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_i_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__i_h_s_s)8 int ostei_i_h_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__i_h_s_s)
13 {
14
15 SIMINT_ASSUME_ALIGN_DBL(work);
16 SIMINT_ASSUME_ALIGN_DBL(INT__i_h_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__i_s_s_s = work + (SIMINT_NSHELL_SIMD * 0);
29 double * const INT__k_s_s_s = work + (SIMINT_NSHELL_SIMD * 28);
30 double * const INT__l_s_s_s = work + (SIMINT_NSHELL_SIMD * 64);
31 double * const INT__m_s_s_s = work + (SIMINT_NSHELL_SIMD * 109);
32 double * const INT__n_s_s_s = work + (SIMINT_NSHELL_SIMD * 164);
33 double * const INT__o_s_s_s = work + (SIMINT_NSHELL_SIMD * 230);
34 SIMINT_DBLTYPE * const primwork = (SIMINT_DBLTYPE *)(work + SIMINT_NSHELL_SIMD*308);
35 SIMINT_DBLTYPE * const restrict PRIM_INT__s_s_s_s = primwork + 0;
36 SIMINT_DBLTYPE * const restrict PRIM_INT__p_s_s_s = primwork + 12;
37 SIMINT_DBLTYPE * const restrict PRIM_INT__d_s_s_s = primwork + 45;
38 SIMINT_DBLTYPE * const restrict PRIM_INT__f_s_s_s = primwork + 105;
39 SIMINT_DBLTYPE * const restrict PRIM_INT__g_s_s_s = primwork + 195;
40 SIMINT_DBLTYPE * const restrict PRIM_INT__h_s_s_s = primwork + 315;
41 SIMINT_DBLTYPE * const restrict PRIM_INT__i_s_s_s = primwork + 462;
42 SIMINT_DBLTYPE * const restrict PRIM_INT__k_s_s_s = primwork + 630;
43 SIMINT_DBLTYPE * const restrict PRIM_INT__l_s_s_s = primwork + 810;
44 SIMINT_DBLTYPE * const restrict PRIM_INT__m_s_s_s = primwork + 990;
45 SIMINT_DBLTYPE * const restrict PRIM_INT__n_s_s_s = primwork + 1155;
46 SIMINT_DBLTYPE * const restrict PRIM_INT__o_s_s_s = primwork + 1287;
47 double * const hrrwork = (double *)(primwork + 1365);
48 double * const HRR_INT__i_p_s_s = hrrwork + 0;
49 double * const HRR_INT__i_d_s_s = hrrwork + 84;
50 double * const HRR_INT__i_f_s_s = hrrwork + 252;
51 double * const HRR_INT__i_g_s_s = hrrwork + 532;
52 double * const HRR_INT__k_p_s_s = hrrwork + 952;
53 double * const HRR_INT__k_d_s_s = hrrwork + 1060;
54 double * const HRR_INT__k_f_s_s = hrrwork + 1276;
55 double * const HRR_INT__k_g_s_s = hrrwork + 1636;
56 double * const HRR_INT__l_p_s_s = hrrwork + 2176;
57 double * const HRR_INT__l_d_s_s = hrrwork + 2311;
58 double * const HRR_INT__l_f_s_s = hrrwork + 2581;
59 double * const HRR_INT__m_p_s_s = hrrwork + 3031;
60 double * const HRR_INT__m_d_s_s = hrrwork + 3196;
61 double * const HRR_INT__n_p_s_s = hrrwork + 3526;
62
63
64 // Create constants
65 const SIMINT_DBLTYPE const_1 = SIMINT_DBLSET1(1);
66 const SIMINT_DBLTYPE const_10 = SIMINT_DBLSET1(10);
67 const SIMINT_DBLTYPE const_2 = SIMINT_DBLSET1(2);
