const vgl1 = VGL;
double* const vgl2 = VGL + ldv;
double* const vgl3 = VGL + (ldv << 1);
double* const vgl4 = VGL + 3*ldv;
double* const vgl5 = VGL + (ldv << 2);
const double Y[3] = { X[0]-R[0],
X[1]-R[1],
double* const vgl3 = VGL + (ldv<<1);
double* const vgl4 = VGL + ldv + (ldv<<1);
double* const vgl5 = VGL + (ldv<<2);
const double Y[3] = { X[0]-R[0],
X[1]-R[1],
X[2]-R[2] };
assert(size_max > lmax+3);
double pows[3][size_max];
for (int32_t i=0 ; i<=2 ; ++i) {
for (int32_t i=0 ; i<3 ; ++i) {
pows[0][i] = 1.0;
pows[1][i] = 1.0;
pows[2][i] = 1.0;
}
for (int32_t i=3 ; i<=lmax+2 ; ++i) {
pows[0][i] = pows[0][i-1] * Y[0];
pows[1][i] = pows[1][i-1] * Y[1];
pows[2][i] = pows[2][i-1] * Y[2];
}
int32_t* l[24];
int32_t* l[size_max];
for (int32_t i=0 ; i<size_max ; ++i) {
l[i] = L + i*ldl;
l[i] = &(L[i*ldl]);
}
for (int32_t i=0 ; i<4 ; ++i) {
l[i][0] = 0;
l[i][1] = 0;
@ -4795,7 +4790,7 @@ qmckl_ao_polynomial_transp_vgl_hpc (const qmckl_context context,
l[1][0] = 1;
l[2][1] = 1;
l[3][2] = 1;
for (int32_t k=0 ; k<4 ; ++k) {
vgl2[k] = 0.0;
vgl3[k] = 0.0;
@ -4811,41 +4806,41 @@ qmckl_ao_polynomial_transp_vgl_hpc (const qmckl_context context,
vgl3[2] = 1.0;
vgl4[3] = 1.0;
m=4;
double dd = 2.0;
for (int32_t d=2 ; d<= lmax ; ++d) {
double da = dd;
for (int32_t a=d ; a>=0 ; --a) {
double db = dd-da;
for (int32_t b=d-a ; b>=0 ; --b) {
const int32_t c = d - a - b;
const int32_t c = d - a - b;
const double dc = dd - da - db;
double xy = pows[0][a+2] * pows[1][b+2];
double yz = pows[1][b+2] * pows[2][c+2];
double xz = pows[0][a+2] * pows[2][c+2];
l[m][0] = a;
l[m][1] = b;
l[m][2] = c;
vgl1[m] = xy * pows[2][c+2];
xy *= dc;
xz *= db;
yz *= da;
vgl2[m] = pows[0][a+1] * yz;
vgl3[m] = pows[1][b+1] * xz;
vgl4[m] = pows[2][c+1] * xy;
vgl5[m] = (da-1.) * pows[0][a] * yz +
(db-1.) * pows[1][b] * xz +
(dc-1.) * pows[2][c] * xy;
db -= 1.0;
++m;
}
@ -5040,13 +5035,13 @@ assert(0 == test_qmckl_ao_polynomial_vgl(context));
double X[3] = { 1.1, 2.2, 3.3 };
double R[3] = { 0.2, 1.1, 3.0 };
int32_t ldv[4] = {1, 4, 10, 20};
for (int32_t ldl=3 ; ldl<5 ; ++ldl) {
int32_t ldv[8] = {1, 4, 10, 20, 35, 56, 84, 120};
for (int32_t ldl=3 ; ldl<=5 ; ++ldl) {
int64_t n;
int32_t L0[24][ldl];
int32_t L1[24][ldl];
int32_t L0[200][ldl];
int32_t L1[200][ldl];
printf("ldl=%d\n", ldl);
for (int32_t lmax=0 ; lmax<=3 ; lmax++) {
for (int32_t lmax=0 ; lmax<=7 ; lmax++) {
double VGL0[5][ldv[lmax]];
double VGL1[5][ldv[lmax]];
memset(&L0[0][0], 0, sizeof(L0));
@ -5064,7 +5059,7 @@ for (int32_t ldl=3 ; ldl<5 ; ++ldl) {
