LIBINT  2.6.0
VRR_GTG_1d_xx_xx.h
1 /*
2  * Copyright (C) 2004-2019 Edward F. Valeev
3  *
4  * This file is part of Libint.
5  *
6  * Libint is free software: you can redistribute it and/or modify
7  * it under the terms of the GNU Lesser General Public License as published by
8  * the Free Software Foundation, either version 3 of the License, or
9  * (at your option) any later version.
10  *
11  * Libint is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14  * GNU Lesser General Public License for more details.
15  *
16  * You should have received a copy of the GNU Lesser General Public License
17  * along with Libint. If not, see <http://www.gnu.org/licenses/>.
18  *
19  */
20 
21 #ifndef _libint2_src_lib_libint_vrrgtg1dxxxx_h_
22 #define _libint2_src_lib_libint_vrrgtg1dxxxx_h_
23 
24 #include <cstdlib>
25 #include <cassert>
26 #include <libint2.h>
27 #include <util_types.h>
28 
29 namespace libint2 {
30 
38  template <unsigned int CartesianAxis, int La, int Lb, int Lc, int Ld, bool vectorize>
40 
41  static void compute(const Libint_t* inteval,
42  LIBINT2_REALTYPE* target,
43  const LIBINT2_REALTYPE* src0) {
44 
45  enum XYZ {x=0, y=1, z=2};
46  assert(CartesianAxis == x || CartesianAxis == y || CartesianAxis == z);
47  assert(vectorize == false);
48 
49  const unsigned int veclen = vectorize ? inteval->veclen : 1;
50 
51  // corner case: (00|00)
52  if (La == 0 && Lb == 0 && Lc == 0 && Ld == 0) {
53  for (unsigned int v=0; v!=veclen; ++v)
54  target[v] = src0[v];
55  return;
56  }
57 
58  //---------------------------------------------
59  // Part (1): build (a+b 0|c+d 0)
60  //---------------------------------------------
61 
62  LIBINT2_REALTYPE apb_0_GTG_cpd_0[La+Lb+1][Lc+Ld+1];
63  apb_0_GTG_cpd_0[0][0] = src0[0];
64 
65  const LIBINT2_REALTYPE *pfac0_0, *pfac0_1;
66  const LIBINT2_REALTYPE *pfac1_0 = inteval->R12kG12_pfac1_0;
67  const LIBINT2_REALTYPE *pfac1_1 = inteval->R12kG12_pfac1_1;
68  const LIBINT2_REALTYPE *pfac2 = inteval->R12kG12_pfac2;
69  switch (CartesianAxis) {
70  case x:
71  pfac0_0 = inteval->R12kG12_pfac0_0_x;
72  pfac0_1 = inteval->R12kG12_pfac0_1_x;
73  break;
74  case y:
75  pfac0_0 = inteval->R12kG12_pfac0_0_y;
76  pfac0_1 = inteval->R12kG12_pfac0_1_y;
77  break;
78  case z:
79  pfac0_0 = inteval->R12kG12_pfac0_0_z;
80  pfac0_1 = inteval->R12kG12_pfac0_1_z;
81  break;
82  default: assert(false);
83  }
84 
85  // build (0 0|1 0)
86  if (Lc+Ld > 0) {
87  apb_0_GTG_cpd_0[0][1] = pfac0_1[0] * apb_0_GTG_cpd_0[0][0];
88 #if LIBINT2_FLOP_COUNT
89  inteval->nflops[0] += 1;
90 #endif
91  }
92 
93  // build (0 0|c+d 0)
94  if (Lc+Ld > 1) {
95  for(int c_plus_d=1; c_plus_d!=Lc+Ld; ++c_plus_d) {
96  apb_0_GTG_cpd_0[0][c_plus_d+1] = pfac0_1[0] * apb_0_GTG_cpd_0[0][c_plus_d] +
97  c_plus_d * pfac1_1[0] * apb_0_GTG_cpd_0[0][c_plus_d-1];
98  }
99 #if LIBINT2_FLOP_COUNT
100  inteval->nflops[0] += 4*(Lc+Ld-1);
101 #endif
102  }
103 
104  // build (1 0|0 0)
105  if (La+Lb > 0) {
106  apb_0_GTG_cpd_0[1][0] = pfac0_0[0] * apb_0_GTG_cpd_0[0][0];
107 #if LIBINT2_FLOP_COUNT
108  inteval->nflops[0] += 1;
109 #endif
110  }
111 
112  // build (a+b 0|0 0)
113  if (La+Lb > 1) {
114  for(int a_plus_b=1; a_plus_b!=La+Lb; ++a_plus_b) {
115  apb_0_GTG_cpd_0[a_plus_b+1][0] = pfac0_0[0] * apb_0_GTG_cpd_0[a_plus_b][0] +
116  a_plus_b * pfac1_0[0] * apb_0_GTG_cpd_0[a_plus_b-1][0];
117  }
118 #if LIBINT2_FLOP_COUNT
119  inteval->nflops[0] += 4*(La+Lb-1);
120 #endif
121  }
122 
123  // build (1 0|c+d 0)
124  if (La+Lb > 0 && Lc+Ld > 0) {
