1 #include "simint/boys/boys.h"
2 #include "simint/ostei/gen/ostei_generated.h"
3 #include "simint/vectorization/vectorization.h"
4 #include <math.h>
5 #include <string.h>
6
7
ostei_h_p_d_s(struct simint_multi_shellpair const P,struct simint_multi_shellpair const Q,double screen_tol,double * const restrict work,double * const restrict INT__h_p_d_s)8 int ostei_h_p_d_s(struct simint_multi_shellpair const P,
9 struct simint_multi_shellpair const Q,
10 double screen_tol,
11 double * const restrict work,
12 double * const restrict INT__h_p_d_s)
13 {
14
15 SIMINT_ASSUME_ALIGN_DBL(work);
16 SIMINT_ASSUME_ALIGN_DBL(INT__h_p_d_s);
17 int ab, cd, abcd;
18 int istart, jstart;
19 int iprimcd, nprim_icd, icd;
20 const int check_screen = (screen_tol > 0.0);
21 int i, j;
22 int n;
23 int not_screened;
24 int real_abcd;
25 int iket;
26
27 // partition workspace
28 double * const INT__h_s_d_s = work + (SIMINT_NSHELL_SIMD * 0);
29 double * const INT__i_s_d_s = work + (SIMINT_NSHELL_SIMD * 126);
30 SIMINT_DBLTYPE * const primwork = (SIMINT_DBLTYPE *)(work + SIMINT_NSHELL_SIMD*294);
31 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_s = primwork + 0;
32 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_s_s = primwork + 9;
33 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_s_s = primwork + 33;
34 SIMINT_DBLTYPE * const restrict PRIM_INT__f_s_s_s = primwork + 75;
35 SIMINT_DBLTYPE * const restrict PRIM_INT__g_s_s_s = primwork + 135;
36 SIMINT_DBLTYPE * const restrict PRIM_INT__g_s_p_s = primwork + 210;
37 SIMINT_DBLTYPE * const restrict PRIM_INT__h_s_s_s = primwork + 300;
38 SIMINT_DBLTYPE * const restrict PRIM_INT__h_s_p_s = primwork + 384;
39 SIMINT_DBLTYPE * const restrict PRIM_INT__h_s_d_s = primwork + 510;
40 SIMINT_DBLTYPE * const restrict PRIM_INT__i_s_s_s = primwork + 636;
41 SIMINT_DBLTYPE * const restrict PRIM_INT__i_s_p_s = primwork + 720;
42 SIMINT_DBLTYPE * const restrict PRIM_INT__i_s_d_s = primwork + 888;
43 double * const hrrwork = (double *)(primwork + 1056);
44
45
46 // Create constants
47 const SIMINT_DBLTYPE const_1 = SIMINT_DBLSET1(1);
48 const SIMINT_DBLTYPE const_2 = SIMINT_DBLSET1(2);
49 const SIMINT_DBLTYPE const_3 = SIMINT_DBLSET1(3);
50 const SIMINT_DBLTYPE const_4 = SIMINT_DBLSET1(4);
51 const SIMINT_DBLTYPE const_5 = SIMINT_DBLSET1(5);
52 const SIMINT_DBLTYPE const_6 = SIMINT_DBLSET1(6);
53 const SIMINT_DBLTYPE one_half = SIMINT_DBLSET1(0.5);
54
55
56 ////////////////////////////////////////
57 // Loop over shells and primitives
58 ////////////////////////////////////////
59
60 real_abcd = 0;
61 istart = 0;
62 for(ab = 0; ab < P.nshell12_clip; ++ab)
63 {
64 const int iend = istart + P.nprim12[ab];
65
66 cd = 0;
67 jstart = 0;
68
69 for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD)
70 {
71 const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD;
72 int jend = jstart;
73 for(i = 0; i < nshellbatch; i++)
74 jend += Q.nprim12[cd+i];
75
76 // Clear the beginning of the workspace (where we are accumulating integrals)
77 memset(work, 0, SIMINT_NSHELL_SIMD * 294 * sizeof(double));
78 abcd = 0;
79
80
81 for(i = istart; i < iend; ++i)
82 {
83 SIMINT_DBLTYPE bra_screen_max; // only used if check_screen
84
85 if(check_screen)
86 {
87 // Skip this whole thing if always insignificant
88 if((P.screen[i] * Q.screen_max) < screen_tol)