assert( L0[l][k] == L1[l][k] );
}
}
for (int32_t k=0 ; k<5 ; ++k) {
for (int32_t l=0 ; l<n ; ++l) {
printf("VGL[%d][%d] = %e %e\n", k, l, VGL0[k][l], VGL1[k][l]);
@ -5167,9 +5162,9 @@ integer function qmckl_compute_ao_vgl_doc_f(context, &
e_coord(2) = coord(ipoint,2)
e_coord(3) = coord(ipoint,3)
do inucl=1,nucl_num
n_coord(1) = nucl_coord(inucl,1)
n_coord(2) = nucl_coord(inucl,2)
n_coord(3) = nucl_coord(inucl,3)
n_coord(1) = nucl_coord(inucl,1)
n_coord(2) = nucl_coord(inucl,2)
n_coord(3) = nucl_coord(inucl,3)
! Test if the point is in the range of the nucleus
x = e_coord(1) - n_coord(1)
@ -5182,7 +5177,7 @@ integer function qmckl_compute_ao_vgl_doc_f(context, &
cycle
end if
! Compute polynomials
! Compute polynomials
info = qmckl_ao_polynomial_vgl_doc_f(context, e_coord, n_coord, &
nucleus_max_ang_mom(inucl), n_poly, powers, 3_8, &
poly_vgl, 5_8)
@ -5237,7 +5232,7 @@ end function qmckl_compute_ao_vgl_doc_f
#+end_src
** HPC version
#+NAME: qmckl_ao_vgl_args_hpc
#+NAME: qmckl_ao_vgl_args_hpc_gaussian
| Variable | Type | In/Out | Description |
|-----------------------+--------------------------------+--------+----------------------------------------------|
| ~context~ | ~qmckl_context~ | in | Global state |
@ -5262,7 +5257,7 @@ end function qmckl_compute_ao_vgl_doc_f
#+begin_src c :comments org :tangle (eval c) :noweb yes :exports none
qmckl_exit_code
qmckl_compute_ao_vgl_hpc (
qmckl_compute_ao_vgl_hpc_gaussian (
const qmckl_context context,
const int64_t ao_num,
const int64_t shell_num,
@ -5283,14 +5278,14 @@ qmckl_compute_ao_vgl_hpc (
const double* coef_normalized,
double* const ao_vgl )
{
int32_t lstart[24];
for (int32_t l=0 ; l<24 ; ++l) {
int32_t lstart[32];
for (int32_t l=0 ; l<32 ; ++l) {
lstart[l] = l*(l+1)*(l+2)/6;
}
int64_t ao_index[shell_num+1];
int64_t size_max = 0;
{
{
int64_t k=0;
for (int inucl=0 ; inucl < nucl_num ; ++inucl) {
const int64_t ishell_start = nucleus_index[inucl];
@ -5309,7 +5304,7 @@ qmckl_compute_ao_vgl_hpc (
double cutoff = -log(1.e-12);
#ifdef HAVE_OPENMP
#pragma omp parallel
#pragma omp parallel
#endif
{
qmckl_exit_code rc;
@ -5321,10 +5316,10 @@ qmckl_compute_ao_vgl_hpc (
{0.0, 1.0, 0.0, 0.0},
{0.0, 0.0, 1.0, 0.0},
{0.0, 0.0, 0.0, 1.0}};
double poly_vgl_l2[5][10] = {{1., 0., 0., 0., 0., 0., 0., 0., 0., 0.},
{0., 1., 0., 0., 0., 0., 0., 0., 0., 0.},
double poly_vgl_l2[5][10] = {{1., 0., 0., 0., 0., 0., 0., 0., 0., 0.},
{0., 1., 0., 0., 0., 0., 0., 0., 0., 0.},
{0., 0., 1., 0., 0., 0., 0., 0., 0., 0.},
