Skip to content

Commit

Permalink
Fix cart2sph transformation bugs for int1e_grids
Browse files Browse the repository at this point in the history
  • Loading branch information
@sunqm
sunqm committed Jun 15, 2022
1 parent 8153147 commit 28100e8
Showing 1 changed file with 36 additions and 27 deletions.
63 changes: 36 additions & 27 deletions src/cart2sph.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -4675,7 +4675,7 @@ static void zcopy_ij(double complex *out, double *gctrR, double *gctrI,
} }
}

static void dcopy_grids_ij(double *out, const double *gctr,
static void dcopy_grids_ij(double *out, const double *gctr, const int counts,
const int ngrids, const int ni, const int nj,
const int mgrids, const int mi, const int mj)
{
Expand All @@ -4686,7 +4686,7 @@ static void dcopy_grids_ij(double *out, const double *gctr,
for (j = 0; j < mj; j++) {
for (i = 0; i < mi; i++) {
#pragma GCC ivdep
for (m = 0; m < mgrids; m++) {
for (m = 0; m < counts; m++) {
out[i*ngrids+m] = gctr[i*mgrids+m];
} }
out += ngi;
Expand All @@ -4695,6 +4695,7 @@ static void dcopy_grids_ij(double *out, const double *gctr,
}

static void zcopy_grids_ij(double complex *out, double *gctrR, double *gctrI,
const int counts,
const int ngrids, const int ni, const int nj,
const int mgrids, const int mi, const int mj)
{
Expand All @@ -4706,7 +4707,7 @@ static void zcopy_grids_ij(double complex *out, double *gctrR, double *gctrI,
for (j = 0; j < mj; j++) {
for (i = 0; i < mi; i++) {
#pragma GCC ivdep
for (m = 0; m < mgrids; m++) {
for (m = 0; m < counts; m++) {
dout[(i*ngrids+m)*OF_CMPLX ] = gctrR[i*mgrids+m];
dout[(i*ngrids+m)*OF_CMPLX+1] = gctrI[i*mgrids+m];
} }
Expand Down Expand Up @@ -5219,7 +5220,7 @@ void c2s_sph_1e_grids(double *out, double *gctr, int *dims,
int nfi = envs->nfi;
int nf = envs->nf;
int ic, jc, grids_offset;
int bgrids, bgrids_di, bgrids_nfi;
int bgrids, bgrids_raw, bgrids_di, bgrids_nfi;
int buflen = GRID_BLKSIZE * nfi * dj;
double *buf1, *buf2;
MALLOC_INSTACK(buf1, buflen);
Expand All @@ -5228,15 +5229,16 @@ void c2s_sph_1e_grids(double *out, double *gctr, int *dims,
double *tmp1;

for (grids_offset = 0; grids_offset < ngrids; grids_offset += GRID_BLKSIZE) {
bgrids = MIN(ngrids - grids_offset, GRID_BLKSIZE);
bgrids_raw = MIN(ngrids - grids_offset, GRID_BLKSIZE);
bgrids = ALIGN_UP(bgrids_raw, SIMDD);
bgrids_di = bgrids * di;
bgrids_nfi = bgrids * nfi;
for (jc = 0; jc < j_ctr; jc++) {
for (ic = 0; ic < i_ctr; ic++) {
tmp1 = (c2s_ket_sph[j_l])(buf1, gctr, bgrids_nfi, bgrids_nfi, j_l);
tmp1 = sph2e_inner(buf2, tmp1, i_l, bgrids, dj, bgrids_di, bgrids_nfi);
pij = out + Ng * (ofj * jc + di * ic) + grids_offset;
dcopy_grids_ij(pij, tmp1, Ng, ni, nj, bgrids, di, dj);
dcopy_grids_ij(pij, tmp1, bgrids_raw, Ng, ni, nj, bgrids, di, dj);
gctr += bgrids * nf;
} }
}
Expand All @@ -5256,15 +5258,16 @@ void c2s_cart_1e_grids(double *out, double *gctr, int *dims,
int nf = envs->nf;
int ofj = ni * nfj;
int ic, jc, grids_offset;
int bgrids;
int bgrids, bgrids_raw;
double *pij;