89 continue;
90 bra_screen_max = SIMINT_DBLSET1(P.screen[i]);
91 }
92
93 icd = 0;
94 iprimcd = 0;
95 nprim_icd = Q.nprim12[cd];
96 double * restrict PRIM_PTR_INT__h_s_d_s = INT__h_s_d_s + abcd * 126;
97 double * restrict PRIM_PTR_INT__i_s_d_s = INT__i_s_d_s + abcd * 168;
98
99
100
101 // Load these one per loop over i
102 const SIMINT_DBLTYPE P_alpha = SIMINT_DBLSET1(P.alpha[i]);
103 const SIMINT_DBLTYPE P_prefac = SIMINT_DBLSET1(P.prefac[i]);
104 const SIMINT_DBLTYPE Pxyz[3] = { SIMINT_DBLSET1(P.x[i]), SIMINT_DBLSET1(P.y[i]), SIMINT_DBLSET1(P.z[i]) };
105
106 const SIMINT_DBLTYPE P_PA[3] = { SIMINT_DBLSET1(P.PA_x[i]), SIMINT_DBLSET1(P.PA_y[i]), SIMINT_DBLSET1(P.PA_z[i]) };
107
108 for(j = jstart; j < jend; j += SIMINT_SIMD_LEN)
109 {
110 // calculate the shell offsets
111 // these are the offset from the shell pointed to by cd
112 // for each element
113 int shelloffsets[SIMINT_SIMD_LEN] = {0};
114 int lastoffset = 0;
115 const int nlane = ( ((j + SIMINT_SIMD_LEN) < jend) ? SIMINT_SIMD_LEN : (jend - j));
116
117 if((iprimcd + SIMINT_SIMD_LEN) >= nprim_icd)
118 {
119 // Handle if the first element of the vector is a new shell
120 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
121 {
122 nprim_icd += Q.nprim12[cd + (++icd)];
123 PRIM_PTR_INT__h_s_d_s += 126;
124 PRIM_PTR_INT__i_s_d_s += 168;
125 }
126 iprimcd++;
127 for(n = 1; n < SIMINT_SIMD_LEN; ++n)
128 {
129 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
130 {
131 shelloffsets[n] = shelloffsets[n-1] + 1;
132 lastoffset++;
133 nprim_icd += Q.nprim12[cd + (++icd)];
134 }
135 else
136 shelloffsets[n] = shelloffsets[n-1];
137 iprimcd++;
138 }
139 }
140 else
141 iprimcd += SIMINT_SIMD_LEN;
142
143 // Do we have to compute this vector (or has it been screened out)?
144 // (not_screened != 0 means we have to do this vector)
145 if(check_screen)
146 {
147 const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j)));
148 if(vmax < screen_tol)
149 {
150 PRIM_PTR_INT__h_s_d_s += lastoffset*126;
151 PRIM_PTR_INT__i_s_d_s += lastoffset*168;
152 continue;
153 }
154 }
155
156 const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j);
157 const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha);
158 const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha);
159 const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum);
160
161
162 /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */
163 SIMINT_DBLTYPE PQ[3];
164 PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j));
165 PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j));
166 PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j));
167 SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]);
168 R2 = SIMINT_FMADD(PQ[1], PQ[1], R2);
169 R2 = SIMINT_FMADD(PQ[2], PQ[2], R2);
170
171 const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST
172 const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha);
173 const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha);
174 const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p);
175 const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q);
176 const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum);
177 const SIMINT_DBLTYPE Q_PA[3] = { SIMINT_DBLLOAD(Q.PA_x, j), SIMINT_DBLLOAD(Q.PA_y, j), SIMINT_DBLLOAD(Q.PA_z, j) };
178
179 // NOTE: Minus sign!