68 const SIMINT_DBLTYPE const_3 = SIMINT_DBLSET1(3);
69 const SIMINT_DBLTYPE const_4 = SIMINT_DBLSET1(4);
70 const SIMINT_DBLTYPE const_5 = SIMINT_DBLSET1(5);
71 const SIMINT_DBLTYPE const_6 = SIMINT_DBLSET1(6);
72 const SIMINT_DBLTYPE const_7 = SIMINT_DBLSET1(7);
73 const SIMINT_DBLTYPE const_8 = SIMINT_DBLSET1(8);
74 const SIMINT_DBLTYPE const_9 = SIMINT_DBLSET1(9);
75 const SIMINT_DBLTYPE one_half = SIMINT_DBLSET1(0.5);
76
77
78 ////////////////////////////////////////
79 // Loop over shells and primitives
80 ////////////////////////////////////////
81
82 real_abcd = 0;
83 istart = 0;
84 for(ab = 0; ab < P.nshell12_clip; ++ab)
85 {
86 const int iend = istart + P.nprim12[ab];
87
88 cd = 0;
89 jstart = 0;
90
91 for(cd = 0; cd < Q.nshell12_clip; cd += SIMINT_NSHELL_SIMD)
92 {
93 const int nshellbatch = ((cd + SIMINT_NSHELL_SIMD) > Q.nshell12_clip) ? Q.nshell12_clip - cd : SIMINT_NSHELL_SIMD;
94 int jend = jstart;
95 for(i = 0; i < nshellbatch; i++)
96 jend += Q.nprim12[cd+i];
97
98 // Clear the beginning of the workspace (where we are accumulating integrals)
99 memset(work, 0, SIMINT_NSHELL_SIMD * 308 * sizeof(double));
100 abcd = 0;
101
102
103 for(i = istart; i < iend; ++i)
104 {
105 SIMINT_DBLTYPE bra_screen_max; // only used if check_screen
106
107 if(check_screen)
108 {
109 // Skip this whole thing if always insignificant
110 if((P.screen[i] * Q.screen_max) < screen_tol)
111 continue;
112 bra_screen_max = SIMINT_DBLSET1(P.screen[i]);
113 }
114
115 icd = 0;
116 iprimcd = 0;
117 nprim_icd = Q.nprim12[cd];
118 double * restrict PRIM_PTR_INT__i_s_s_s = INT__i_s_s_s + abcd * 28;
119 double * restrict PRIM_PTR_INT__k_s_s_s = INT__k_s_s_s + abcd * 36;
120 double * restrict PRIM_PTR_INT__l_s_s_s = INT__l_s_s_s + abcd * 45;
121 double * restrict PRIM_PTR_INT__m_s_s_s = INT__m_s_s_s + abcd * 55;
122 double * restrict PRIM_PTR_INT__n_s_s_s = INT__n_s_s_s + abcd * 66;
123 double * restrict PRIM_PTR_INT__o_s_s_s = INT__o_s_s_s + abcd * 78;
124
125
126
127 // Load these one per loop over i
128 const SIMINT_DBLTYPE P_alpha = SIMINT_DBLSET1(P.alpha[i]);
129 const SIMINT_DBLTYPE P_prefac = SIMINT_DBLSET1(P.prefac[i]);
130 const SIMINT_DBLTYPE Pxyz[3] = { SIMINT_DBLSET1(P.x[i]), SIMINT_DBLSET1(P.y[i]), SIMINT_DBLSET1(P.z[i]) };
131
132 const SIMINT_DBLTYPE P_PA[3] = { SIMINT_DBLSET1(P.PA_x[i]), SIMINT_DBLSET1(P.PA_y[i]), SIMINT_DBLSET1(P.PA_z[i]) };
133
134 for(j = jstart; j < jend; j += SIMINT_SIMD_LEN)
135 {
136 // calculate the shell offsets
137 // these are the offset from the shell pointed to by cd
138 // for each element
139 int shelloffsets[SIMINT_SIMD_LEN] = {0};
140 int lastoffset = 0;
