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_p_s_d_f(struct simint_multi_shellpair const P,struct simint_multi_shellpair const Q,double screen_tol,double * const restrict work,double * const restrict INT__p_s_d_f)8 int ostei_p_s_d_f(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__p_s_d_f)
13 {
14
15 SIMINT_ASSUME_ALIGN_DBL(work);
16 SIMINT_ASSUME_ALIGN_DBL(INT__p_s_d_f);
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 ibra;
26
27 // partition workspace
28 double * const INT__p_s_s_f = work + (SIMINT_NSHELL_SIMD * 0);
29 double * const INT__p_s_s_g = work + (SIMINT_NSHELL_SIMD * 30);
30 double * const INT__p_s_s_h = work + (SIMINT_NSHELL_SIMD * 75);
31 SIMINT_DBLTYPE * const primwork = (SIMINT_DBLTYPE *)(work + SIMINT_NSHELL_SIMD*138);
32 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_s = primwork + 0;
33 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_p = primwork + 7;
34 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_d = primwork + 25;
35 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_f = primwork + 55;
36 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_g = primwork + 95;
37 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_h = primwork + 140;
38 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_s_f = primwork + 182;
39 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_s_g = primwork + 212;
40 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_s_h = primwork + 257;
41 double * const hrrwork = (double *)(primwork + 320);
42 double * const HRR_INT__p_s_p_f = hrrwork + 0;
43 double * const HRR_INT__p_s_p_g = hrrwork + 90;
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 one_half = SIMINT_DBLSET1(0.5);
53
54
55 ////////////////////////////////////////
56 // Loop over shells and primitives
57 ////////////////////////////////////////
58
59 real_abcd = 0;
60 istart = 0;
61 for(ab = 0; ab < P.nshell12_clip; ++ab)
62 {
63 const int iend = istart + P.nprim12[ab];
64
65 cd = 0;
66 jstart = 0;
67
68 for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD)
69 {
70 const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD;
71 int jend = jstart;
72 for(i = 0; i < nshellbatch; i++)
73 jend += Q.nprim12[cd+i];
74
75 // Clear the beginning of the workspace (where we are accumulating integrals)
76 memset(work, 0, SIMINT_NSHELL_SIMD * 138 * sizeof(double));
77 abcd = 0;
78
79
80 for(i = istart; i < iend; ++i)
81 {
82 SIMINT_DBLTYPE bra_screen_max; // only used if check_screen
83
84 if(check_screen)
85 {
86 // Skip this whole thing if always insignificant
87 if((P.screen[i] * Q.screen_max) < screen_tol)
88 continue;
89 bra_screen_max = SIMINT_DBLSET1(P.screen[i]);
90 }
91
92 icd = 0;
93 iprimcd = 0;
94 nprim_icd = Q.nprim12[cd];
95 double * restrict PRIM_PTR_INT__p_s_s_f = INT__p_s_s_f + abcd * 30;
96 double * restrict PRIM_PTR_INT__p_s_s_g = INT__p_s_s_g + abcd * 45;
97 double * restrict PRIM_PTR_INT__p_s_s_h = INT__p_s_s_h + abcd * 63;
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__p_s_s_f += 30;
124 PRIM_PTR_INT__p_s_s_g += 45;
125 PRIM_PTR_INT__p_s_s_h += 63;
126 }
127 iprimcd++;
128 for(n = 1; n < SIMINT_SIMD_LEN; ++n)
129 {
130 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
131 {
132 shelloffsets[n] = shelloffsets[n-1] + 1;
133 lastoffset++;
134 nprim_icd += Q.nprim12[cd + (++icd)];
135 }
136 else
137 shelloffsets[n] = shelloffsets[n-1];
138 iprimcd++;
139 }
140 }
141 else
142 iprimcd += SIMINT_SIMD_LEN;
143
144 // Do we have to compute this vector (or has it been screened out)?
