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_g_s_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_g_s_s)8 int ostei_h_g_s_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_g_s_s)
13 {
14
15 SIMINT_ASSUME_ALIGN_DBL(work);
16 SIMINT_ASSUME_ALIGN_DBL(INT__h_g_s_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_s_s = work + (SIMINT_NSHELL_SIMD * 0);
29 double * const INT__i_s_s_s = work + (SIMINT_NSHELL_SIMD * 21);
30 double * const INT__k_s_s_s = work + (SIMINT_NSHELL_SIMD * 49);
31 double * const INT__l_s_s_s = work + (SIMINT_NSHELL_SIMD * 85);
32 double * const INT__m_s_s_s = work + (SIMINT_NSHELL_SIMD * 130);
33 SIMINT_DBLTYPE * const primwork = (SIMINT_DBLTYPE *)(work + SIMINT_NSHELL_SIMD*185);
34 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_s = primwork + 0;
35 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_s_s = primwork + 10;
36 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_s_s = primwork + 37;
37 SIMINT_DBLTYPE * const restrict PRIM_INT__f_s_s_s = primwork + 85;
38 SIMINT_DBLTYPE * const restrict PRIM_INT__g_s_s_s = primwork + 155;
39 SIMINT_DBLTYPE * const restrict PRIM_INT__h_s_s_s = primwork + 245;
40 SIMINT_DBLTYPE * const restrict PRIM_INT__i_s_s_s = primwork + 350;
41 SIMINT_DBLTYPE * const restrict PRIM_INT__k_s_s_s = primwork + 462;
42 SIMINT_DBLTYPE * const restrict PRIM_INT__l_s_s_s = primwork + 570;
43 SIMINT_DBLTYPE * const restrict PRIM_INT__m_s_s_s = primwork + 660;
44 double * const hrrwork = (double *)(primwork + 715);
45 double * const HRR_INT__h_p_s_s = hrrwork + 0;
46 double * const HRR_INT__h_d_s_s = hrrwork + 63;
47 double * const HRR_INT__h_f_s_s = hrrwork + 189;
48 double * const HRR_INT__i_p_s_s = hrrwork + 399;
49 double * const HRR_INT__i_d_s_s = hrrwork + 483;
50 double * const HRR_INT__i_f_s_s = hrrwork + 651;
51 double * const HRR_INT__k_p_s_s = hrrwork + 931;
52 double * const HRR_INT__k_d_s_s = hrrwork + 1039;
53 double * const HRR_INT__l_p_s_s = hrrwork + 1255;
54
55
56 // Create constants
57 const SIMINT_DBLTYPE const_1 = SIMINT_DBLSET1(1);
58 const SIMINT_DBLTYPE const_2 = SIMINT_DBLSET1(2);
59 const SIMINT_DBLTYPE const_3 = SIMINT_DBLSET1(3);
60 const SIMINT_DBLTYPE const_4 = SIMINT_DBLSET1(4);
61 const SIMINT_DBLTYPE const_5 = SIMINT_DBLSET1(5);
62 const SIMINT_DBLTYPE const_6 = SIMINT_DBLSET1(6);
63 const SIMINT_DBLTYPE const_7 = SIMINT_DBLSET1(7);
64 const SIMINT_DBLTYPE const_8 = SIMINT_DBLSET1(8);
65 const SIMINT_DBLTYPE one_half = SIMINT_DBLSET1(0.5);
66
67
68 ////////////////////////////////////////
69 // Loop over shells and primitives
70 ////////////////////////////////////////
71
72 real_abcd = 0;
73 istart = 0;
74 for(ab = 0; ab < P.nshell12_clip; ++ab)
75 {
76 const int iend = istart + P.nprim12[ab];
77
78 cd = 0;
79 jstart = 0;
80
81 for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD)
82 {
83 const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD;
84 int jend = jstart;
85 for(i = 0; i < nshellbatch; i++)
86 jend += Q.nprim12[cd+i];
87
88 // Clear the beginning of the workspace (where we are accumulating integrals)
89 memset(work, 0, SIMINT_NSHELL_SIMD * 185 * sizeof(double));
90 abcd = 0;
91
92
93 for(i = istart; i < iend; ++i)
94 {
95 SIMINT_DBLTYPE bra_screen_max; // only used if check_screen
96
97 if(check_screen)
98 {
99 // Skip this whole thing if always insignificant
100 if((P.screen[i] * Q.screen_max) < screen_tol)
101 continue;
102 bra_screen_max = SIMINT_DBLSET1(P.screen[i]);
103 }
104
105 icd = 0;
106 iprimcd = 0;
107 nprim_icd = Q.nprim12[cd];