141 const int nlane = ( ((j + SIMINT_SIMD_LEN) < jend) ? SIMINT_SIMD_LEN : (jend - j));
142
143 if((iprimcd + SIMINT_SIMD_LEN) >= nprim_icd)
144 {
145 // Handle if the first element of the vector is a new shell
146 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
147 {
148 nprim_icd += Q.nprim12[cd + (++icd)];
149 PRIM_PTR_INT__i_s_s_s += 28;
150 PRIM_PTR_INT__k_s_s_s += 36;
151 PRIM_PTR_INT__l_s_s_s += 45;
152 PRIM_PTR_INT__m_s_s_s += 55;
153 PRIM_PTR_INT__n_s_s_s += 66;
154 PRIM_PTR_INT__o_s_s_s += 78;
155 }
156 iprimcd++;
157 for(n = 1; n < SIMINT_SIMD_LEN; ++n)
158 {
159 if(iprimcd >= nprim_icd && ((icd+1) < nshellbatch))
160 {
161 shelloffsets[n] = shelloffsets[n-1] + 1;
162 lastoffset++;
163 nprim_icd += Q.nprim12[cd + (++icd)];
164 }
165 else
166 shelloffsets[n] = shelloffsets[n-1];
167 iprimcd++;
168 }
169 }
170 else
171 iprimcd += SIMINT_SIMD_LEN;
172
173 // Do we have to compute this vector (or has it been screened out)?
174 // (not_screened != 0 means we have to do this vector)
175 if(check_screen)
176 {
177 const double vmax = vector_max(SIMINT_MUL(bra_screen_max, SIMINT_DBLLOAD(Q.screen, j)));
178 if(vmax < screen_tol)
179 {
180 PRIM_PTR_INT__i_s_s_s += lastoffset*28;
181 PRIM_PTR_INT__k_s_s_s += lastoffset*36;
182 PRIM_PTR_INT__l_s_s_s += lastoffset*45;
183 PRIM_PTR_INT__m_s_s_s += lastoffset*55;
184 PRIM_PTR_INT__n_s_s_s += lastoffset*66;
185 PRIM_PTR_INT__o_s_s_s += lastoffset*78;
186 continue;
187 }
188 }
189
190 const SIMINT_DBLTYPE Q_alpha = SIMINT_DBLLOAD(Q.alpha, j);
191 const SIMINT_DBLTYPE PQalpha_mul = SIMINT_MUL(P_alpha, Q_alpha);
192 const SIMINT_DBLTYPE PQalpha_sum = SIMINT_ADD(P_alpha, Q_alpha);
193 const SIMINT_DBLTYPE one_over_PQalpha_sum = SIMINT_DIV(const_1, PQalpha_sum);
194
195
196 /* construct R2 = (Px - Qx)**2 + (Py - Qy)**2 + (Pz -Qz)**2 */
197 SIMINT_DBLTYPE PQ[3];
198 PQ[0] = SIMINT_SUB(Pxyz[0], SIMINT_DBLLOAD(Q.x, j));
199 PQ[1] = SIMINT_SUB(Pxyz[1], SIMINT_DBLLOAD(Q.y, j));
200 PQ[2] = SIMINT_SUB(Pxyz[2], SIMINT_DBLLOAD(Q.z, j));
201 SIMINT_DBLTYPE R2 = SIMINT_MUL(PQ[0], PQ[0]);
202 R2 = SIMINT_FMADD(PQ[1], PQ[1], R2);
203 R2 = SIMINT_FMADD(PQ[2], PQ[2], R2);
204
205 const SIMINT_DBLTYPE alpha = SIMINT_MUL(PQalpha_mul, one_over_PQalpha_sum); // alpha from MEST
206 const SIMINT_DBLTYPE one_over_p = SIMINT_DIV(const_1, P_alpha);
207 const SIMINT_DBLTYPE one_over_q = SIMINT_DIV(const_1, Q_alpha);
208 const SIMINT_DBLTYPE one_over_2p = SIMINT_MUL(one_half, one_over_p);
209 const SIMINT_DBLTYPE one_over_2q = SIMINT_MUL(one_half, one_over_q);
210 const SIMINT_DBLTYPE one_over_2pq = SIMINT_MUL(one_half, one_over_PQalpha_sum);
211
212 // NOTE: Minus sign!