145 // (not_screened != 0 means we have to do this vector)
146 if(check_screen)
147 {
148 const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j)));
149 if(vmax < screen_tol)
150 {
151 PRIM_PTR_INT__p_s_s_f += lastoffset*30;
152 PRIM_PTR_INT__p_s_s_g += lastoffset*45;
153 PRIM_PTR_INT__p_s_s_h += lastoffset*63;
154 continue;
155 }
156 }
157
158 const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j);
159 const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha);
160 const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha);
161 const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum);
162
163
164 /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */
165 SIMINT_DBLTYPE PQ[3];
166 PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j));
167 PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j));
168 PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j));
169 SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]);
170 R2 = SIMINT_FMADD(PQ[1], PQ[1], R2);
171 R2 = SIMINT_FMADD(PQ[2], PQ[2], R2);
172
173 const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST
174 const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha);
175 const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha);
176 const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p);
177 const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q);
178 const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum);
179 const SIMINT_DBLTYPE Q_PB[3] = { SIMINT_DBLLOAD(Q.PB_x, j), SIMINT_DBLLOAD(Q.PB_y, j), SIMINT_DBLLOAD(Q.PB_z, j) };
180
181 // NOTE: Minus sign!
182 const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p);
183 SIMINT_DBLTYPE aop_PQ[3];
184 aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]);
185 aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]);
186 aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]);
187
188 SIMINT_DBLTYPE a_over_q = SIMINT_MUL(alpha, one_over_q);
189 SIMINT_DBLTYPE aoq_PQ[3];
190 aoq_PQ[0] = SIMINT_MUL(a_over_q, PQ[0]);
191 aoq_PQ[1] = SIMINT_MUL(a_over_q, PQ[1]);
192 aoq_PQ[2] = SIMINT_MUL(a_over_q, PQ[2]);
193 // Put a minus sign here so we don't have to in RR routines
194 a_over_q = SIMINT_NEG(a_over_q);
195
196
197 //////////////////////////////////////////////
198 // Fjt function section
199 // Maximum v value: 6
200 //////////////////////////////////////////////
201 // The parameter to the Fjt function
202 const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha);
203
204
205 const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j);
206
207
208 boys_F_split(PRIM_INT__s_s_s_s, F_x, 6);
209 SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum);
210 prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac);
211 for(n = 0; n <= 6; n++)
212 PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac);
213
214 //////////////////////////////////////////////
215 // Primitive integrals: Vertical recurrance
216 //////////////////////////////////////////////