{0., 0., 0., 1., 0., 0., 0., 0., 0., 0.},
{0., 0., 0., 1., 0., 0., 0., 0., 0., 0.},
{0., 0., 0., 0., 2., 0., 0., 2., 0., 2.}};
double poly_vgl[5][size_max];
@ -5332,12 +5327,12 @@ qmckl_compute_ao_vgl_hpc (
#pragma omp for
#endif
for (int64_t ipoint=0 ; ipoint < point_num ; ++ipoint) {
const double e_coord[3] = { coord[ipoint],
const double e_coord[3] = { coord[ipoint],
coord[ipoint + point_num],
coord[ipoint + 2*point_num] };
for (int64_t inucl=0 ; inucl < nucl_num ; ++inucl) {
const double n_coord[3] = { nucl_coord[inucl],
const double n_coord[3] = { nucl_coord[inucl],
nucl_coord[inucl + nucl_num],
nucl_coord[inucl + 2*nucl_num] };
@ -5357,13 +5352,13 @@ qmckl_compute_ao_vgl_hpc (
case 0:
break;
case 1:
case 1:
poly_vgl_l1[0][1] = x;
poly_vgl_l1[0][2] = y;
poly_vgl_l1[0][3] = z;
break;
case 2:
case 2:
poly_vgl_l2[0][1] = x;
poly_vgl_l2[0][2] = y;
poly_vgl_l2[0][3] = z;
@ -5385,20 +5380,19 @@ qmckl_compute_ao_vgl_hpc (
break;
default:
rc = qmckl_ao_polynomial_transp_vgl_hpc(context, e_coord, n_coord,
rc = qmckl_ao_polynomial_transp_vgl_hpc(context, e_coord, n_coord,
nucleus_max_ang_mom[inucl],
&n_poly, powers, (int64_t) 3,
&n_poly, powers, (int64_t) 3,
&(poly_vgl[0][0]), size_max);
assert (rc == QMCKL_SUCCESS);
break;
}
const int64_t ishell_start = nucleus_index[inucl];
const int64_t ishell_end = nucleus_index[inucl] + nucleus_shell_num[inucl];
for (int64_t ishell = ishell_start ; ishell < ishell_end ; ++ishell) {
/* /!\ Gaussian fuctions */
int64_t nidx = 0;
const int64_t iprim_start = shell_prim_index[ishell];
const int64_t iprim_end = shell_prim_index[ishell] + shell_prim_num[ishell];
@ -5429,9 +5423,10 @@ qmckl_compute_ao_vgl_hpc (
const double s4 = s6_*z;
const double s5 = s5_;
const int32_t l = shell_ang_mom[ishell];
const int32_t n = lstart[l+1]-lstart[l];
const int64_t k = ao_index[ishell];
const int64_t k = ao_index[ishell];
double* const ao_vgl_1 = &(ao_vgl[ipoint*5*ao_num+k]);
double* const ao_vgl_2 = &(ao_vgl_1[ ao_num]);
@ -5467,7 +5462,7 @@ qmckl_compute_ao_vgl_hpc (
ao_vgl_3[il] = (poly_vgl_3[il] * s1 + poly_vgl_1[il] * s3) * f[il];
ao_vgl_4[il] = (poly_vgl_4[il] * s1 + poly_vgl_1[il] * s4) * f[il];
ao_vgl_5[il] = (poly_vgl_1[il] * s5 +
2.0*(poly_vgl_2[il] * s2 +
2.0*(poly_vgl_2[il] * s2 +
poly_vgl_3[il] * s3 +
poly_vgl_4[il] * s4 )) * f[il];
}
@ -5484,7 +5479,7 @@ qmckl_compute_ao_vgl_hpc (
ao_vgl_3[il] = (poly_vgl_3[il] * s1 + poly_vgl_1[il] * s3) * f[il];
ao_vgl_4[il] = (poly_vgl_4[il] * s1 + poly_vgl_1[il] * s4) * f[il];