180 const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p);
181 SIMINT_DBLTYPE aop_PQ[3];
182 aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]);
183 aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]);
184 aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]);
185
186 SIMINT_DBLTYPE a_over_q = SIMINT_MUL(alpha, one_over_q);
187 SIMINT_DBLTYPE aoq_PQ[3];
188 aoq_PQ[0] = SIMINT_MUL(a_over_q, PQ[0]);
189 aoq_PQ[1] = SIMINT_MUL(a_over_q, PQ[1]);
190 aoq_PQ[2] = SIMINT_MUL(a_over_q, PQ[2]);
191 // Put a minus sign here so we don't have to in RR routines
192 a_over_q = SIMINT_NEG(a_over_q);
193
194
195 //////////////////////////////////////////////
196 // Fjt function section
197 // Maximum v value: 8
198 //////////////////////////////////////////////
199 // The parameter to the Fjt function
200 const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha);
201
202
203 const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j);
204
205
206 boys_F_split(PRIM_INT__s_s_s_s, F_x, 8);
207 SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum);
208 prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac);
209 for(n = 0; n <= 8; n++)
210 PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac);
211
212 //////////////////////////////////////////////
213 // Primitive integrals: Vertical recurrance
214 //////////////////////////////////////////////
215
216 const SIMINT_DBLTYPE vrr_const_1_over_2p = one_over_2p;
217 const SIMINT_DBLTYPE vrr_const_2_over_2p = SIMINT_MUL(const_2, one_over_2p);
218 const SIMINT_DBLTYPE vrr_const_3_over_2p = SIMINT_MUL(const_3, one_over_2p);
219 const SIMINT_DBLTYPE vrr_const_4_over_2p = SIMINT_MUL(const_4, one_over_2p);
220 const SIMINT_DBLTYPE vrr_const_5_over_2p = SIMINT_MUL(const_5, one_over_2p);
221 const SIMINT_DBLTYPE vrr_const_1_over_2q = one_over_2q;
222 const SIMINT_DBLTYPE vrr_const_1_over_2pq = one_over_2pq;
223 const SIMINT_DBLTYPE vrr_const_2_over_2pq = SIMINT_MUL(const_2, one_over_2pq);
224 const SIMINT_DBLTYPE vrr_const_3_over_2pq = SIMINT_MUL(const_3, one_over_2pq);
225 const SIMINT_DBLTYPE vrr_const_4_over_2pq = SIMINT_MUL(const_4, one_over_2pq);
226 const SIMINT_DBLTYPE vrr_const_5_over_2pq = SIMINT_MUL(const_5, one_over_2pq);
227 const SIMINT_DBLTYPE vrr_const_6_over_2pq = SIMINT_MUL(const_6, one_over_2pq);
228
229
230
231 // Forming PRIM_INT__p_s_s_s[8 * 3];
232 for(n = 0; n < 8; ++n) // loop over orders of auxiliary function
233 {
234
235 PRIM_INT__p_s_s_s[n * 3 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_s_s[n * 1 + 0]);
236 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]);
237
238 PRIM_INT__p_s_s_s[n * 3 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_s_s[n * 1 + 0]);
239 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]);
240
241 PRIM_INT__p_s_s_s[n * 3 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_s_s[n * 1 + 0]);
242 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]);
243
244 }
245
246
247
248 // Forming PRIM_INT__d_s_s_s[7 * 6];
249 for(n = 0; n < 7; ++n) // loop over orders of auxiliary function
250 {
251
252 PRIM_INT__d_s_s_s[n * 6 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__p_s_s_s[n * 3 + 0]);
253 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]);