217
218 const SIMINT_DBLTYPE vrr_const_1_over_2q = one_over_2q;
219 const SIMINT_DBLTYPE vrr_const_2_over_2q = SIMINT_MUL(const_2, one_over_2q);
220 const SIMINT_DBLTYPE vrr_const_3_over_2q = SIMINT_MUL(const_3, one_over_2q);
221 const SIMINT_DBLTYPE vrr_const_4_over_2q = SIMINT_MUL(const_4, 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
228
229
230 // Forming PRIM_INT__s_s_s_p[6 * 3];
231 for(n = 0; n < 6; ++n) // loop over orders of auxiliary function
232 {
233
234 PRIM_INT__s_s_s_p[n * 3 + 0] = SIMINT_MUL(Q_PB[0], PRIM_INT__s_s_s_s[n * 1 + 0]);
235 PRIM_INT__s_s_s_p[n * 3 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_p[n * 3 + 0]);
236
237 PRIM_INT__s_s_s_p[n * 3 + 1] = SIMINT_MUL(Q_PB[1], PRIM_INT__s_s_s_s[n * 1 + 0]);
238 PRIM_INT__s_s_s_p[n * 3 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_p[n * 3 + 1]);
239
240 PRIM_INT__s_s_s_p[n * 3 + 2] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_s[n * 1 + 0]);
241 PRIM_INT__s_s_s_p[n * 3 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_p[n * 3 + 2]);
242
243 }
244
245
246
247 // Forming PRIM_INT__s_s_s_d[5 * 6];
248 for(n = 0; n < 5; ++n) // loop over orders of auxiliary function
249 {
250
251 PRIM_INT__s_s_s_d[n * 6 + 0] = SIMINT_MUL(Q_PB[0], PRIM_INT__s_s_s_p[n * 3 + 0]);
252 PRIM_INT__s_s_s_d[n * 6 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_s_p[(n+1) * 3 + 0], PRIM_INT__s_s_s_d[n * 6 + 0]);
253 PRIM_INT__s_s_s_d[n * 6 + 0] = SIMINT_FMADD( vrr_const_1_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__s_s_s_d[n * 6 + 0]);
254
255 PRIM_INT__s_s_s_d[n * 6 + 1] = SIMINT_MUL(Q_PB[1], PRIM_INT__s_s_s_p[n * 3 + 0]);
256 PRIM_INT__s_s_s_d[n * 6 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_s_p[(n+1) * 3 + 0], PRIM_INT__s_s_s_d[n * 6 + 1]);
257
258 PRIM_INT__s_s_s_d[n * 6 + 2] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_p[n * 3 + 0]);
259 PRIM_INT__s_s_s_d[n * 6 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_p[(n+1) * 3 + 0], PRIM_INT__s_s_s_d[n * 6 + 2]);
260
261 PRIM_INT__s_s_s_d[n * 6 + 3] = SIMINT_MUL(Q_PB[1], PRIM_INT__s_s_s_p[n * 3 + 1]);
262 PRIM_INT__s_s_s_d[n * 6 + 3] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_s_p[(n+1) * 3 + 1], PRIM_INT__s_s_s_d[n * 6 + 3]);
263 PRIM_INT__s_s_s_d[n * 6 + 3] = SIMINT_FMADD( vrr_const_1_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__s_s_s_d[n * 6 + 3]);
264
265 PRIM_INT__s_s_s_d[n * 6 + 4] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_p[n * 3 + 1]);
266 PRIM_INT__s_s_s_d[n * 6 + 4] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_p[(n+1) * 3 + 1], PRIM_INT__s_s_s_d[n * 6 + 4]);
267
268 PRIM_INT__s_s_s_d[n * 6 + 5] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_p[n * 3 + 2]);
269 PRIM_INT__s_s_s_d[n * 6 + 5] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_p[(n+1) * 3 + 2], PRIM_INT__s_s_s_d[n * 6 + 5]);
270 PRIM_INT__s_s_s_d[n * 6 + 5] = SIMINT_FMADD( vrr_const_1_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_s[(n+1) * 1 + 0], PRIM_INT__s_s_s_s[n * 1 + 0]), PRIM_INT__s_s_s_d[n * 6 + 5]);