108 double * restrict PRIM_PTR_INT__h_s_s_s = INT__h_s_s_s + abcd * 21;
109 double * restrict PRIM_PTR_INT__i_s_s_s = INT__i_s_s_s + abcd * 28;
110 double * restrict PRIM_PTR_INT__k_s_s_s = INT__k_s_s_s + abcd * 36;
111 double * restrict PRIM_PTR_INT__l_s_s_s = INT__l_s_s_s + abcd * 45;
112 double * restrict PRIM_PTR_INT__m_s_s_s = INT__m_s_s_s + abcd * 55;
113
114
115
116 // Load these one per loop over i
117 const SIMINT_DBLTYPE P_alpha = SIMINT_DBLSET1(P.alpha[i]);
118 const SIMINT_DBLTYPE P_prefac = SIMINT_DBLSET1(P.prefac[i]);
119 const SIMINT_DBLTYPE Pxyz[3] = { SIMINT_DBLSET1(P.x[i]), SIMINT_DBLSET1(P.y[i]), SIMINT_DBLSET1(P.z[i]) };
120
121 const SIMINT_DBLTYPE P_PA[3] = { SIMINT_DBLSET1(P.PA_x[i]), SIMINT_DBLSET1(P.PA_y[i]), SIMINT_DBLSET1(P.PA_z[i]) };
122
123 for(j = jstart; j < jend; j += SIMINT_SIMD_LEN)
124 {
125 // calculate the shell offsets
126 // these are the offset from the shell pointed to by cd
127 // for each element
128 int shelloffsets[SIMINT_SIMD_LEN] = {0};
129 int lastoffset = 0;
130 const int nlane = ( ((j + SIMINT_SIMD_LEN) < jend) ? SIMINT_SIMD_LEN : (jend - j));
131
132 if((iprimcd + SIMINT_SIMD_LEN) >= nprim_icd)
133 {
134 // Handle if the first element of the vector is a new shell
135 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
136 {
137 nprim_icd += Q.nprim12[cd + (++icd)];
138 PRIM_PTR_INT__h_s_s_s += 21;
139 PRIM_PTR_INT__i_s_s_s += 28;
140 PRIM_PTR_INT__k_s_s_s += 36;
141 PRIM_PTR_INT__l_s_s_s += 45;
142 PRIM_PTR_INT__m_s_s_s += 55;
143 }
144 iprimcd++;
145 for(n = 1; n < SIMINT_SIMD_LEN; ++n)
146 {
147 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
148 {
149 shelloffsets[n] = shelloffsets[n-1] + 1;
150 lastoffset++;
151 nprim_icd += Q.nprim12[cd + (++icd)];
152 }
153 else
154 shelloffsets[n] = shelloffsets[n-1];
155 iprimcd++;
156 }
157 }
158 else
159 iprimcd += SIMINT_SIMD_LEN;
160
161 // Do we have to compute this vector (or has it been screened out)?
162 // (not_screened != 0 means we have to do this vector)
163 if(check_screen)
164 {
165 const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j)));
166 if(vmax < screen_tol)
167 {
168 PRIM_PTR_INT__h_s_s_s += lastoffset*21;
169 PRIM_PTR_INT__i_s_s_s += lastoffset*28;
170 PRIM_PTR_INT__k_s_s_s += lastoffset*36;
171 PRIM_PTR_INT__l_s_s_s += lastoffset*45;
172 PRIM_PTR_INT__m_s_s_s += lastoffset*55;
173 continue;
174 }
175 }
176
177 const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j);
178 const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha);
179 const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha);
180 const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum);
181
182
183 /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */
184 SIMINT_DBLTYPE PQ[3];
185 PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j));
186 PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j));
187 PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j));
188 SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]);
189 R2 = SIMINT_FMADD(PQ[1], PQ[1], R2);
190 R2 = SIMINT_FMADD(PQ[2], PQ[2], R2);
191
192 const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST
193 const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha);
194 const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha);
195 const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p);
196 const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q);
197 const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum);
198
199 // NOTE: Minus sign!