254 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]);
255
256 PRIM_INT__d_s_s_s[n * 6 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__p_s_s_s[n * 3 + 0]);
257 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]);
258
259 PRIM_INT__d_s_s_s[n * 6 + 3] = SIMINT_MUL(P_PA[1], PRIM_INT__p_s_s_s[n * 3 + 1]);
260 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]);
261 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]);
262
263 PRIM_INT__d_s_s_s[n * 6 + 5] = SIMINT_MUL(P_PA[2], PRIM_INT__p_s_s_s[n * 3 + 2]);
264 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]);
265 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]);
266
267 }
268
269
270
271 // Forming PRIM_INT__f_s_s_s[6 * 10];
272 for(n = 0; n < 6; ++n) // loop over orders of auxiliary function
273 {
274
275 PRIM_INT__f_s_s_s[n * 10 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__d_s_s_s[n * 6 + 0]);
276 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]);
277 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]);
278
279 PRIM_INT__f_s_s_s[n * 10 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 0]);
280 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]);
281
282 PRIM_INT__f_s_s_s[n * 10 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 0]);
283 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]);
284
285 PRIM_INT__f_s_s_s[n * 10 + 3] = SIMINT_MUL(P_PA[0], PRIM_INT__d_s_s_s[n * 6 + 3]);
286 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]);
287
288 PRIM_INT__f_s_s_s[n * 10 + 4] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 1]);
289 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]);
290
291 PRIM_INT__f_s_s_s[n * 10 + 5] = SIMINT_MUL(P_PA[0], PRIM_INT__d_s_s_s[n * 6 + 5]);
292 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]);
293
294 PRIM_INT__f_s_s_s[n * 10 + 6] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 3]);
295 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]);
296 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]);
297
298 PRIM_INT__f_s_s_s[n * 10 + 7] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 3]);
299 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]);
300
301 PRIM_INT__f_s_s_s[n * 10 + 8] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 5]);
302 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]);
303
304 PRIM_INT__f_s_s_s[n * 10 + 9] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 5]);
305 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]);
306 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]);
307
308 }
309
310
311 VRR_I_g_s_s_s(
312 PRIM_INT__g_s_s_s,
313 PRIM_INT__f_s_s_s,
314 PRIM_INT__d_s_s_s,
315 P_PA,
316 a_over_p,
317 aop_PQ,
318 one_over_2p,
319 5);
320
321
322 VRR_I_h_s_s_s(
323 PRIM_INT__h_s_s_s,
324 PRIM_INT__g_s_s_s,
325 PRIM_INT__f_s_s_s,
326 P_PA,
327 a_over_p,
328 aop_PQ,
329 one_over_2p,
330 4);
331
332
333 ostei_general_vrr_K(5, 0, 1, 0, 2,
334 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
335 PRIM_INT__h_s_s_s, NULL, NULL, PRIM_INT__g_s_s_s, NULL, PRIM_INT__h_s_p_s);
336
337
338 VRR_K_g_s_p_s(
339 PRIM_INT__g_s_p_s,
340 PRIM_INT__g_s_s_s,
341 PRIM_INT__f_s_s_s,
342 Q_PA,
343 aoq_PQ,
344 one_over_2pq,
345 2);
346
347
348 ostei_general_vrr_K(5, 0, 2, 0, 1,
