#include "simint/boys/boys.h" #include "simint/ostei/gen/ostei_generated.h" #include "simint/vectorization/vectorization.h" #include #include int ostei_h_s_d_p(struct simint_multi_shellpair const P, struct simint_multi_shellpair const Q, double screen_tol, double * const restrict work, double * const restrict INT__h_s_d_p) { SIMINT_ASSUME_ALIGN_DBL(work); SIMINT_ASSUME_ALIGN_DBL(INT__h_s_d_p); int ab, cd, abcd; int istart, jstart; int iprimcd, nprim_icd, icd; const int check_screen = (screen_tol > 0.0); int i, j; int n; int not_screened; int real_abcd; int ibra; // partition workspace double * const INT__h_s_d_s = work + (SIMINT_NSHELL_SIMD * 0); double * const INT__h_s_f_s = work + (SIMINT_NSHELL_SIMD * 126); SIMINT_DBLTYPE * const primwork = (SIMINT_DBLTYPE *)(work + SIMINT_NSHELL_SIMD*336); SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_s = primwork + 0; SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_s_s = primwork + 9; SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_s_s = primwork + 33; SIMINT_DBLTYPE * const restrict PRIM_INT__f_s_s_s = primwork + 75; SIMINT_DBLTYPE * const restrict PRIM_INT__f_s_p_s = primwork + 135; SIMINT_DBLTYPE * const restrict PRIM_INT__g_s_s_s = primwork + 225; SIMINT_DBLTYPE * const restrict PRIM_INT__g_s_p_s = primwork + 300; SIMINT_DBLTYPE * const restrict PRIM_INT__g_s_d_s = primwork + 435; SIMINT_DBLTYPE * const restrict PRIM_INT__h_s_s_s = primwork + 615; SIMINT_DBLTYPE * const restrict PRIM_INT__h_s_p_s = primwork + 699; SIMINT_DBLTYPE * const restrict PRIM_INT__h_s_d_s = primwork + 888; SIMINT_DBLTYPE * const restrict PRIM_INT__h_s_f_s = primwork + 1140; double * const hrrwork = (double *)(primwork + 1350); // Create constants const SIMINT_DBLTYPE const_1 = SIMINT_DBLSET1(1); const SIMINT_DBLTYPE const_2 = SIMINT_DBLSET1(2); const SIMINT_DBLTYPE const_3 = SIMINT_DBLSET1(3); const SIMINT_DBLTYPE const_4 = SIMINT_DBLSET1(4); const SIMINT_DBLTYPE const_5 = SIMINT_DBLSET1(5); const SIMINT_DBLTYPE one_half = SIMINT_DBLSET1(0.5); //////////////////////////////////////// // Loop over shells and primitives //////////////////////////////////////// real_abcd = 0; istart = 0; for(ab = 0; ab < P.nshell12_clip; ++ab) { const int iend = istart + P.nprim12[ab]; cd = 0; jstart = 0; for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD) { const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD; int jend = jstart; for(i = 0; i < nshellbatch; i++) jend += Q.nprim12[cd+i]; // Clear the beginning of the workspace (where we are accumulating integrals) memset(work, 0, SIMINT_NSHELL_SIMD * 336 * sizeof(double)); abcd = 0; for(i = istart; i < iend; ++i) { SIMINT_DBLTYPE bra_screen_max; // only used if check_screen if(check_screen) { // Skip this whole thing if always insignificant if((P.screen[i] * Q.screen_max) < screen_tol) continue; bra_screen_max = SIMINT_DBLSET1(P.screen[i]); } icd = 0; iprimcd = 0; nprim_icd = Q.nprim12[cd]; double * restrict PRIM_PTR_INT__h_s_d_s = INT__h_s_d_s + abcd * 126; double * restrict