271
272 }
273
274
275
276 // Forming PRIM_INT__s_s_s_f[4 * 10];
277 for(n = 0; n < 4; ++n) // loop over orders of auxiliary function
278 {
279
280 PRIM_INT__s_s_s_f[n * 10 + 0] = SIMINT_MUL(Q_PB[0], PRIM_INT__s_s_s_d[n * 6 + 0]);
281 PRIM_INT__s_s_s_f[n * 10 + 0] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_s_d[(n+1) * 6 + 0], PRIM_INT__s_s_s_f[n * 10 + 0]);
282 PRIM_INT__s_s_s_f[n * 10 + 0] = SIMINT_FMADD( vrr_const_2_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_p[(n+1) * 3 + 0], PRIM_INT__s_s_s_p[n * 3 + 0]), PRIM_INT__s_s_s_f[n * 10 + 0]);
283
284 PRIM_INT__s_s_s_f[n * 10 + 1] = SIMINT_MUL(Q_PB[1], PRIM_INT__s_s_s_d[n * 6 + 0]);
285 PRIM_INT__s_s_s_f[n * 10 + 1] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_s_d[(n+1) * 6 + 0], PRIM_INT__s_s_s_f[n * 10 + 1]);
286
287 PRIM_INT__s_s_s_f[n * 10 + 2] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_d[n * 6 + 0]);
288 PRIM_INT__s_s_s_f[n * 10 + 2] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_d[(n+1) * 6 + 0], PRIM_INT__s_s_s_f[n * 10 + 2]);
289
290 PRIM_INT__s_s_s_f[n * 10 + 3] = SIMINT_MUL(Q_PB[0], PRIM_INT__s_s_s_d[n * 6 + 3]);
291 PRIM_INT__s_s_s_f[n * 10 + 3] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_s_d[(n+1) * 6 + 3], PRIM_INT__s_s_s_f[n * 10 + 3]);
292
293 PRIM_INT__s_s_s_f[n * 10 + 4] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_d[n * 6 + 1]);
294 PRIM_INT__s_s_s_f[n * 10 + 4] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_d[(n+1) * 6 + 1], PRIM_INT__s_s_s_f[n * 10 + 4]);
295
296 PRIM_INT__s_s_s_f[n * 10 + 5] = SIMINT_MUL(Q_PB[0], PRIM_INT__s_s_s_d[n * 6 + 5]);
297 PRIM_INT__s_s_s_f[n * 10 + 5] = SIMINT_FMADD( aoq_PQ[0], PRIM_INT__s_s_s_d[(n+1) * 6 + 5], PRIM_INT__s_s_s_f[n * 10 + 5]);
298
299 PRIM_INT__s_s_s_f[n * 10 + 6] = SIMINT_MUL(Q_PB[1], PRIM_INT__s_s_s_d[n * 6 + 3]);
300 PRIM_INT__s_s_s_f[n * 10 + 6] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_s_d[(n+1) * 6 + 3], PRIM_INT__s_s_s_f[n * 10 + 6]);
301 PRIM_INT__s_s_s_f[n * 10 + 6] = SIMINT_FMADD( vrr_const_2_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_p[(n+1) * 3 + 1], PRIM_INT__s_s_s_p[n * 3 + 1]), PRIM_INT__s_s_s_f[n * 10 + 6]);
302
303 PRIM_INT__s_s_s_f[n * 10 + 7] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_d[n * 6 + 3]);
304 PRIM_INT__s_s_s_f[n * 10 + 7] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_d[(n+1) * 6 + 3], PRIM_INT__s_s_s_f[n * 10 + 7]);
305
306 PRIM_INT__s_s_s_f[n * 10 + 8] = SIMINT_MUL(Q_PB[1], PRIM_INT__s_s_s_d[n * 6 + 5]);
307 PRIM_INT__s_s_s_f[n * 10 + 8] = SIMINT_FMADD( aoq_PQ[1], PRIM_INT__s_s_s_d[(n+1) * 6 + 5], PRIM_INT__s_s_s_f[n * 10 + 8]);
308
309 PRIM_INT__s_s_s_f[n * 10 + 9] = SIMINT_MUL(Q_PB[2], PRIM_INT__s_s_s_d[n * 6 + 5]);
310 PRIM_INT__s_s_s_f[n * 10 + 9] = SIMINT_FMADD( aoq_PQ[2], PRIM_INT__s_s_s_d[(n+1) * 6 + 5], PRIM_INT__s_s_s_f[n * 10 + 9]);