200 const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p);
201 SIMINT_DBLTYPE aop_PQ[3];
202 aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]);
203 aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]);
204 aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]);
205
206
207 //////////////////////////////////////////////
208 // Fjt function section
209 // Maximum v value: 9
210 //////////////////////////////////////////////
211 // The parameter to the Fjt function
212 const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha);
213
214
215 const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j);
216
217
218 boys_F_split(PRIM_INT__s_s_s_s, F_x, 9);
219 SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum);
220 prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac);
221 for(n = 0; n <= 9; n++)
222 PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac);
223
224 //////////////////////////////////////////////
225 // Primitive integrals: Vertical recurrance
226 //////////////////////////////////////////////
227
228 const SIMINT_DBLTYPE vrr_const_1_over_2p = one_over_2p;
229 const SIMINT_DBLTYPE vrr_const_2_over_2p = SIMINT_MUL(const_2, one_over_2p);
230 const SIMINT_DBLTYPE vrr_const_3_over_2p = SIMINT_MUL(const_3, one_over_2p);
231 const SIMINT_DBLTYPE vrr_const_4_over_2p = SIMINT_MUL(const_4, one_over_2p);
232 const SIMINT_DBLTYPE vrr_const_5_over_2p = SIMINT_MUL(const_5, one_over_2p);
233 const SIMINT_DBLTYPE vrr_const_6_over_2p = SIMINT_MUL(const_6, one_over_2p);
234 const SIMINT_DBLTYPE vrr_const_7_over_2p = SIMINT_MUL(const_7, one_over_2p);
235 const SIMINT_DBLTYPE vrr_const_8_over_2p = SIMINT_MUL(const_8, one_over_2p);
236
237
238
239 // Forming PRIM_INT__p_s_s_s[9 * 3];
240 for(n = 0; n < 9; ++n) // loop over orders of auxiliary function
241 {
242
243 PRIM_INT__p_s_s_s[n * 3 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_s_s[n * 1 + 0]);
244 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]);
245
246 PRIM_INT__p_s_s_s[n * 3 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_s_s[n * 1 + 0]);
247 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]);
248
249 PRIM_INT__p_s_s_s[n * 3 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_s_s[n * 1 + 0]);
250 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]);
251
252 }
253
254
255
256 // Forming PRIM_INT__d_s_s_s[8 * 6];
257 for(n = 0; n < 8; ++n) // loop over orders of auxiliary function
258 {
259
260 PRIM_INT__d_s_s_s[n * 6 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__p_s_s_s[n * 3 + 0]);
261 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]);
262 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]);
263
264 PRIM_INT__d_s_s_s[n * 6 + 3] = SIMINT_MUL(P_PA[1], PRIM_INT__p_s_s_s[n * 3 + 1]);
265 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]);
266 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]);
267
268 PRIM_INT__d_s_s_s[n * 6 + 5] = SIMINT_MUL(P_PA[2], PRIM_INT__p_s_s_s[n * 3 + 2]);
269 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]);
270 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]);
271
272 }
273
274
275
276 // Forming PRIM_INT__f_s_s_s[7 * 10];
277 for(n = 0; n < 7; ++n) // loop over orders of auxiliary function
278 {
279
280 PRIM_INT__f_s_s_s[n * 10 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__d_s_s_s[n * 6 + 0]);