PRIM_PTR_INT__h_s_f_s = INT__h_s_f_s + abcd * 210; // Load these one per loop over i const SIMINT_DBLTYPE P_alpha = SIMINT_DBLSET1(P.alpha[i]); const SIMINT_DBLTYPE P_prefac = SIMINT_DBLSET1(P.prefac[i]); const SIMINT_DBLTYPE Pxyz[3] = { SIMINT_DBLSET1(P.x[i]), SIMINT_DBLSET1(P.y[i]), SIMINT_DBLSET1(P.z[i]) }; const SIMINT_DBLTYPE P_PA[3] = { SIMINT_DBLSET1(P.PA_x[i]), SIMINT_DBLSET1(P.PA_y[i]), SIMINT_DBLSET1(P.PA_z[i]) }; for(j = jstart; j < jend; j += SIMINT_SIMD_LEN) { // calculate the shell offsets // these are the offset from the shell pointed to by cd // for each element int shelloffsets[SIMINT_SIMD_LEN] = {0}; int lastoffset = 0; const int nlane = ( ((j + SIMINT_SIMD_LEN) < jend) ? SIMINT_SIMD_LEN : (jend - j)); if((iprimcd + SIMINT_SIMD_LEN) >= nprim_icd) { // Handle if the first element of the vector is a new shell if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch)) { nprim_icd += Q.nprim12[cd + (++icd)]; PRIM_PTR_INT__h_s_d_s += 126; PRIM_PTR_INT__h_s_f_s += 210; } iprimcd++; for(n = 1; n < SIMINT_SIMD_LEN; ++n) { if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch)) { shelloffsets[n] = shelloffsets[n-1] + 1; lastoffset++; nprim_icd += Q.nprim12[cd + (++icd)]; } else shelloffsets[n] = shelloffsets[n-1]; iprimcd++; } } else iprimcd += SIMINT_SIMD_LEN; // Do we have to compute this vector (or has it been screened out)? // (not_screened != 0 means we have to do this vector) if(check_screen) { const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j))); if(vmax < screen_tol) { PRIM_PTR_INT__h_s_d_s += lastoffset*126; PRIM_PTR_INT__h_s_f_s += lastoffset*210; continue; } } const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j); const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha); const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha); const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum); /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */ SIMINT_DBLTYPE PQ[3]; PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j)); PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j)); PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j)); SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]); R2 = SIMINT_FMADD(PQ[1], PQ[1], R2); R2 = SIMINT_FMADD(PQ[2], PQ[2], R2); const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha); const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha); const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p); const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q); const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum); const SIMINT_DBLTYPE Q_PA[3] = { SIMINT_DBLLOAD(Q.PA_x, j), SIMINT_DBLLOAD(Q.PA_y, j), SIMINT_DBLLOAD(Q.PA_z, j) }; // NOTE: Minus sign! const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p); SIMINT_DBLTYPE aop_PQ[3]; aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]); aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]); aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]); SIMINT_DBLTYPE a_over_q = SIMINT_MUL(alpha, one_over_q); SIMINT_DBLTYPE aoq_PQ[3]; aoq_PQ[0] = SIMINT_MUL(a_over_q, PQ[0]); aoq_PQ[1] = SIMINT_MUL(a_over_q, PQ[1]); aoq_PQ[2] = SIMINT_MUL(a_over_q, PQ[2]); // Put a minus sign here so we don't have to in RR routines a_over_q = SIMINT_NEG(a_over_q); ////////////////////////////////////////////// // Fjt function section // Maximum v value: 8 ////////////////////////////////////////////// // The parameter to the Fjt function const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha); const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j); boys_F_split(PRIM_INT__s_s_s_s, F_x, 8); SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum); prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac); for(n = 0; n <= 8; n++) PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac); ////////////////////////////////////////////// // Primitive integrals: Vertical recurrance ////////////////////////////////////////////// const SIMINT_DBLTYPE vrr_const_1_over_2p = one_over_2p; const SIMINT_DBLTYPE vrr_const_2_over_2p = SIMINT_MUL(const_2, one_over_2p); const SIMINT_DBLTYPE vrr_const_3_over_2p = SIMINT_MUL(const_3, one_over_2p); const SIMINT_DBLTYPE vrr_const_4_over_2p = SIMINT_MUL(const_4, one_over_2p); const SIMINT_DBLTYPE vrr_const_1_over_2q = one_over_2q; const SIMINT_DBLTYPE vrr_const_2_over_2q = SIMINT_MUL(const_2, one_over_2q); const SIMINT_DBLTYPE vrr_const_1_over_2pq = one_over_2pq; const SIMINT_DBLTYPE vrr_const_2_over_2pq = SIMINT_MUL(const_2, one_over_2pq); const SIMINT_DBLTYPE vrr_const_3_over_2pq = SIMINT_MUL(const_3, one_over_2pq); const SIMINT_DBLTYPE vrr_const_4_over_2pq = SIMINT_MUL(const_4, one_over_2pq); const SIMINT_DBLTYPE vrr_const_5_over_2pq = SIMINT_MUL(const_5, one_over_2pq); // Forming PRIM_INT__p_s_s_s[8 * 3]; for(n = 0; n < 8; ++n) // loop over orders of auxiliary function { PRIM_INT__p_s_s_s[n * 3 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_s_s[n * 1 + 0]); PRIM_INT__p_s_s_s[n * 3 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__p_s_s_s[n * 3 + 0]); PRIM_INT__p_s_s_s[n * 3 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_s_s[n * 1 + 0]); PRIM_INT__p_s_s_s[n * 3 + 1] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__p_s_s_s[n * 3 + 1]); PRIM_INT__p_s_s_s[n * 3 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_s_s[n * 1 + 0]); PRIM_INT__p_s_s_s[n * 3 + 2] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__p_s_s_s[n * 3 + 2]); } // Forming PRIM_INT__d_s_s_s[7 * 6]; for(n = 0; n < 7; ++n) // loop over orders of auxiliary function { PRIM_INT__d_s_s_s[n * 6 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__p_s_s_s[n * 3 + 0]); PRIM_INT__d_s_s_s[n * 6 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__p_s_s_s[(n+1) * 3 + 0], PRIM_INT__d_s_s_s[n * 6 + 0]); PRIM_INT__d_s_s_s[n * 6 + 0] = SIMINT_FMADD( vrr_const_1_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__d_s_s_s[n * 6 + 0]); PRIM_INT__d_s_s_s[n * 6 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__p_s_s_s[n * 3 + 