311 PRIM_INT__s_s_s_f[n * 10 + 9] = SIMINT_FMADD( vrr_const_2_over_2q, SIMINT_FMADD(a_over_q, PRIM_INT__s_s_s_p[(n+1) * 3 + 2], PRIM_INT__s_s_s_p[n * 3 + 2]), PRIM_INT__s_s_s_f[n * 10 + 9]);
312
313 }
314
315
316 VRR_I_p_s_s_f(
317 PRIM_INT__p_s_s_f,
318 PRIM_INT__s_s_s_f,
319 PRIM_INT__s_s_s_d,
320 P_PA,
321 aop_PQ,
322 one_over_2pq,
323 1);
324
325
326 VRR_L_s_s_s_g(
327 PRIM_INT__s_s_s_g,
328 PRIM_INT__s_s_s_f,
329 PRIM_INT__s_s_s_d,
330 Q_PB,
331 a_over_q,
332 aoq_PQ,
333 one_over_2q,
334 3);
335
336
337 VRR_I_p_s_s_g(
338 PRIM_INT__p_s_s_g,
339 PRIM_INT__s_s_s_g,
340 PRIM_INT__s_s_s_f,
341 P_PA,
342 aop_PQ,
343 one_over_2pq,
344 1);
345
346
347 VRR_L_s_s_s_h(
348 PRIM_INT__s_s_s_h,
349 PRIM_INT__s_s_s_g,
350 PRIM_INT__s_s_s_f,
351 Q_PB,
352 a_over_q,
353 aoq_PQ,
354 one_over_2q,
355 2);
356
357
358 ostei_general_vrr_I(1, 0, 0, 5, 1,
359 one_over_2p, a_over_p, one_over_2pq, aop_PQ, P_PA,
360 PRIM_INT__s_s_s_h, NULL, NULL, NULL, PRIM_INT__s_s_s_g, PRIM_INT__p_s_s_h);
361
362
363
364
365 ////////////////////////////////////
366 // Accumulate contracted integrals
367 ////////////////////////////////////
368 if(lastoffset == 0)
369 {
370 contract_all(30, PRIM_INT__p_s_s_f, PRIM_PTR_INT__p_s_s_f);
371 contract_all(45, PRIM_INT__p_s_s_g, PRIM_PTR_INT__p_s_s_g);
372 contract_all(63, PRIM_INT__p_s_s_h, PRIM_PTR_INT__p_s_s_h);
373 }
374 else
375 {
376 contract(30, shelloffsets, PRIM_INT__p_s_s_f, PRIM_PTR_INT__p_s_s_f);
377 contract(45, shelloffsets, PRIM_INT__p_s_s_g, PRIM_PTR_INT__p_s_s_g);
378 contract(63, shelloffsets, PRIM_INT__p_s_s_h, PRIM_PTR_INT__p_s_s_h);
379 PRIM_PTR_INT__p_s_s_f += lastoffset*30;
380 PRIM_PTR_INT__p_s_s_g += lastoffset*45;
381 PRIM_PTR_INT__p_s_s_h += lastoffset*63;
382 }
383
384 } // close loop over j
385 } // close loop over i
386
387 //Advance to the next batch
388 jstart = SIMINT_SIMD_ROUND(jend);
389
390 //////////////////////////////////////////////
391 // Contracted integrals: Horizontal recurrance
392 //////////////////////////////////////////////
393
394
395
396
397 for(abcd = 0; abcd < nshellbatch; ++abcd, ++real_abcd)
398 {
399 const double hCD[3] = { Q.AB_x[cd+abcd], Q.AB_y[cd+abcd], Q.AB_z[cd+abcd] };
400
401 // set up HRR pointers
402 double const * restrict HRR_INT__p_s_s_f = INT__p_s_s_f + abcd * 30;
403 double const * restrict HRR_INT__p_s_s_g = INT__p_s_s_g + abcd * 45;
404 double const * restrict HRR_INT__p_s_s_h = INT__p_s_s_h + abcd * 63;
405 double * restrict HRR_INT__p_s_d_f = INT__p_s_d_f + real_abcd * 180;
406
407 // form INT__p_s_p_f
408 HRR_K_p_f(
409 HRR_INT__p_s_p_f,
410 HRR_INT__p_s_s_f,
411 HRR_INT__p_s_s_g,
412 hCD, 3);
413
414 // form INT__p_s_p_g
415 HRR_K_p_g(
416 HRR_INT__p_s_p_g,
417 HRR_INT__p_s_s_g,
418 HRR_INT__p_s_s_h,
419 hCD, 3);
420
421 // form INT__p_s_d_f
422 HRR_K_d_f(
423 HRR_INT__p_s_d_f,
424 HRR_INT__p_s_p_f,
425 HRR_INT__p_s_p_g,
426 hCD, 3);
427
428
429 } // close HRR loop
430
431
432 } // close loop cdbatch
433
434 istart = iend;
435 } // close loop over ab
436
437 return P.nshell12_clip * Q.nshell12_clip;
438 }
439
440