ao_vgl_5[il] = (poly_vgl_5[il] * s1 + poly_vgl_1[il] * s5 +
2.0*(poly_vgl_2[il] * s2 +
2.0*(poly_vgl_2[il] * s2 +
poly_vgl_3[il] * s3 +
poly_vgl_4[il] * s4 )) * f[il];
}
@ -5501,7 +5496,7 @@ qmckl_compute_ao_vgl_hpc (
ao_vgl_3[il] = (poly_vgl_3[il] * s1 + poly_vgl_1[il] * s3) * f[il];
ao_vgl_4[il] = (poly_vgl_4[il] * s1 + poly_vgl_1[il] * s4) * f[il];
ao_vgl_5[il] = (poly_vgl_5[il] * s1 + poly_vgl_1[il] * s5 +
2.0*(poly_vgl_2[il] * s2 +
2.0*(poly_vgl_2[il] * s2 +
poly_vgl_3[il] * s3 +
poly_vgl_4[il] * s4 )) * f[il];
}
@ -5550,7 +5545,7 @@ qmckl_compute_ao_vgl_hpc (
double* const ao_vgl );
#+end_src
#+begin_src c :tangle (eval h_private_func) :comments org
qmckl_exit_code qmckl_compute_ao_vgl_hpc (
qmckl_exit_code qmckl_compute_ao_vgl_hpc_gaussian (
const qmckl_context context,
const int64_t ao_num,
const int64_t shell_num,
@ -5633,7 +5628,7 @@ qmckl_compute_ao_vgl_hpc (
end function qmckl_compute_ao_vgl_doc
#+end_src
*** Provide :noexport:
@ -5691,7 +5686,8 @@ qmckl_exit_code qmckl_provide_ao_vgl(qmckl_context context)
ctx->ao_basis.ao_vgl = ao_vgl;
}
#ifdef HAVE_HPC
rc = qmckl_compute_ao_vgl_hpc(context,
if (ctx->ao_basis.type == 'G') {
rc = qmckl_compute_ao_vgl_hpc_gaussian(context,
ctx->ao_basis.ao_num,
ctx->ao_basis.shell_num,
ctx->ao_basis.prim_num,
@ -5710,6 +5706,45 @@ qmckl_exit_code qmckl_provide_ao_vgl(qmckl_context context)
ctx->ao_basis.exponent,
ctx->ao_basis.coefficient_normalized,
ctx->ao_basis.ao_vgl);
/*
} else if (ctx->ao_basis.type == 'S') {
rc = qmck_compute_ao_vgl_hpc_slater(context,
ctx->ao_basis.ao_num,
ctx->ao_basis.shell_num,
ctx->ao_basis.prim_num,
ctx->point.num,
ctx->nucleus.num,
ctx->point.coord.data,
ctx->nucleus.coord.data,
ctx->ao_basis.nucleus_index,
ctx->ao_basis.nucleus_shell_num,
ctx->ao_basis.nucleus_range,
ctx->ao_basis.nucleus_max_ang_mom,
ctx->ao_basis.shell_ang_mom,
ctx->ao_basis.shell_prim_index,
ctx->ao_basis.shell_prim_num,
ctx->ao_basis.ao_factor,
ctx->ao_basis.exponent,
ctx->ao_basis.coefficient_normalized,
ctx->ao_basis.ao_vgl);
,*/
} else {
rc = qmckl_compute_ao_vgl_doc(context,
ctx->ao_basis.ao_num,
ctx->ao_basis.shell_num,
ctx->point.num,
ctx->nucleus.num,
ctx->point.coord.data,
ctx->nucleus.coord.data,
ctx->ao_basis.nucleus_index,
ctx->ao_basis.nucleus_shell_num,
ctx->ao_basis.nucleus_range,
ctx->ao_basis.nucleus_max_ang_mom,
ctx->ao_basis.shell_ang_mom,
ctx->ao_basis.ao_factor,
ctx->ao_basis.shell_vgl,
ctx->ao_basis.ao_vgl);
}
#else
rc = qmckl_compute_ao_vgl_doc(context,
ctx->ao_basis.ao_num,