213 const SIMINT_DBLTYPE a_over_p = SIMINT_MUL(SIMINT_NEG(alpha), one_over_p);
214 SIMINT_DBLTYPE aop_PQ[3];
215 aop_PQ[0] = SIMINT_MUL(a_over_p, PQ[0]);
216 aop_PQ[1] = SIMINT_MUL(a_over_p, PQ[1]);
217 aop_PQ[2] = SIMINT_MUL(a_over_p, PQ[2]);
218
219
220 //////////////////////////////////////////////
221 // Fjt function section
222 // Maximum v value: 11
223 //////////////////////////////////////////////
224 // The parameter to the Fjt function
225 const SIMINT_DBLTYPE F_x = SIMINT_MUL(R2, alpha);
226
227
228 const SIMINT_DBLTYPE Q_prefac = mask_load(nlane, Q.prefac + j);
229
230
231 boys_F_split(PRIM_INT__s_s_s_s, F_x, 11);
232 SIMINT_DBLTYPE prefac = SIMINT_SQRT(one_over_PQalpha_sum);
233 prefac = SIMINT_MUL(SIMINT_MUL(P_prefac, Q_prefac), prefac);
234 for(n = 0; n <= 11; n++)
235 PRIM_INT__s_s_s_s[n] = SIMINT_MUL(PRIM_INT__s_s_s_s[n], prefac);
236
237 //////////////////////////////////////////////
238 // Primitive integrals: Vertical recurrance
239 //////////////////////////////////////////////
240
241 const SIMINT_DBLTYPE vrr_const_1_over_2p = one_over_2p;
242 const SIMINT_DBLTYPE vrr_const_2_over_2p = SIMINT_MUL(const_2, one_over_2p);
243 const SIMINT_DBLTYPE vrr_const_3_over_2p = SIMINT_MUL(const_3, one_over_2p);
244 const SIMINT_DBLTYPE vrr_const_4_over_2p = SIMINT_MUL(const_4, one_over_2p);
245 const SIMINT_DBLTYPE vrr_const_5_over_2p = SIMINT_MUL(const_5, one_over_2p);
246 const SIMINT_DBLTYPE vrr_const_6_over_2p = SIMINT_MUL(const_6, one_over_2p);
247 const SIMINT_DBLTYPE vrr_const_7_over_2p = SIMINT_MUL(const_7, one_over_2p);
248 const SIMINT_DBLTYPE vrr_const_8_over_2p = SIMINT_MUL(const_8, one_over_2p);
249 const SIMINT_DBLTYPE vrr_const_9_over_2p = SIMINT_MUL(const_9, one_over_2p);
250 const SIMINT_DBLTYPE vrr_const_10_over_2p = SIMINT_MUL(const_10, one_over_2p);
251
252
253
254 // Forming PRIM_INT__p_s_s_s[11 * 3];
255 for(n = 0; n < 11; ++n) // loop over orders of auxiliary function
256 {
257
258 PRIM_INT__p_s_s_s[n * 3 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__s_s_s_s[n * 1 + 0]);
259 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]);
260
261 PRIM_INT__p_s_s_s[n * 3 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__s_s_s_s[n * 1 + 0]);
262 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]);
263
264 PRIM_INT__p_s_s_s[n * 3 + 2] = SIMINT_MUL(P_PA[2], PRIM_INT__s_s_s_s[n * 1 + 0]);
265 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]);
266
267 }
268
269
270
271 // Forming PRIM_INT__d_s_s_s[10 * 6];
272 for(n = 0; n < 10; ++n) // loop over orders of auxiliary function
273 {
274
275 PRIM_INT__d_s_s_s[n * 6 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__p_s_s_s[n * 3 + 0]);