125  for(int c_plus_d=1; c_plus_d<=Lc+Ld; ++c_plus_d) {
126  apb_0_GTG_cpd_0[1][c_plus_d] = pfac0_0[0] * apb_0_GTG_cpd_0[0][c_plus_d] +
127  c_plus_d * pfac2[0] * apb_0_GTG_cpd_0[0][c_plus_d-1];
128  }
129 #if LIBINT2_FLOP_COUNT
130  inteval->nflops[0] += 4*(Lc+Ld-1);
131 #endif
132  }
133 
134  // build (a+b 0|c+d 0)
135  if (La+Lb > 1 && Lc+Ld > 0) {
136  for(int a_plus_b=1; a_plus_b!=La+Lb; ++a_plus_b) {
137  for(int c_plus_d=1; c_plus_d<=Lc+Ld; ++c_plus_d) {
138  apb_0_GTG_cpd_0[a_plus_b+1][c_plus_d] = pfac0_0[0] * apb_0_GTG_cpd_0[a_plus_b][c_plus_d] +
139  a_plus_b * pfac1_0[0] * apb_0_GTG_cpd_0[a_plus_b-1][c_plus_d] +
140  c_plus_d * pfac2[0] * apb_0_GTG_cpd_0[a_plus_b][c_plus_d-1];
141  }
142  }
143 #if LIBINT2_FLOP_COUNT
144  inteval->nflops[0] += 7*(La+Lb-1)*(Lc+Ld-1);
145 #endif
146  }
147 
148  //---------------------------------------------
149  // Part (2): build (a b|c+d 0)
150  //---------------------------------------------
151 
152  const LIBINT2_REALTYPE* AB;
153  switch (CartesianAxis) {
154  case x:
155  AB = inteval->AB_x;
156  break;
157  case y:
158  AB = inteval->AB_y;
159  break;
160  case z:
161  AB = inteval->AB_z;
162  break;
163  default: assert(false);
164  }
165 
166  LIBINT2_REALTYPE a_b_GTG_cpd_0[La+1][Lb+1][Lc+Ld+1];
167  for(int c_plus_d=0; c_plus_d<=Lc+Ld; ++c_plus_d) {
168  // copy (a+b 0| to a local 0,a+b buffer
169  LIBINT2_REALTYPE b_a_GTG[La+Lb+1][La+Lb+1];
170  for(int a_plus_b=0; a_plus_b<=La+Lb; ++a_plus_b) {
171  b_a_GTG[0][a_plus_b] = apb_0_GTG_cpd_0[a_plus_b][c_plus_d];
172  }
173  // use HRR to compute b,a
174  for(int b=1; b<=Lb; ++b) {
175  for(int a=0; a<=La+Lb-b; ++a) {
176  b_a_GTG[b][a] = b_a_GTG[b-1][a+1] + AB[0] * b_a_GTG[b-1][a];
177  }
178 #if LIBINT2_FLOP_COUNT
179  inteval->nflops[0] += 2 * (La+Lb-b+1);
180 #endif
181  }
182  // copy b,a to (a b|
183  for(int b=0; b<=Lb; ++b) {
184  for(int a=0; a<=La; ++a) {
185  a_b_GTG_cpd_0[a][b][c_plus_d] = b_a_GTG[b][a];
186  }
187  }
188  }
189 
190  //---------------------------------------------
191  // Part (3): build (a b|c d)
192  //---------------------------------------------
193 
194  const LIBINT2_REALTYPE* CD;
195  switch (CartesianAxis) {
196  case x:
197  CD = inteval->CD_x;
198  break;
199  case y:
200  CD = inteval->CD_y;
201  break;
202  case z:
203  CD = inteval->CD_z;
204  break;
205  default: assert(false);
206  }
207 
208  LIBINT2_REALTYPE* target_a_b_blk_ptr = target;
209  const int Nd = (Ld+1);
210  const int Ncd = (Lc+1)*Nd;
211  for(int a=0; a<=La; ++a) {
212  for(int b=0; b<=Lb; ++b, target_a_b_blk_ptr+=Ncd) {
213  // copy |c+d 0) to a local 0,c+d buffer
214  LIBINT2_REALTYPE d_c_GTG[Lc+Ld+1][Lc+Ld+1];
215  for(int c_plus_d=0; c_plus_d<=Lc+Ld; ++c_plus_d) {
216  d_c_GTG[0][c_plus_d] = a_b_GTG_cpd_0[a][b][c_plus_d];
217  }
218  // use HRR to compute d,c
219  for(int d=1; d<=Ld; ++d) {
220  for(int c=0; c<=Lc+Ld-d; ++c) {
221  d_c_GTG[d][c] = d_c_GTG[d-1][c+1] + CD[0] * d_c_GTG[d-1][c];
222  }
223 #if LIBINT2_FLOP_COUNT
224  inteval->nflops[0] += 2 * (Lc+Ld-d+1);
225 #endif
226  }
227  // copy d,c to |c d)
228  for(int d=0; d<=Ld; ++d) {
229  for(int c=0, cd=d; c<=Lc; ++c, cd+=Nd) {
230  target_a_b_blk_ptr[cd] = d_c_GTG[d][c];
231  }
232  }
233  }
234  }
235 
236  // done
237  }
238 
239  };
240 
241 };
242 
243 #endif // header guard
244 
Defaults definitions for various parameters assumed by Libint.
Definition: algebra.cc:24
builds (ab| GTG_1d |cd), the shell set of 2-dimensional integrals needed for Rys quadrature evaluatio...
Definition: VRR_GTG_1d_xx_xx.h:39