349 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
350 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);
351
352
353 ostei_general_vrr1_I(6, 3,
354 one_over_2p, a_over_p, aop_PQ, P_PA,
355 PRIM_INT__h_s_s_s, PRIM_INT__g_s_s_s, PRIM_INT__i_s_s_s);
356
357
358 ostei_general_vrr_K(6, 0, 1, 0, 2,
359 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
360 PRIM_INT__i_s_s_s, NULL, NULL, PRIM_INT__h_s_s_s, NULL, PRIM_INT__i_s_p_s);
361
362
363 ostei_general_vrr_K(6, 0, 2, 0, 1,
364 one_over_2q, a_over_q, one_over_2pq, aoq_PQ, Q_PA,
365 PRIM_INT__i_s_p_s, PRIM_INT__i_s_s_s, NULL, PRIM_INT__h_s_p_s, NULL, PRIM_INT__i_s_d_s);
366
367
368
369
370 ////////////////////////////////////
371 // Accumulate contracted integrals
372 ////////////////////////////////////
373 if(lastoffset == 0)
374 {
375 contract_all(126, PRIM_INT__h_s_d_s, PRIM_PTR_INT__h_s_d_s);
376 contract_all(168, PRIM_INT__i_s_d_s, PRIM_PTR_INT__i_s_d_s);
377 }
378 else
379 {
380 contract(126, shelloffsets, PRIM_INT__h_s_d_s, PRIM_PTR_INT__h_s_d_s);
381 contract(168, shelloffsets, PRIM_INT__i_s_d_s, PRIM_PTR_INT__i_s_d_s);
382 PRIM_PTR_INT__h_s_d_s += lastoffset*126;
383 PRIM_PTR_INT__i_s_d_s += lastoffset*168;
384 }
385
386 } // close loop over j
387 } // close loop over i
388
389 //Advance to the next batch
390 jstart = SIMINT_SIMD_ROUND(jend);
391
392 //////////////////////////////////////////////
393 // Contracted integrals: Horizontal recurrance
394 //////////////////////////////////////////////
395
396
397 const double hAB[3] = { P.AB_x[ab], P.AB_y[ab], P.AB_z[ab] };
398
399
400 for(abcd = 0; abcd < nshellbatch; ++abcd, ++real_abcd)
401 {
402
403 // set up HRR pointers
404 double const * restrict HRR_INT__h_s_d_s = INT__h_s_d_s + abcd * 126;
405 double const * restrict HRR_INT__i_s_d_s = INT__i_s_d_s + abcd * 168;
406 double * restrict HRR_INT__h_p_d_s = INT__h_p_d_s + real_abcd * 378;
407
408 // form INT__h_p_d_s
409 ostei_general_hrr_J(5, 1, 2, 0, hAB, HRR_INT__i_s_d_s, HRR_INT__h_s_d_s, HRR_INT__h_p_d_s);
410
411
412 } // close HRR loop
413
414
415 } // close loop cdbatch
416
417 istart = iend;
418 } // close loop over ab
419
420 return P.nshell12_clip * Q.nshell12_clip;
421 }
422
ostei_p_h_d_s(struct simint_multi_shellpair const P,struct simint_multi_shellpair const Q,double screen_tol,double * const restrict work,double * const restrict INT__p_h_d_s)423 int ostei_p_h_d_s(struct simint_multi_shellpair const P,
424 struct simint_multi_shellpair const Q,
425 double screen_tol,
426 double * const restrict work,
427 double * const restrict INT__p_h_d_s)
428 {
429 double P_AB[3*P.nshell12];
430 struct simint_multi_shellpair P_tmp = P;
431 P_tmp.PA_x = P.PB_x; P_tmp.PA_y = P.PB_y; P_tmp.PA_z = P.PB_z;
432 P_tmp.PB_x = P.PA_x; P_tmp.PB_y = P.PA_y; P_tmp.PB_z = P.PA_z;
433 P_tmp.AB_x = P_AB;
434 P_tmp.AB_y = P_AB + P.nshell12;
435 P_tmp.AB_z = P_AB + 2*P.nshell12;
436
437 for(int i = 0; i < P.nshell12; i++)
438 {
439 P_tmp.AB_x[i] = -P.AB_x[i];
440 P_tmp.AB_y[i] = -P.AB_y[i];
441 P_tmp.AB_z[i] = -P.AB_z[i];
442 }
443
444 int ret = ostei_h_p_d_s(P_tmp, Q, screen_tol, work, INT__p_h_d_s);
445 double buffer[378] SIMINT_ALIGN_ARRAY_DBL;
446
447 for(int q = 0; q < ret; q++)
448 {
449 int idx = 0;
450 for(int a = 0; a < 3; ++a)