0]); PRIM_INT__d_s_s_s[n * 6 + 1] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__p_s_s_s[(n+1) * 3 + 0], PRIM_INT__d_s_s_s[n * 6 + 1]); PRIM_INT__d_s_s_s[n * 6 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__p_s_s_s[n * 3 + 0]); PRIM_INT__d_s_s_s[n * 6 + 2] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__p_s_s_s[(n+1) * 3 + 0], PRIM_INT__d_s_s_s[n * 6 + 2]); PRIM_INT__d_s_s_s[n * 6 + 3] = SIMINT_MUL(P_PA[1], PRIM_INT__p_s_s_s[n * 3 + 1]); PRIM_INT__d_s_s_s[n * 6 + 3] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__p_s_s_s[(n+1) * 3 + 1], PRIM_INT__d_s_s_s[n * 6 + 3]); PRIM_INT__d_s_s_s[n * 6 + 3] = SIMINT_FMADD( vrr_const_1_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__d_s_s_s[n * 6 + 3]); PRIM_INT__d_s_s_s[n * 6 + 4] = SIMINT_MUL(P_PA[2], PRIM_INT__p_s_s_s[n * 3 + 1]); PRIM_INT__d_s_s_s[n * 6 + 4] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__p_s_s_s[(n+1) * 3 + 1], PRIM_INT__d_s_s_s[n * 6 + 4]); PRIM_INT__d_s_s_s[n * 6 + 5] = SIMINT_MUL(P_PA[2], PRIM_INT__p_s_s_s[n * 3 + 2]); PRIM_INT__d_s_s_s[n * 6 + 5] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__p_s_s_s[(n+1) * 3 + 2], PRIM_INT__d_s_s_s[n * 6 + 5]); PRIM_INT__d_s_s_s[n * 6 + 5] = SIMINT_FMADD( vrr_const_1_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__d_s_s_s[n * 6 + 5]); } // Forming PRIM_INT__f_s_s_s[6 * 10]; for(n = 0; n < 6; ++n) // loop over orders of auxiliary function { PRIM_INT__f_s_s_s[n * 10 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__d_s_s_s[n * 6 + 0]); PRIM_INT__f_s_s_s[n * 10 + 0] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__d_s_s_s[(n+1) * 6 + 0], PRIM_INT__f_s_s_s[n * 10 + 0]); PRIM_INT__f_s_s_s[n * 10 + 0] = SIMINT_FMADD( vrr_const_2_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__p_s_s_s[(n+1) * 3 + 0], PRIM_INT__p_s_s_s[n * 3 + 0]), PRIM_INT__f_s_s_s[n * 10 + 0]); PRIM_INT__f_s_s_s[n * 10 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 0]); PRIM_INT__f_s_s_s[n * 10 + 1] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__d_s_s_s[(n+1) * 6 + 0], PRIM_INT__f_s_s_s[n * 10 + 1]); PRIM_INT__f_s_s_s[n * 10 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 0]); PRIM_INT__f_s_s_s[n * 10 + 2] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__d_s_s_s[(n+1) * 6 + 0], PRIM_INT__f_s_s_s[n * 10 + 2]); PRIM_INT__f_s_s_s[n * 10 + 3] = SIMINT_MUL(P_PA[0], PRIM_INT__d_s_s_s[n * 6 + 3]); PRIM_INT__f_s_s_s[n * 10 + 3] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__d_s_s_s[(n+1) * 6 + 3], PRIM_INT__f_s_s_s[n * 10 + 3]); PRIM_INT__f_s_s_s[n * 10 + 4] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 1]); PRIM_INT__f_s_s_s[n * 10 + 4] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__d_s_s_s[(n+1) * 6 + 1], PRIM_INT__f_s_s_s[n * 10 + 4]); PRIM_INT__f_s_s_s[n * 10 + 5] = SIMINT_MUL(P_PA[0], PRIM_INT__d_s_s_s[n * 6 + 5]); PRIM_INT__f_s_s_s[n * 10 + 5] = SIMINT_FMADD( aop_PQ[0], PRIM_INT__d_s_s_s[(n+1) * 6 + 5], PRIM_INT__f_s_s_s[n * 10 + 5]); PRIM_INT__f_s_s_s[n * 10 + 6] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 3]); PRIM_INT__f_s_s_s[n * 10 + 6] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__d_s_s_s[(n+1) * 6 + 3], PRIM_INT__f_s_s_s[n * 10 + 6]); PRIM_INT__f_s_s_s[n * 10 + 6] = SIMINT_FMADD( vrr_const_2_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__p_s_s_s[(n+1) * 3 + 1], PRIM_INT__p_s_s_s[n * 3 + 1]), PRIM_INT__f_s_s_s[n * 10 + 6]); PRIM_INT__f_s_s_s[n * 10 + 7] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 3]); PRIM_INT__f_s_s_s[n * 10 + 7] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__d_s_s_s[(n+1) * 6 + 3], PRIM_INT__f_s_s_s[n * 10 + 7]); PRIM_INT__f_s_s_s[n * 10 + 8] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 5]); PRIM_INT__f_s_s_s[n * 10 + 8] = SIMINT_FMADD( aop_PQ[1], PRIM_INT__d_s_s_s[(n+1) * 6 + 5], PRIM_INT__f_s_s_s[n * 10 + 8]); PRIM_INT__f_s_s_s[n * 10 + 9] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 5]); PRIM_INT__f_s_s_s[n * 10 + 9] = SIMINT_FMADD( aop_PQ[2], PRIM_INT__d_s_s_s[(n+1) * 6 + 5], PRIM_INT__f_s_s_s[n * 10 + 9]); PRIM_INT__f_s_s_s[n * 10 + 9] = SIMINT_FMADD( vrr_const_2_over_2p, SIMINT_FMADD(a_over_p, PRIM_INT__p_s_s_s[(n+1) * 3 + 2], PRIM_INT__p_s_s_s[n * 3 + 2]), PRIM_INT__f_s_s_s[n * 10 + 9]); } VRR_I_g_s_s_s( PRIM_INT__g_s_s_s, PRIM_INT__f_s_s_s, PRIM_INT__d_s_s_s, P_PA, a_over_p, aop_PQ, one_over_2p, 5); VRR_I_h_s_s_s( PRIM_INT__h_s_s_s, PRIM_INT__g_s_s_s, PRIM_INT__f_s_s_s, P_PA, a_over_p, aop_PQ, one_over_2p, 4); ostei_general_vrr_K(5, 0, 1, 0, 3, one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA, PRIM_INT__h_s_s_s, NULL, NULL, PRIM_INT__g_s_s_s, NULL, PRIM_INT__h_s_p_s); VRR_K_g_s_p_s( PRIM_INT__g_s_p_s, PRIM_INT__g_s_s_s, PRIM_INT__f_s_s_s, Q_PA, aoq_PQ, one_over_2pq, 3); ostei_general_vrr_K(5, 0, 2, 0, 2, one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA, PRIM_INT__h_s_p_s, PRIM_INT__h_s_s_s, NULL, PRIM_INT__g_s_p_s, NULL, PRIM_INT__h_s_d_s); VRR_K_f_s_p_s( PRIM_INT__f_s_p_s, PRIM_INT__f_s_s_s, PRIM_INT__d_s_s_s, Q_PA, aoq_PQ, one_over_2pq, 3); ostei_general_vrr_K(4, 0, 2, 0, 2, one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA, PRIM_INT__g_s_p_s, PRIM_INT__g_s_s_s, NULL, PRIM_INT__f_s_p_s, NULL, PRIM_INT__g_s_d_s); ostei_general_vrr_K(5, 0, 3, 0, 1, one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA, PRIM_INT__h_s_d_s, PRIM_INT__h_s_p_s, NULL, PRIM_INT__g_s_d_s, NULL, PRIM_INT__h_s_f_s); //////////////////////////////////// // Accumulate contracted integrals //////////////////////////////////// if(lastoffset == 0) { contract_all(126, PRIM_INT__h_s_d_s, PRIM_PTR_INT__h_s_d_s); contract_all(210, PRIM_INT__h_s_f_s, PRIM_PTR_INT__h_s_f_s); } else { contract(126, shelloffsets, PRIM_INT__h_s_d_s, PRIM_PTR_INT__h_s_d_s); contract(210, shelloffsets, PRIM_INT__h_s_f_s, PRIM_PTR_INT__h_s_f_s); PRIM_PTR_INT__h_s_d_s += lastoffset*126; PRIM_PTR_INT__h_s_f_s += lastoffset*210; } } // close loop over j } // close loop over i //Advance to the next batch jstart = SIMINT_SIMD_ROUND(jend); ////////////////////////////////////////////// // Contracted