281 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]);
282 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]);
283
284 PRIM_INT__f_s_s_s[n * 10 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 0]);
285 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]);
286
287 PRIM_INT__f_s_s_s[n * 10 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 0]);
288 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]);
289
290 PRIM_INT__f_s_s_s[n * 10 + 6] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 3]);
291 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]);
292 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]);
293
294 PRIM_INT__f_s_s_s[n * 10 + 7] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 3]);
295 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]);
296
297 PRIM_INT__f_s_s_s[n * 10 + 9] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 5]);
298 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]);
299 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]);
300
301 }
302
303
304 VRR_I_g_s_s_s(
305 PRIM_INT__g_s_s_s,
306 PRIM_INT__f_s_s_s,
307 PRIM_INT__d_s_s_s,
308 P_PA,
309 a_over_p,
310 aop_PQ,
311 one_over_2p,
312 6);
313
314
315 VRR_I_h_s_s_s(
316 PRIM_INT__h_s_s_s,
317 PRIM_INT__g_s_s_s,
318 PRIM_INT__f_s_s_s,
319 P_PA,
320 a_over_p,
321 aop_PQ,
322 one_over_2p,
323 5);
324
325
326 ostei_general_vrr1_I(6, 4,
327 one_over_2p, a_over_p, aop_PQ, P_PA,
328 PRIM_INT__h_s_s_s, PRIM_INT__g_s_s_s, PRIM_INT__i_s_s_s);
329
330
331 ostei_general_vrr1_I(7, 3,
332 one_over_2p, a_over_p, aop_PQ, P_PA,
333 PRIM_INT__i_s_s_s, PRIM_INT__h_s_s_s, PRIM_INT__k_s_s_s);
334
335
336 ostei_general_vrr1_I(8, 2,
337 one_over_2p, a_over_p, aop_PQ, P_PA,
338 PRIM_INT__k_s_s_s, PRIM_INT__i_s_s_s, PRIM_INT__l_s_s_s);
339
340
341 ostei_general_vrr1_I(9, 1,
342 one_over_2p, a_over_p, aop_PQ, P_PA,
343 PRIM_INT__l_s_s_s, PRIM_INT__k_s_s_s, PRIM_INT__m_s_s_s);
344
345
346
347
348 ////////////////////////////////////
349 // Accumulate contracted integrals
350 ////////////////////////////////////
351 if(lastoffset == 0)
352 {
353 contract_all(21, PRIM_INT__h_s_s_s, PRIM_PTR_INT__h_s_s_s);
354 contract_all(28, PRIM_INT__i_s_s_s, PRIM_PTR_INT__i_s_s_s);
355 contract_all(36, PRIM_INT__k_s_s_s, PRIM_PTR_INT__k_s_s_s);
356 contract_all(45, PRIM_INT__l_s_s_s, PRIM_PTR_INT__l_s_s_s);
357 contract_all(55, PRIM_INT__m_s_s_s, PRIM_PTR_INT__m_s_s_s);
358 }
359 else
360 {
361 contract(21, shelloffsets, PRIM_INT__h_s_s_s, PRIM_PTR_INT__h_s_s_s);
362 contract(28, shelloffsets, PRIM_INT__i_s_s_s, PRIM_PTR_INT__i_s_s_s);
363 contract(36, shelloffsets, PRIM_INT__k_s_s_s, PRIM_PTR_INT__k_s_s_s);
364 contract(45, shelloffsets, PRIM_INT__l_s_s_s, PRIM_PTR_INT__l_s_s_s);
365 contract(55, shelloffsets, PRIM_INT__m_s_s_s, PRIM_PTR_INT__m_s_s_s);
366 PRIM_PTR_INT__h_s_s_s += lastoffset*21;
367 PRIM_PTR_INT__i_s_s_s += lastoffset*28;
368 PRIM_PTR_INT__k_s_s_s += lastoffset*36;
369 PRIM_PTR_INT__l_s_s_s += lastoffset*45;
370 PRIM_PTR_INT__m_s_s_s += lastoffset*55;
371 }
372
373 } // close loop over j
374 } // close loop over i
375
376 //Advance to the next batch
377 jstart = SIMINT_SIMD_ROUND(jend);
378
379 //////////////////////////////////////////////
380 // Contracted integrals: Horizontal recurrance