451 for(int b = 0; b < 21; ++b)
452 for(int c = 0; c < 6; ++c)
453 for(int d = 0; d < 1; ++d)
454 buffer[idx++] = INT__p_h_d_s[q*378+b*18+a*6+c*1+d];
455
456 memcpy(INT__p_h_d_s+q*378, buffer, 378*sizeof(double));
457 }
458
459 return ret;
460 }
461
ostei_h_p_s_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_p_s_d)462 int ostei_h_p_s_d(struct simint_multi_shellpair const P,
463 struct simint_multi_shellpair const Q,
464 double screen_tol,
465 double * const restrict work,
466 double * const restrict INT__h_p_s_d)
467 {
468 double Q_AB[3*Q.nshell12];
469 struct simint_multi_shellpair Q_tmp = Q;
470 Q_tmp.PA_x = Q.PB_x; Q_tmp.PA_y = Q.PB_y; Q_tmp.PA_z = Q.PB_z;
471 Q_tmp.PB_x = Q.PA_x; Q_tmp.PB_y = Q.PA_y; Q_tmp.PB_z = Q.PA_z;
472 Q_tmp.AB_x = Q_AB;
473 Q_tmp.AB_y = Q_AB + Q.nshell12;
474 Q_tmp.AB_z = Q_AB + 2*Q.nshell12;
475
476 for(int i = 0; i < Q.nshell12; i++)
477 {
478 Q_tmp.AB_x[i] = -Q.AB_x[i];
479 Q_tmp.AB_y[i] = -Q.AB_y[i];
480 Q_tmp.AB_z[i] = -Q.AB_z[i];
481 }
482
483 int ret = ostei_h_p_d_s(P, Q_tmp, screen_tol, work, INT__h_p_s_d);
484 double buffer[378] SIMINT_ALIGN_ARRAY_DBL;
485
486 for(int q = 0; q < ret; q++)
487 {
488 int idx = 0;
489 for(int a = 0; a < 21; ++a)
490 for(int b = 0; b < 3; ++b)
491 for(int c = 0; c < 1; ++c)
492 for(int d = 0; d < 6; ++d)
493 buffer[idx++] = INT__h_p_s_d[q*378+a*18+b*6+d*1+c];
494
495 memcpy(INT__h_p_s_d+q*378, buffer, 378*sizeof(double));
496 }
497
498 return ret;
499 }
500
ostei_p_h_s_d(struct simint_multi_shellpair const P,struct simint_multi_shellpair const Q,double screen_tol,double * const restrict work,double * const restrict INT__p_h_s_d)501 int ostei_p_h_s_d(struct simint_multi_shellpair const P,
502 struct simint_multi_shellpair const Q,
503 double screen_tol,
504 double * const restrict work,
505 double * const restrict INT__p_h_s_d)
506 {
507 double P_AB[3*P.nshell12];
508 struct simint_multi_shellpair P_tmp = P;
509 P_tmp.PA_x = P.PB_x; P_tmp.PA_y = P.PB_y; P_tmp.PA_z = P.PB_z;
510 P_tmp.PB_x = P.PA_x; P_tmp.PB_y = P.PA_y; P_tmp.PB_z = P.PA_z;
511 P_tmp.AB_x = P_AB;
512 P_tmp.AB_y = P_AB + P.nshell12;
513 P_tmp.AB_z = P_AB + 2*P.nshell12;
514
515 for(int i = 0; i < P.nshell12; i++)
516 {
517 P_tmp.AB_x[i] = -P.AB_x[i];
518 P_tmp.AB_y[i] = -P.AB_y[i];
519 P_tmp.AB_z[i] = -P.AB_z[i];
520 }
521
522 double Q_AB[3*Q.nshell12];
523 struct simint_multi_shellpair Q_tmp = Q;
524 Q_tmp.PA_x = Q.PB_x; Q_tmp.PA_y = Q.PB_y; Q_tmp.PA_z = Q.PB_z;
525 Q_tmp.PB_x = Q.PA_x; Q_tmp.PB_y = Q.PA_y; Q_tmp.PB_z = Q.PA_z;
526 Q_tmp.AB_x = Q_AB;
527 Q_tmp.AB_y = Q_AB + Q.nshell12;
528 Q_tmp.AB_z = Q_AB + 2*Q.nshell12;
529
530 for(int i = 0; i < Q.nshell12; i++)
531 {
532 Q_tmp.AB_x[i] = -Q.AB_x[i];
533 Q_tmp.AB_y[i] = -Q.AB_y[i];
534 Q_tmp.AB_z[i] = -Q.AB_z[i];
535 }
536
537 int ret = ostei_h_p_d_s(P_tmp, Q_tmp, screen_tol, work, INT__p_h_s_d);
538 double buffer[378] SIMINT_ALIGN_ARRAY_DBL;
539
540 for(int q = 0; q < ret; q++)
541 {
542 int idx = 0;
543 for(int a = 0; a < 3; ++a)
544 for(int b = 0; b < 21; ++b)
545 for(int c = 0; c < 1; ++c)
546 for(int d = 0; d < 6; ++d)
547 buffer[idx++] = INT__p_h_s_d[q*378+b*18+a*6+d*1+c];
548
549 memcpy(INT__p_h_s_d+q*378, buffer, 378*sizeof(double));
550 }
551
552 return ret;
553 }
554
555