276 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]);
277 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]);
278
279 PRIM_INT__d_s_s_s[n * 6 + 3] = SIMINT_MUL(P_PA[1], PRIM_INT__p_s_s_s[n * 3 + 1]);
280 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]);
281 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]);
282
283 PRIM_INT__d_s_s_s[n * 6 + 5] = SIMINT_MUL(P_PA[2], PRIM_INT__p_s_s_s[n * 3 + 2]);
284 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]);
285 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]);
286
287 }
288
289
290
291 // Forming PRIM_INT__f_s_s_s[9 * 10];
292 for(n = 0; n < 9; ++n) // loop over orders of auxiliary function
293 {
294
295 PRIM_INT__f_s_s_s[n * 10 + 0] = SIMINT_MUL(P_PA[0], PRIM_INT__d_s_s_s[n * 6 + 0]);
296 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]);
297 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]);
298
299 PRIM_INT__f_s_s_s[n * 10 + 1] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 0]);
300 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]);
301
302 PRIM_INT__f_s_s_s[n * 10 + 6] = SIMINT_MUL(P_PA[1], PRIM_INT__d_s_s_s[n * 6 + 3]);
303 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]);
304 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]);
305
306 PRIM_INT__f_s_s_s[n * 10 + 9] = SIMINT_MUL(P_PA[2], PRIM_INT__d_s_s_s[n * 6 + 5]);
307 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]);
308 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]);
309
310 }
311
312
313 VRR_I_g_s_s_s(
314 PRIM_INT__g_s_s_s,
315 PRIM_INT__f_s_s_s,
316 PRIM_INT__d_s_s_s,
317 P_PA,
318 a_over_p,
319 aop_PQ,
320 one_over_2p,
321 8);
322
323
324 VRR_I_h_s_s_s(
325 PRIM_INT__h_s_s_s,
326 PRIM_INT__g_s_s_s,
327 PRIM_INT__f_s_s_s,
328 P_PA,
329 a_over_p,
330 aop_PQ,
331 one_over_2p,
332 7);
333
334
335 ostei_general_vrr1_I(6, 6,
336 one_over_2p, a_over_p, aop_PQ, P_PA,
337 PRIM_INT__h_s_s_s, PRIM_INT__g_s_s_s, PRIM_INT__i_s_s_s);
338
339
340 ostei_general_vrr1_I(7, 5,
341 one_over_2p, a_over_p, aop_PQ, P_PA,
342 PRIM_INT__i_s_s_s, PRIM_INT__h_s_s_s, PRIM_INT__k_s_s_s);
343
344
345 ostei_general_vrr1_I(8, 4,
346 one_over_2p, a_over_p, aop_PQ, P_PA,
347 PRIM_INT__k_s_s_s, PRIM_INT__i_s_s_s, PRIM_INT__l_s_s_s);
348
349
350 ostei_general_vrr1_I(9, 3,
351 one_over_2p, a_over_p, aop_PQ, P_PA,
352 PRIM_INT__l_s_s_s, PRIM_INT__k_s_s_s, PRIM_INT__m_s_s_s);
353
354
355 ostei_general_vrr1_I(10, 2,
356 one_over_2p, a_over_p, aop_PQ, P_PA,
357 PRIM_INT__m_s_s_s, PRIM_INT__l_s_s_s, PRIM_INT__n_s_s_s);
358
359
360 ostei_general_vrr1_I(11, 1,
361 one_over_2p, a_over_p, aop_PQ, P_PA,
362 PRIM_INT__n_s_s_s, PRIM_INT__m_s_s_s, PRIM_INT__o_s_s_s);