integrals: Horizontal recurrance ////////////////////////////////////////////// for(abcd = 0; abcd < nshellbatch; ++abcd, ++real_abcd) { const double hCD[3] = { Q.AB_x[cd+abcd], Q.AB_y[cd+abcd], Q.AB_z[cd+abcd] }; // set up HRR pointers double const * restrict HRR_INT__h_s_d_s = INT__h_s_d_s + abcd * 126; double const * restrict HRR_INT__h_s_f_s = INT__h_s_f_s + abcd * 210; double * restrict HRR_INT__h_s_d_p = INT__h_s_d_p + real_abcd * 378; // form INT__h_s_d_p for(ibra = 0; ibra < 21; ++ibra) { HRR_INT__h_s_d_p[ibra * 18 + 0] = HRR_INT__h_s_f_s[ibra * 10 + 0] + ( hCD[0] * HRR_INT__h_s_d_s[ibra * 6 + 0] ); HRR_INT__h_s_d_p[ibra * 18 + 1] = HRR_INT__h_s_f_s[ibra * 10 + 1] + ( hCD[1] * HRR_INT__h_s_d_s[ibra * 6 + 0] ); HRR_INT__h_s_d_p[ibra * 18 + 2] = HRR_INT__h_s_f_s[ibra * 10 + 2] + ( hCD[2] * HRR_INT__h_s_d_s[ibra * 6 + 0] ); HRR_INT__h_s_d_p[ibra * 18 + 3] = HRR_INT__h_s_f_s[ibra * 10 + 1] + ( hCD[0] * HRR_INT__h_s_d_s[ibra * 6 + 1] ); HRR_INT__h_s_d_p[ibra * 18 + 4] = HRR_INT__h_s_f_s[ibra * 10 + 3] + ( hCD[1] * HRR_INT__h_s_d_s[ibra * 6 + 1] ); HRR_INT__h_s_d_p[ibra * 18 + 5] = HRR_INT__h_s_f_s[ibra * 10 + 4] + ( hCD[2] * HRR_INT__h_s_d_s[ibra * 6 + 1] ); HRR_INT__h_s_d_p[ibra * 18 + 6] = HRR_INT__h_s_f_s[ibra * 10 + 2] + ( hCD[0] * HRR_INT__h_s_d_s[ibra * 6 + 2] ); HRR_INT__h_s_d_p[ibra * 18 + 7] = HRR_INT__h_s_f_s[ibra * 10 + 4] + ( hCD[1] * HRR_INT__h_s_d_s[ibra * 6 + 2] ); HRR_INT__h_s_d_p[ibra * 18 + 8] = HRR_INT__h_s_f_s[ibra * 10 + 5] + ( hCD[2] * HRR_INT__h_s_d_s[ibra * 6 + 2] ); HRR_INT__h_s_d_p[ibra * 18 + 9] = HRR_INT__h_s_f_s[ibra * 10 + 3] + ( hCD[0] * HRR_INT__h_s_d_s[ibra * 6 + 3] ); HRR_INT__h_s_d_p[ibra * 18 + 10] = HRR_INT__h_s_f_s[ibra * 10 + 6] + ( hCD[1] * HRR_INT__h_s_d_s[ibra * 6 + 3] ); HRR_INT__h_s_d_p[ibra * 18 + 11] = HRR_INT__h_s_f_s[ibra * 10 + 7] + ( hCD[2] * HRR_INT__h_s_d_s[ibra * 6 + 3] ); HRR_INT__h_s_d_p[ibra * 18 + 12] = HRR_INT__h_s_f_s[ibra * 10 + 4] + ( hCD[0] * HRR_INT__h_s_d_s[ibra * 6 + 4] ); HRR_INT__h_s_d_p[ibra * 18 + 13] = HRR_INT__h_s_f_s[ibra * 10 + 7] + ( hCD[1] * HRR_INT__h_s_d_s[ibra * 6 + 4] ); HRR_INT__h_s_d_p[ibra * 18 + 14] = HRR_INT__h_s_f_s[ibra * 10 + 8] + ( hCD[2] * HRR_INT__h_s_d_s[ibra * 6 + 4] ); HRR_INT__h_s_d_p[ibra * 18 + 15] = HRR_INT__h_s_f_s[ibra * 10 + 5] + ( hCD[0] * HRR_INT__h_s_d_s[ibra * 6 + 5] ); HRR_INT__h_s_d_p[ibra * 18 + 16] = HRR_INT__h_s_f_s[ibra * 10 + 8] + ( hCD[1] * HRR_INT__h_s_d_s[ibra * 6 + 5] ); HRR_INT__h_s_d_p[ibra * 18 + 17] = HRR_INT__h_s_f_s[ibra * 10 + 9] + ( hCD[2] * HRR_INT__h_s_d_s[ibra * 6 + 5] ); } } // close HRR loop } // close loop cdbatch istart = iend; } // close loop over ab return P.nshell12_clip * Q.nshell12_clip; } int ostei_s_h_d_p(struct simint_multi_shellpair const P, struct simint_multi_shellpair const Q, double screen_tol, double * const restrict work, double * const restrict INT__s_h_d_p) { double P_AB[3*P.nshell12]; struct simint_multi_shellpair P_tmp = P; P_tmp.PA_x = P.PB_x; P_tmp.PA_y = P.PB_y; P_tmp.PA_z = P.PB_z; P_tmp.PB_x = P.PA_x; P_tmp.PB_y = P.PA_y; P_tmp.PB_z = P.PA_z; P_tmp.AB_x = P_AB; P_tmp.AB_y = P_AB + P.nshell12; P_tmp.AB_z = P_AB + 2*P.nshell12; for(int i = 0; i < P.nshell12; i++) { P_tmp.AB_x[i] = -P.AB_x[i]; P_tmp.AB_y[i] = -P.AB_y[i]; P_tmp.AB_z[i] = -P.AB_z[i]; } int ret = ostei_h_s_d_p(P_tmp, Q, screen_tol, work, INT__s_h_d_p); double buffer[378] SIMINT_ALIGN_ARRAY_DBL; for(int q = 0; q < ret; q++) { int idx = 0; for(int a = 0; a < 1; ++a) for(int b = 0; b < 21; ++b) for(int c = 0; c < 6; ++c) for(int d = 0; d < 3; ++d) buffer[idx++] = INT__s_h_d_p[q*378+b*18+a*18+c*3+d]; memcpy(INT__s_h_d_p+q*378, buffer, 378*sizeof(double)); } return ret; } int ostei_h_s_p_d(struct simint_multi_shellpair const P, struct simint_multi_shellpair const Q, double screen_tol, double * const restrict work, double * const restrict INT__h_s_p_d) { double Q_AB[3*Q.nshell12]; struct simint_multi_shellpair Q_tmp = Q; Q_tmp.PA_x = Q.PB_x; Q_tmp.PA_y = Q.PB_y; Q_tmp.PA_z = Q.PB_z; Q_tmp.PB_x = Q.PA_x; Q_tmp.PB_y = Q.PA_y; Q_tmp.PB_z = Q.PA_z; Q_tmp.AB_x = Q_AB; Q_tmp.AB_y = Q_AB + Q.nshell12; Q_tmp.AB_z = Q_AB + 2*Q.nshell12; for(int i = 0; i < Q.nshell12; i++) { Q_tmp.AB_x[i] = -Q.AB_x[i]; Q_tmp.AB_y[i] = -Q.AB_y[i]; Q_tmp.AB_z[i] = -Q.AB_z[i]; } int ret = ostei_h_s_d_p(P, Q_tmp, screen_tol, work, INT__h_s_p_d); double buffer[378] SIMINT_ALIGN_ARRAY_DBL; for(int q = 0; q < ret; q++) { int idx = 0; for(int a = 0; a < 21; ++a) for(int b = 0; b < 1; ++b) for(int c = 0; c < 3; ++c) for(int d = 0; d < 6; ++d) buffer[idx++] = INT__h_s_p_d[q*378+a*18+b*18+d*3+c]; memcpy(INT__h_s_p_d+q*378, buffer, 378*sizeof(double)); } return ret; } int ostei_s_h_p_d(struct simint_multi_shellpair const P, struct simint_multi_shellpair const Q, double screen_tol, double * const restrict work, double * const restrict INT__s_h_p_d) { double P_AB[3*P.nshell12]; struct simint_multi_shellpair P_tmp = P; P_tmp.PA_x = P.PB_x; P_tmp.PA_y = P.PB_y; P_tmp.PA_z = P.PB_z; P_tmp.PB_x = P.PA_x; P_tmp.PB_y = P.PA_y; P_tmp.PB_z = P.PA_z; P_tmp.AB_x = P_AB; P_tmp.AB_y = P_AB + P.nshell12; P_tmp.AB_z = P_AB + 2*P.nshell12; for(int i = 0; i < P.nshell12; i++) { P_tmp.AB_x[i] = -P.AB_x[i]; P_tmp.AB_y[i] = -P.AB_y[i]; P_tmp.AB_z[i] = -P.AB_z[i]; } double Q_AB[3*Q.nshell12]; struct simint_multi_shellpair Q_tmp = Q; Q_tmp.PA_x = Q.PB_x; Q_tmp.PA_y = Q.PB_y; Q_tmp.PA_z = Q.PB_z; Q_tmp.PB_x = Q.PA_x; Q_tmp.PB_y = Q.PA_y; Q_tmp.PB_z = Q.PA_z; Q_tmp.AB_x = Q_AB; Q_tmp.AB_y = Q_AB + Q.nshell12; Q_tmp.AB_z = Q_AB + 2*Q.nshell12; for(int i = 0; i < Q.nshell12; i++) { Q_tmp.AB_x[i] = -Q.AB_x[i]; Q_tmp.AB_y[i] = -Q.AB_y[i]; Q_tmp.AB_z[i] = -Q.AB_z[i]; } int ret = ostei_h_s_d_p(P_tmp, Q_tmp, screen_tol, work, INT__s_h_p_d); double buffer[378] SIMINT_ALIGN_ARRAY_DBL; for(int q = 0; q < ret; q++) { int idx = 0; for(int a = 0; a < 1; ++a) for(int b = 0; b < 21; ++b) for(int c = 0; c < 3; ++c) for(int d = 0; d < 6; ++d) buffer[idx++] = INT__s_h_p_d[q*378+b*18+a*18+d*3+c]; memcpy(INT__s_h_p_d+q*378, buffer, 378*sizeof(double)); } return ret; }