for (grids_offset = 0; grids_offset < ngrids; grids_offset += GRID_BLKSIZE) {
bgrids = MIN(ngrids - grids_offset, GRID_BLKSIZE);
bgrids_raw = MIN(ngrids - grids_offset, GRID_BLKSIZE);
bgrids = ALIGN_UP(bgrids_raw, SIMDD);
for (jc = 0; jc < j_ctr; jc++) {
for (ic = 0; ic < i_ctr; ic++) {
pij = out + Ng * (ofj * jc + nfi * ic) + grids_offset;
dcopy_grids_ij(pij, gctr, Ng, ni, nj, bgrids, nfi, nfj);
dcopy_grids_ij(pij, gctr, bgrids_raw, Ng, ni, nj, bgrids, nfi, nfj);
gctr += bgrids * nf;
} }
}
Expand Down Expand Up @@ -5298,7 +5301,7 @@ void c2s_sf_1e_grids(double complex *out, double *gctr, int *dims,
int nf2j = nfj + nfj;
int nf = envs->nf;
int ic, jc, grids_offset;
int bgrids, bgrids_di;
int bgrids, bgrids_raw, bgrids_di;
int buflen = GRID_BLKSIZE * di * nf2j;
double *tmp1R, *tmp1I, *tmp2R, *tmp2I;
MALLOC_INSTACK(tmp1R, buflen);
Expand All @@ -5308,14 +5311,15 @@ void c2s_sf_1e_grids(double complex *out, double *gctr, int *dims,
double complex *pij;

for (grids_offset = 0; grids_offset < ngrids; grids_offset += GRID_BLKSIZE) {
bgrids = MIN(ngrids - grids_offset, GRID_BLKSIZE);
bgrids_raw = MIN(ngrids - grids_offset, GRID_BLKSIZE);
bgrids = ALIGN_UP(bgrids_raw, SIMDD);
bgrids_di = bgrids * di;
for (jc = 0; jc < j_ctr; jc++) {
for (ic = 0; ic < i_ctr; ic++) {
a_bra1_cart2spinor_sf(tmp1R, tmp1I, NULL, NULL, NULL, gctr, bgrids, nfj, i_kp, i_l);
a_ket_cart2spinor(tmp2R, tmp2I, tmp1R, tmp1I, bgrids_di, j_kp, j_l);
pij = out + Ng * (ofj * jc + di * ic) + grids_offset;
zcopy_grids_ij(pij, tmp2R, tmp2I, Ng, ni, nj, bgrids, di, dj);
zcopy_grids_ij(pij, tmp2R, tmp2I, bgrids_raw, Ng, ni, nj, bgrids, di, dj);
gctr += bgrids * nf;
} }
}
Expand Down Expand Up @@ -5345,7 +5349,7 @@ void c2s_sf_1e_gridsi(double complex *out, double *gctr, int *dims,
int nf2j = nfj + nfj;
int nf = envs->nf;
int ic, jc, grids_offset;
int bgrids, bgrids_di;
int bgrids, bgrids_raw, bgrids_di;
int buflen = GRID_BLKSIZE * di * nf2j;
double *tmp1R, *tmp1I, *tmp2R, *tmp2I;
MALLOC_INSTACK(tmp1R, buflen);
Expand All @@ -5355,14 +5359,15 @@ void c2s_sf_1e_gridsi(double complex *out, double *gctr, int *dims,
double complex *pij;

for (grids_offset = 0; grids_offset < ngrids; grids_offset += GRID_BLKSIZE) {
bgrids = MIN(ngrids - grids_offset, GRID_BLKSIZE);
bgrids_raw = MIN(ngrids - grids_offset, GRID_BLKSIZE);
bgrids = ALIGN_UP(bgrids_raw, SIMDD);
bgrids_di = bgrids * di;
for (jc = 0; jc < j_ctr; jc++) {
for (ic = 0; ic < i_ctr; ic++) {
a_bra1_cart2spinor_sf(tmp1R, tmp1I, NULL, NULL, NULL, gctr, bgrids, nfj, i_kp, i_l);
a_iket_cart2spinor(tmp2R, tmp2I, tmp1R, tmp1I, bgrids_di, j_kp, j_l);
pij = out + Ng * (ofj * jc + di * ic) + grids_offset;
zcopy_grids_ij(pij, tmp2R, tmp2I, Ng, ni, nj, bgrids, di, dj);
zcopy_grids_ij(pij, tmp2R, tmp2I, bgrids_raw, Ng, ni, nj, bgrids, di, dj);
gctr += bgrids * nf;
} }
}
Expand Down Expand Up @@ -5392,12 +5397,13 @@ void c2s_si_1e_grids(double complex *out, double *gctr, int *dims,
int nf2i = nfi + nfi;
int nf2j = nfj + nfj;
int nf = envs->nf;
int ngrids_nf = ALIGN_UP(ngrids, SIMDD) * nf;
int ic, jc, grids_offset;
int bgrids, bgrids_di, bgrids_nf;
int bgrids, bgrids_raw, bgrids_di, bgrids_nf;
double *gc_x = gctr;
double *gc_y = gc_x + ngrids * nf * i_ctr * j_ctr;
double *gc_z = gc_y + ngrids * nf * i_ctr * j_ctr;
double *gc_1 = gc_z + ngrids * nf * i_ctr * j_ctr;
double *gc_y = gc_x + ngrids_nf * i_ctr * j_ctr;
double *gc_z = gc_y + ngrids_nf * i_ctr * j_ctr;
double *gc_1 = gc_z + ngrids_nf * i_ctr * j_ctr;
int buflen = GRID_BLKSIZE * di * nf2j;
double *tmp1R, *tmp1I, *tmp2R, *tmp2I;
MALLOC_INSTACK(tmp1R, buflen);
Expand All @@ -5407,15 +5413,16 @@ void c2s_si_1e_grids(double complex *out, double *gctr, int *dims,
double complex *pij;