363
364
365
366
367 ////////////////////////////////////
368 // Accumulate contracted integrals
369 ////////////////////////////////////
370 if(lastoffset == 0)
371 {
372 contract_all(28, PRIM_INT__i_s_s_s, PRIM_PTR_INT__i_s_s_s);
373 contract_all(36, PRIM_INT__k_s_s_s, PRIM_PTR_INT__k_s_s_s);
374 contract_all(45, PRIM_INT__l_s_s_s, PRIM_PTR_INT__l_s_s_s);
375 contract_all(55, PRIM_INT__m_s_s_s, PRIM_PTR_INT__m_s_s_s);
376 contract_all(66, PRIM_INT__n_s_s_s, PRIM_PTR_INT__n_s_s_s);
377 contract_all(78, PRIM_INT__o_s_s_s, PRIM_PTR_INT__o_s_s_s);
378 }
379 else
380 {
381 contract(28, shelloffsets, PRIM_INT__i_s_s_s, PRIM_PTR_INT__i_s_s_s);
382 contract(36, shelloffsets, PRIM_INT__k_s_s_s, PRIM_PTR_INT__k_s_s_s);
383 contract(45, shelloffsets, PRIM_INT__l_s_s_s, PRIM_PTR_INT__l_s_s_s);
384 contract(55, shelloffsets, PRIM_INT__m_s_s_s, PRIM_PTR_INT__m_s_s_s);
385 contract(66, shelloffsets, PRIM_INT__n_s_s_s, PRIM_PTR_INT__n_s_s_s);
386 contract(78, shelloffsets, PRIM_INT__o_s_s_s, PRIM_PTR_INT__o_s_s_s);
387 PRIM_PTR_INT__i_s_s_s += lastoffset*28;
388 PRIM_PTR_INT__k_s_s_s += lastoffset*36;
389 PRIM_PTR_INT__l_s_s_s += lastoffset*45;
390 PRIM_PTR_INT__m_s_s_s += lastoffset*55;
391 PRIM_PTR_INT__n_s_s_s += lastoffset*66;
392 PRIM_PTR_INT__o_s_s_s += lastoffset*78;
393 }
394
395 } // close loop over j
396 } // close loop over i
397
398 //Advance to the next batch
399 jstart = SIMINT_SIMD_ROUND(jend);
400
401 //////////////////////////////////////////////
402 // Contracted integrals: Horizontal recurrance
403 //////////////////////////////////////////////
404
405
406 const double hAB[3] = { P.AB_x[ab], P.AB_y[ab], P.AB_z[ab] };
407
408
409 for(abcd = 0; abcd < nshellbatch; ++abcd, ++real_abcd)
410 {
411
412 // set up HRR pointers
413 double const * restrict HRR_INT__i_s_s_s = INT__i_s_s_s + abcd * 28;
414 double const * restrict HRR_INT__k_s_s_s = INT__k_s_s_s + abcd * 36;
415 double const * restrict HRR_INT__l_s_s_s = INT__l_s_s_s + abcd * 45;
416 double const * restrict HRR_INT__m_s_s_s = INT__m_s_s_s + abcd * 55;
417 double const * restrict HRR_INT__n_s_s_s = INT__n_s_s_s + abcd * 66;
418 double const * restrict HRR_INT__o_s_s_s = INT__o_s_s_s + abcd * 78;
419 double * restrict HRR_INT__i_h_s_s = INT__i_h_s_s + real_abcd * 588;
420
421 // form INT__i_p_s_s
422 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);
423
424 // form INT__k_p_s_s
425 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);
426
427 // form INT__l_p_s_s
428 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);
429
430 // form INT__m_p_s_s
431 ostei_general_hrr_J(9, 1, 0, 0, hAB, HRR_INT__n_s_s_s, HRR_INT__m_s_s_s, HRR_INT__m_p_s_s);