381 //////////////////////////////////////////////
382
383
384 const double hAB[3] = { P.AB_x[ab], P.AB_y[ab], P.AB_z[ab] };
385
386
387 for(abcd = 0; abcd < nshellbatch; ++abcd, ++real_abcd)
388 {
389
390 // set up HRR pointers
391 double const * restrict HRR_INT__h_s_s_s = INT__h_s_s_s + abcd * 21;
392 double const * restrict HRR_INT__i_s_s_s = INT__i_s_s_s + abcd * 28;
393 double const * restrict HRR_INT__k_s_s_s = INT__k_s_s_s + abcd * 36;
394 double const * restrict HRR_INT__l_s_s_s = INT__l_s_s_s + abcd * 45;
395 double const * restrict HRR_INT__m_s_s_s = INT__m_s_s_s + abcd * 55;
396 double * restrict HRR_INT__h_g_s_s = INT__h_g_s_s + real_abcd * 315;
397
398 // form INT__h_p_s_s
399 ostei_general_hrr_J(5, 1, 0, 0, hAB, HRR_INT__i_s_s_s, HRR_INT__h_s_s_s, HRR_INT__h_p_s_s);
400
401 // form INT__i_p_s_s
402 ostei_general_hrr_J(6, 1, 0, 0, hAB, HRR_INT__k_s_s_s, HRR_INT__i_s_s_s, HRR_INT__i_p_s_s);
403
404 // form INT__k_p_s_s
405 ostei_general_hrr_J(7, 1, 0, 0, hAB, HRR_INT__l_s_s_s, HRR_INT__k_s_s_s, HRR_INT__k_p_s_s);
406
407 // form INT__l_p_s_s
408 ostei_general_hrr_J(8, 1, 0, 0, hAB, HRR_INT__m_s_s_s, HRR_INT__l_s_s_s, HRR_INT__l_p_s_s);
409
410 // form INT__h_d_s_s
411 ostei_general_hrr_J(5, 2, 0, 0, hAB, HRR_INT__i_p_s_s, HRR_INT__h_p_s_s, HRR_INT__h_d_s_s);
412
413 // form INT__i_d_s_s
414 ostei_general_hrr_J(6, 2, 0, 0, hAB, HRR_INT__k_p_s_s, HRR_INT__i_p_s_s, HRR_INT__i_d_s_s);
415
416 // form INT__k_d_s_s
417 ostei_general_hrr_J(7, 2, 0, 0, hAB, HRR_INT__l_p_s_s, HRR_INT__k_p_s_s, HRR_INT__k_d_s_s);
418
419 // form INT__h_f_s_s
420 ostei_general_hrr_J(5, 3, 0, 0, hAB, HRR_INT__i_d_s_s, HRR_INT__h_d_s_s, HRR_INT__h_f_s_s);
421
422 // form INT__i_f_s_s
423 ostei_general_hrr_J(6, 3, 0, 0, hAB, HRR_INT__k_d_s_s, HRR_INT__i_d_s_s, HRR_INT__i_f_s_s);
424
425 // form INT__h_g_s_s
426 ostei_general_hrr_J(5, 4, 0, 0, hAB, HRR_INT__i_f_s_s, HRR_INT__h_f_s_s, HRR_INT__h_g_s_s);
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
ostei_g_h_s_s(struct simint_multi_shellpair const P,struct simint_multi_shellpair const Q,double screen_tol,double * const restrict work,double * const restrict INT__g_h_s_s)440 int ostei_g_h_s_s(struct simint_multi_shellpair const P,
441 struct simint_multi_shellpair const Q,
442 double screen_tol,
443 double * const restrict work,
444 double * const restrict INT__g_h_s_s)
445 {
446 double P_AB[3*P.nshell12];
447 struct simint_multi_shellpair P_tmp = P;
448 P_tmp.PA_x = P.PB_x; P_tmp.PA_y = P.PB_y; P_tmp.PA_z = P.PB_z;
449 P_tmp.PB_x = P.PA_x; P_tmp.PB_y = P.PA_y; P_tmp.PB_z = P.PA_z;
450 P_tmp.AB_x = P_AB;
451 P_tmp.AB_y = P_AB + P.nshell12;
452 P_tmp.AB_z = P_AB + 2*P.nshell12;
453
454 for(int i = 0; i < P.nshell12; i++)
455 {
456 P_tmp.AB_x[i] = -P.AB_x[i];
457 P_tmp.AB_y[i] = -P.AB_y[i];
458 P_tmp.AB_z[i] = -P.AB_z[i];
459 }
460
461 int ret = ostei_h_g_s_s(P_tmp, Q, screen_tol, work, INT__g_h_s_s);
462 double buffer[315] SIMINT_ALIGN_ARRAY_DBL;
463
464 for(int q = 0; q < ret; q++)
465 {
466 int idx = 0;
467 for(int a = 0; a < 15; ++a)
468 for(int b = 0; b < 21; ++b)
469 for(int c = 0; c < 1; ++c)
470 for(int d = 0; d < 1; ++d)
471 buffer[idx++] = INT__g_h_s_s[q*315+b*15+a*1+c*1+d];
472
473 memcpy(INT__g_h_s_s+q*315, buffer, 315*sizeof(double));
474 }
475
476 return ret;
477 }
478
479