for (grids_offset = 0; grids_offset < ngrids; grids_offset += GRID_BLKSIZE) {
bgrids = MIN(ngrids - grids_offset, GRID_BLKSIZE);
bgrids_raw = MIN(ngrids - grids_offset, GRID_BLKSIZE);
bgrids = ALIGN_UP(bgrids_raw, SIMDD);
bgrids_di = bgrids * di;
bgrids_nf = bgrids * nf;
for (jc = 0; jc < j_ctr; jc++) {
for (ic = 0; ic < i_ctr; ic++) {
a_bra1_cart2spinor_si(tmp1R, tmp1I, gc_x, gc_y, gc_z, gc_1, bgrids, nfj, i_kp, i_l);
a_ket_cart2spinor(tmp2R, tmp2I, tmp1R, tmp1I, bgrids_di, j_kp, j_l);
pij = out + Ng * (ofj * jc + di * ic) + grids_offset;
zcopy_grids_ij(pij, tmp2R, tmp2I, Ng, ni, nj, bgrids, di, dj);
zcopy_grids_ij(pij, tmp2R, tmp2I, bgrids_raw, Ng, ni, nj, bgrids, di, dj);
gc_x += bgrids_nf;
gc_y += bgrids_nf;
gc_z += bgrids_nf;
Expand Down Expand Up @@ -5448,12 +5455,13 @@ void c2s_si_1e_gridsi(double complex *out, double *gctr, int *dims,
int nf2i = nfi + nfi;
int nf2j = nfj + nfj;
int nf = envs->nf;
int ngrids_nf = ALIGN_UP(ngrids, SIMDD) * nf;
int ic, jc, grids_offset;
int bgrids, bgrids_di, bgrids_nf;
int bgrids, bgrids_raw, bgrids_di, bgrids_nf;
double *gc_x = gctr;
double *gc_y = gc_x + ngrids * nf * i_ctr * j_ctr;
double *gc_z = gc_y + ngrids * nf * i_ctr * j_ctr;
double *gc_1 = gc_z + ngrids * nf * i_ctr * j_ctr;
double *gc_y = gc_x + ngrids_nf * i_ctr * j_ctr;
double *gc_z = gc_y + ngrids_nf * i_ctr * j_ctr;
double *gc_1 = gc_z + ngrids_nf * i_ctr * j_ctr;
int buflen = GRID_BLKSIZE * di * nf2j;
double *tmp1R, *tmp1I, *tmp2R, *tmp2I;
MALLOC_INSTACK(tmp1R, buflen);
Expand All @@ -5463,15 +5471,16 @@ void c2s_si_1e_gridsi(double complex *out, double *gctr, int *dims,
double complex *pij;

for (grids_offset = 0; grids_offset < ngrids; grids_offset += GRID_BLKSIZE) {
bgrids = MIN(ngrids - grids_offset, GRID_BLKSIZE);
bgrids_raw = MIN(ngrids - grids_offset, GRID_BLKSIZE);
bgrids = ALIGN_UP(bgrids_raw, SIMDD);
bgrids_di = bgrids * di;
bgrids_nf = bgrids * nf;
for (jc = 0; jc < j_ctr; jc++) {
for (ic = 0; ic < i_ctr; ic++) {
a_bra1_cart2spinor_si(tmp1R, tmp1I, gc_x, gc_y, gc_z, gc_1, bgrids, nfj, i_kp, i_l);
a_iket_cart2spinor(tmp2R, tmp2I, tmp1R, tmp1I, bgrids_di, j_kp, j_l);
pij = out + Ng * (ofj * jc + di * ic) + grids_offset;
zcopy_grids_ij(pij, tmp2R, tmp2I, Ng, ni, nj, bgrids, di, dj);
zcopy_grids_ij(pij, tmp2R, tmp2I, bgrids_raw, Ng, ni, nj, bgrids, di, dj);
gc_x += bgrids_nf;
gc_y += bgrids_nf;
gc_z += bgrids_nf;
Expand Down

0 comments on commit 28100e8

Please sign in to comment.