432
433 // form INT__n_p_s_s
434 ostei_general_hrr_J(10, 1, 0, 0, hAB, HRR_INT__o_s_s_s, HRR_INT__n_s_s_s, HRR_INT__n_p_s_s);
435
436 // form INT__i_d_s_s
437 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);
438
439 // form INT__k_d_s_s
440 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);
441
442 // form INT__l_d_s_s
443 ostei_general_hrr_J(8, 2, 0, 0, hAB, HRR_INT__m_p_s_s, HRR_INT__l_p_s_s, HRR_INT__l_d_s_s);
444
445 // form INT__m_d_s_s
446 ostei_general_hrr_J(9, 2, 0, 0, hAB, HRR_INT__n_p_s_s, HRR_INT__m_p_s_s, HRR_INT__m_d_s_s);
447
448 // form INT__i_f_s_s
449 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);
450
451 // form INT__k_f_s_s
452 ostei_general_hrr_J(7, 3, 0, 0, hAB, HRR_INT__l_d_s_s, HRR_INT__k_d_s_s, HRR_INT__k_f_s_s);
453
454 // form INT__l_f_s_s
455 ostei_general_hrr_J(8, 3, 0, 0, hAB, HRR_INT__m_d_s_s, HRR_INT__l_d_s_s, HRR_INT__l_f_s_s);
456
457 // form INT__i_g_s_s
458 ostei_general_hrr_J(6, 4, 0, 0, hAB, HRR_INT__k_f_s_s, HRR_INT__i_f_s_s, HRR_INT__i_g_s_s);
459
460 // form INT__k_g_s_s
461 ostei_general_hrr_J(7, 4, 0, 0, hAB, HRR_INT__l_f_s_s, HRR_INT__k_f_s_s, HRR_INT__k_g_s_s);
462
463 // form INT__i_h_s_s
464 ostei_general_hrr_J(6, 5, 0, 0, hAB, HRR_INT__k_g_s_s, HRR_INT__i_g_s_s, HRR_INT__i_h_s_s);
465
466
467 } // close HRR loop
468
469
470 } // close loop cdbatch
471
472 istart = iend;
473 } // close loop over ab
474
475 return P.nshell12_clip * Q.nshell12_clip;
476 }
477
ostei_h_i_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_i_s_s)478 int ostei_h_i_s_s(struct simint_multi_shellpair const P,
479 struct simint_multi_shellpair const Q,
480 double screen_tol,
481 double * const restrict work,
482 double * const restrict INT__h_i_s_s)
483 {
484 double P_AB[3*P.nshell12];
485 struct simint_multi_shellpair P_tmp = P;
486 P_tmp.PA_x = P.PB_x; P_tmp.PA_y = P.PB_y; P_tmp.PA_z = P.PB_z;
487 P_tmp.PB_x = P.PA_x; P_tmp.PB_y = P.PA_y; P_tmp.PB_z = P.PA_z;
488 P_tmp.AB_x = P_AB;
489 P_tmp.AB_y = P_AB + P.nshell12;
490 P_tmp.AB_z = P_AB + 2*P.nshell12;
491
492 for(int i = 0; i < P.nshell12; i++)
493 {
494 P_tmp.AB_x[i] = -P.AB_x[i];
495 P_tmp.AB_y[i] = -P.AB_y[i];
496 P_tmp.AB_z[i] = -P.AB_z[i];
497 }
498
499 int ret = ostei_i_h_s_s(P_tmp, Q, screen_tol, work, INT__h_i_s_s);
500 double buffer[588] SIMINT_ALIGN_ARRAY_DBL;
501
502 for(int q = 0; q < ret; q++)
503 {
504 int idx = 0;
505 for(int a = 0; a < 21; ++a)
506 for(int b = 0; b < 28; ++b)
507 for(int c = 0; c < 1; ++c)
508 for(int d = 0; d < 1; ++d)
509 buffer[idx++] = INT__h_i_s_s[q*588+b*21+a*1+c*1+d];
510
511 memcpy(INT__h_i_s_s+q*588, buffer, 588*sizeof(double));
512 }
513
514 return ret;
515 }
516
517